Join 3,512 readers in helping fund MetaFilter (Hide)


And a 1 and a 2, a 1, 2, 3, 3.984
July 23, 2012 7:35 AM   Subscribe

"People prefer music that deviates from perfection in a natural way." Researchers into rhythm are trying to figure out the nature of these deviations, and what implications this has for audio engineering and neuroscience.
posted by EvaDestruction (50 comments total) 36 users marked this as a favorite

 
It rather depends on what you consider "perfection" to be, no? Audio engineers have been introducing distortion to recordings on purpose for as long as there was tape.
posted by LogicalDash at 7:42 AM on July 23, 2012


The discovery that human errors in musical rhythm follow a pattern could influence how audio engineers “humanize” computer-generated music.

To even call them "errors" in the first place is, you know... wrong.

"People prefer music that deviates from perfection in a natural way."

What people prefer, then is perfection. The "natural" is perfection. Metronomic time is just metronomic time.

OK, I'm outta here. Gonna go listen to The Meters now.
posted by flapjax at midnite at 7:49 AM on July 23, 2012 [11 favorites]


Oh, but... good post title, EvaDestruction!
posted by flapjax at midnite at 7:50 AM on July 23, 2012 [1 favorite]


FWIW, Physics Today isn't a peer-reviewed journal, but a sort of compilation of research from members of the various societies within American Institute of Physics. And here's the PLoS ONE article.
posted by knile at 7:58 AM on July 23, 2012


Me loves me some Physics Today every month.
posted by BentFranklin at 7:59 AM on July 23, 2012


The expectation that a performer would keep mechanical time is a byproduct of using drum machines. For people doing music that isn't derived from that mechanical tempo (everything from classical to swing to blues), making a machine play a rhythm well is still an unsolved problem (not to mention getting the machine to follow a rhythm set by a human).
posted by idiopath at 8:00 AM on July 23, 2012 [1 favorite]


Physics Today isn't a peer-reviewed journal

True, but it is written by scientists for scientists, so they don't have to review all of quantum mechanics for every article, and they don't apologize for having equations.
posted by BentFranklin at 8:03 AM on July 23, 2012


Me loves me some Physics Today every month.

This statement hurts my brain.
posted by Mister Fabulous at 8:05 AM on July 23, 2012 [4 favorites]


I wonder if the writer knows that musicians are perfectly aware of this, and are by no means even attempting, in most cases, to play exactly on the metronomic beat?
posted by thelonius at 8:08 AM on July 23, 2012 [2 favorites]


The amazing thing about tuning is, mathematically, there is no possibility of perfection. Fundamental theorem of arithmetic. Octaves (2s) are incommensurate with fifths (3s) and major thirds (5s). Everything is a compromise.
posted by iotic at 8:10 AM on July 23, 2012 [6 favorites]


This is why you shouldn't quantize everything. In most pop music, live drums are usually quantized and every vocal track is auto-tuned (even the ones that sound natural). It sounds mechanical and boring.
posted by bhnyc at 8:20 AM on July 23, 2012


You know what else I hate? People doing research on optics. Don't talk to me about "breaking light up into its constituent colors," man! The rainbow is not "broken"! It's beautiful! If a bunch of eggheads can't see that, then they obviously have nothing of value to say. Sincerely, the Sturm und Drang movement.
posted by No-sword at 8:32 AM on July 23, 2012 [9 favorites]


This is why Ringo Starr was so good for the Beatles. His lack of metronomic "perfection" gave his drumming the kind of personality that a drum machine can never, ever replicate. Listen to "Don't Pass Me By" for a good example.
posted by Faint of Butt at 8:36 AM on July 23, 2012


Science meets pocket and soul. Good times!
posted by Ice Cream Socialist at 8:37 AM on July 23, 2012


I think this is probably understood by people who do lots of computer-based composition.

When I write a MIDI drumbeat (or any live sounding instrument for that matter), I usually quantize until the final mix (more for information management and reducing mistakes than anything else), at which point I do a rubato pass (deterministic randomness following a swing feel) and then add in a random time and velocity delta (to add some human error noise), and then hand-tweak time and velocity one final time so I don't hear any machine gunning or awkward hits. I think most composers who are trying to use live sounding samples have their own version of this.
posted by hanoixan at 8:39 AM on July 23, 2012 [4 favorites]


I'll play devil's advocate here and suggest that Kraftwerk's absolute precision has always felt otherworldly and perfect to me. They conjure up a sound that seems non-human, which is quite an artistic achievement. That illusion would be completely ruined by deviations, I think. But, maybe they are incorporating tiny variations that are stealthily working me over. Whatever they're up to, it works - I've heard tons of electronic music and nobody else seems to be able to create this man/machine sound environment quite like them.
posted by davebush at 8:43 AM on July 23, 2012


The sound of Buddy Rich turning in his grave. In perfect time.
posted by hal9k at 8:45 AM on July 23, 2012 [1 favorite]


I wonder if the writer knows that musicians are perfectly aware of this, and are by no means even attempting, in most cases, to play exactly on the metronomic beat?

Those really aren't the kinds of variations that these guys are researching or describing. The tiny variations they are finding are being produced under laboratory conditions where they are asking professional musicians to keep a metronomic time. They are also finding that they vary systematically across timescales that the musician is simply not consciously aware of and that auditors, similarly, would not be consciously aware of.

In my experience this is one of those issues on which informed commentators are hopelessly inconsistent. I'm interested in the comment above, for example, about Ringo and that his "imperfection" was a humanizing influence on the Beatles' sound. I'm very prepared to believe that--but at the same time I have found innumerable comments from people with impressive credentials to the effect that while Ringo's drumming was never very complex, the one thing he did provide for the Beatles was an astonishingly metronomically solid and consistent beat.

I have never seen an internet argument about a drummer's imperfections (or perfections) in fact not be utterly inconsistent (at least, over time) on this point. Lots of drummers will--pace what is said above about "every" drummer deliberately avoiding metronomic consistency--praise a drummer's ability to keep "perfect" time. By the same token, lots will savage a drummer for being "just like a drum machine." The oddest thing, to me, is that the same drummers will come in for both kinds of criticism--that they "can't even keep time" and that they are "interchangeable with a drum machine." I can't count the number of times I've read that Meg White "can't keep time" and is hopelessly ahead or behind the beat. And yet when the criticism of White first got going it was the opposite: that she was simply boring--like a drum machine. In fact, it was widely alleged that they did use a drum machine on the CDs.

I don't claim to know anything at all about drumming--but I can tell from the flatly contradictory things said about it by practitioners that it badly needs more scientific analysis of the kind provided in this FPP.
posted by yoink at 8:55 AM on July 23, 2012 [11 favorites]


There's a Frank Sinatra song (can't recall which) that's a great example of how musicians have been doing this purposefully for a long while. During the verses he sings sharp staccato, just a little bit ahead of the beat, that makes that section driving and energetic. Then during the choruses he sings softer and longer, just a tiny bit behind the beat, giving that section a lazy feel. The tempo doesn't change, but the expression of it does, and it makes the one sound faster than the other.
posted by echo target at 9:09 AM on July 23, 2012


Ray Charles, accurate to within 2.5 milliseconds. Of course, his ability to stay behind that beat is what makes Ray so great vocally. I've been forced to listen to so many tedious discusssions about what "pocket" means, I welcome scientific inquiry into this fascinating area.
posted by Lorin at 9:13 AM on July 23, 2012


There's a Frank Sinatra song (can't recall which) that's a great example of how musicians have been doing this purposefully for a long while.

Again, that's not what they're talking about. They're talking about musicians trying their hardest to hit the beat dead on and missing it by very small amounts of which they are quite unconscious. They're not talking about the deliberate choice to sing behind or ahead of the beat.
posted by yoink at 9:16 AM on July 23, 2012


Fuckin' beats; how do they work?

It's magic, yo.
posted by ZenMasterThis at 9:19 AM on July 23, 2012


Dropped in to make a comment about Bonzo droppin that groove waaaaay back.... then realized that I, like pretty much every other commenter except Yoink, hadn't RTFA.
posted by armoir from antproof case at 9:20 AM on July 23, 2012


Again, that's not what they're talking about. They're talking about musicians trying their hardest to hit the beat dead on and missing it by very small amounts of which they are quite unconscious. They're not talking about the deliberate choice to sing behind or ahead of the beat.

I can't really find anything in the article to suggest that they're familiar with rhythmic feel and are controlling for it. Ten to twenty milliseconds is absolutely in the neighborhood of what's sufficient to produce an audible change in feel.
posted by invitapriore at 9:26 AM on July 23, 2012


Wait a minute, did they have the drummers just play steady quarter notes for five minutes? It's not really clear in the article. I thought they were improvising patterns for a while. If they were really just trying to match the metronome, it changes the interpretation a bit.
posted by echo target at 9:30 AM on July 23, 2012


flapjax at midnite: The discovery that human errors in musical rhythm follow a pattern could influence how audio engineers “humanize” computer-generated music.

To even call them "errors" in the first place is, you know... wrong.

"People prefer music that deviates from perfection in a natural way."

What people prefer, then is perfection. The "natural" is perfection. Metronomic time is just metronomic time.
Half-right. To call "natural" meter perfection is to create a circular argument: it's perfect because whatever it is, is perfect.

But not all metric errors are in fact errors; that is true. Many are. Some are not (see: Billie Holiday's entire opus).
posted by IAmBroom at 9:33 AM on July 23, 2012


Yeah, no, they just asked the drummer to imitate a metronome. Figure 1 in the PLoS article makes that clear.
posted by nebulawindphone at 9:39 AM on July 23, 2012


Brian Eno:
«I think we’re sort of deep in the grid period of making music— well, we’re probably emerging from it a little bit now, I would say. You know how eras always have a sound to them and you don’t realize it until the era has gone? I remember when in the early days of rock’n’roll, when everything sounded totally different, all amazing and blah blah blah blah blah. Now you can play me one second of any record from that time, and I’ll say “1959” or “1961.” I can hear precisely. It’s like it has a huge date stamp on it. And I think we’re all capable of doing that. You can hear the profile of a sound, in retrospect, so much more clearly than you did at the time. And I think one of the things that’s going to be nauseatingly characteristic about so much music of now is its glossy production values and its griddedness, the tightness of the way everything is locked together.

I just got an amazing 10-CD set, it’s the music that Alan Lomax recorded in Haiti in 1936. And what’s incredible is how fantastic the drummers are and how off-the-grid they are. The liveliness is astonishing; they’re just totally alive, these recordings. It’s very interesting, to me, to be reminded of that, that there was a time when things were not that tight. And we’re going through this super-uptight era, which I think comes entirely from literacy, actually. It’s the result of machines that were designed as word processors being used for making music. Because that’s what we’re doing, after all. All the programs we’re using started their lives, really, as word processing programs and the concepts that typify word processing, like “cut and paste,” “change typeface.”»

posted by procrastinator at 9:51 AM on July 23, 2012 [6 favorites]


Well, I'm busted - I didn't read the article very carefully.
posted by thelonius at 10:01 AM on July 23, 2012


The artist BT has been known to study how imperfect rhythms affect the mind. He uses the golden ratio in many ways, sometimes to offset what would be a rhythm on an otherwise perfect grid. 1.618 is one such track.
posted by hellphish at 10:04 AM on July 23, 2012 [1 favorite]


"There was once a word used---swing. Swing went in one direction, it was linear, and everything had to be played with an obvious pulse and that's very restrictive. But I use the term 'rotary perception.' If you get a mental picture of the beat existing within a circle, you're more free to improvise. People used to think the notes had to fall on the center of the beats in the bar at intervals like a metronome, with three or four men in the rhythm section accenting the same pulse. That's like parade music or dance music. But imagine a circle surrounding each beat-each guy can play his notes anywhere in that circle, and it gives him a feeling he has more space. The notes fall anywhere inside the circle, but the original feeling for the beat isn't changed. If one in the group loses confidence, somebody hits the beat again. The pulse is inside you. When you're playing with musicians who think this way you can do anything. Anybody can stop and let the others go on. It's called strolling…."

---Charles Mingus, Beneath the Underdog
posted by TwoToneRow at 10:07 AM on July 23, 2012 [4 favorites]


Okay, I think I understand what's going on here.

So no drummer ever plays totally square. Start there.

In a real performance, there are two sources of "unsquareness." First: if he's good, the drummer is aiming for a rhythm that isn't totally square. But second: even compared to the rhythm he's aiming for, the drummer still strays a little bit ahead or behind, rather than perfectly nailing what he intends to play. The first set of deviations from square are determined by the drummer's skill and stylistic choices, and are in some sense deliberate. The second set are caused by the imperfection of his human body and brain; they're unintentional, totally random motor errors. Let's call the first thing "style" and the second thing "error," just to keep them straight.

In this study, they filtered out the effects of style by asking people to play like a metronome. They were just looking at genuine random error.

Here's where it gets interesting. For a human player, even their error still seems to follow a certain pattern. To be precise: it forms a time series with some long-range correlation, like the Brownian motion of a dust speck in a drop of water. If the dust speck was way over at the left side of the water drop at n milliseconds, then you know where to look for it at n+1 milliseconds — It'll be somewhere over to the left side still. Maybe it'll be a little further to the left than it used to be, maybe it'll be a little less far, but it won't just vanish and reappear way over to the right. So too, they found, with the error of a human drummer. If the guy is way ahead of the metronome on beat n, odds are he'll still be ahead of the metronome on beat n+1 — maybe a little further ahead, maybe a less far ahead, but he's unlikely to jump from way ahead to way behind.

Long story short, we should think of the drummer's error as wandering ahead of or behind the the rhythm that he's aiming for, like a speck of dust or a drunk guy staggering around, rather than blipping around completely unpredictably like the specks of static on an old-fashioned TV.

Here's where it gets really interesting. Computer-generated music, by default, has no error. You can program in a square beat, and it'll come out perfectly square. You can program in a beat with some style — "put the second beat of the bar 50 milliseconds late" — and by golly that beat will be 50 milliseconds late every time. But if you want, you can also add a bit of simulated error to the computer-generated beat. (In some sense it's not real error, because you told the computer to put it there. But it's like human error in the sense that it's outside the musician's control.) This is sometimes called "humanizing" the beat, and in moderation it tends to make electronic music less dull and more ear-catching.

So the question is, how should we go about adding simulated error to our computer-generated beat? Well, there are two obvious possibilities. You can throw in some uncorrelated randomness. (For each note, roll a die, and move it forward that many milliseconds.) Or you can use random-walk-style randomness that has some long-range correlation to it. (Store a number representing how far "ahead" or "behind" we are. For each note, flip a coin to decide whether we should wander one ms further ahead than we were before, or one ms further behind.) The first possibility is computationally simpler. The second possibility is closer to what a human drummer would do. So they checked, and it turns out, people prefer the second way of doing it — the way that follows the pattern of human error, gradually wandering ahead or behind the beat.


Tl;dr: We already knew that people preferred human style to robotic style. But now we know that people prefer human-like mistakes to other kinds of mistakes. And I think that's pretty cool.
posted by nebulawindphone at 10:44 AM on July 23, 2012 [22 favorites]


I'm drastically oversimplifying "style" up there just to keep things clear. It also includes one-time deliberate rhythmic shifts: "I'm gonna play this here note just a hair behind the beat, because it sounds better that way."

And then there's the possibility that a high error rate could itself count as a kind of style. When Mo Tucker was playing Heroin I'm betting she didn't choose which notes were going to be early or late. But she did choose not to be all uptight about the whole thing — and that choice mattered. Replacing her with Clyde Stubbelfield would have been unhelpful.

And then there's the stories about people like Frank Zappa and Captain Beefheart cultivating this deliberate, controlled style that just sounded like the notes were all randomly shoved off the beat this way and that. Like if he'd wanted to, Drumbo from the Magic Band could have repeated the drum part from Heroin note-for-note and gotten it all exactly the same every single time. And who knows, maybe he could have.

Whatever. Screw this. Let's go dance.

posted by nebulawindphone at 10:58 AM on July 23, 2012 [1 favorite]


Didn't Jeff Bilmes establish this about ten years ago?

He made high-quality isolated recordings of a Cuban percussion group, extracted the timings, then resynthesized it with (a) quantized timing, (b) quantized timing + random error, and (c) original timing. The results were, respectively: (a) cold, (b) sloppy, and (c) funky.
posted by nixt at 11:06 AM on July 23, 2012


No, as I'm reading it, what they did here was (b1) quantized timing + human-like random error, (b2) quantized timing + non-human-like random error, and found that people prefer (b1) to (b2).

Of course, (c) good funky drumming would be better than either, but that's old news.
posted by nebulawindphone at 11:11 AM on July 23, 2012


Even knowing this is not 100% on point, which seems somehow more appropriate that it is not, while reading I am reminded of the live and cuddly version of metronome by nomeansno.
posted by safetyfork at 11:32 AM on July 23, 2012


The amazing thing about tuning is, mathematically, there is no possibility of perfection. Fundamental theorem of arithmetic. Octaves (2s) are incommensurate with fifths (3s) and major thirds (5s). Everything is a compromise.

What?! You can totally create a just intonation scale that has perfect exact ratios for major and minor thirds, fourths, fifths, etc. The only problem is that if you tune your instrument to such a scale you can only play in that one key. Equal temperament is a compromise that sacrifices perfection of the intervals for the ability to play in any key. If you're using an instrument that is infinitely tunable, such as the human voice, then that's not a concern and you can always play in just intonation, which is why the chords sung by barbershop quartets always sound so perfect.
posted by Rhomboid at 11:33 AM on July 23, 2012 [1 favorite]


nebulawindphone, I think Bilmes followed up his analysis work with methods for adding human-like variation to quantized drum patterns, which would be the (b2) that you're describing. But it looks like the authors of the PLoS ONE article (and associated patent) only reference publications by other physicists.
posted by nixt at 11:50 AM on July 23, 2012


Yeah, then you may be right. I don't know this literature at all, I just dig music and can speak statistics.
posted by nebulawindphone at 12:21 PM on July 23, 2012


people prefer music that deviates from perfection in a natural way: “a mix of predictability and surprise,”

What is "perfection"? What is "natural"? As for surprise, guess why Haydn's "Surprise" Symphony was called that. In the 1790s.

Physicists have discovered these misbeats follow a pattern.

This is an article from Harvard Science. No less. Seems they've finally caught up with what any halfways-talented teen musician already knows. "Humanization" was a failure a quarter-century ago. Maybe Harvard "physicists" should leave studies of music to people who actually know something already.

This is more "fluffy" science journalism (most of it is) about "studies" which "prove" the obvious, written by a sloppy writer, with utterly nothing new to say.
posted by Twang at 12:32 PM on July 23, 2012


What?! You can totally create a just intonation scale that has perfect exact ratios for major and minor thirds, fourths, fifths, etc.

Yes, but if you then play say, a major second to a major sixth, you get a horribly dissonant fifth.

What I have recently discovered, as I am making an app simulating sympathetic strings for practicing Indian music, is that the just scale is not even used in Indian music. If you listen to ragas played by master musicians, you will find the thirds and sixths are pretty far from those in the just scale, even though plenty of articles will tell you the just scale is used for Indian music.

Note that this is for melodic, rather than harmonic reasons. Even against a drone and with no chords, the just scale sounds wrong. You still have to compromise.

I also wouldn't assume that barbershop quartets, string quartets etc tune to just intervals - although doubtless they don't temper the thirds and sixths to as extreme a degree as equal temperament. It would sound odd to be shifting the notes constantly in response to the harmony.
posted by iotic at 12:33 PM on July 23, 2012


What?! You can totally create a just intonation scale that has perfect exact ratios for major and minor thirds, fourths, fifths, etc. The only problem is that if you tune your instrument to such a scale you can only play in that one key.

Well, there are actually kinds of music that are unplayable in JI.

Simple but unrealistic example: if you want to be able to modulate all the way around the circle of fifths and end up "back where you started," you'll need to be playing in some sort of temperment, because in JI you'll end up a Pythagorean comma off.

Slightly more plausible example: if you want to be able to go up three major thirds and end up an octave away from where you started, you'll need to be playing in some sort of temperment, because in JI you'll end up out of tune by a Diesis. There's a lot of late-Romantic and 20th-century music where it matters that the octave can be symmetrically divided into major thirds, and breaking that symmetry would make things go haywire.

Equal temperments in particular have this sort of M.C. Escher staircase quality where you can set off in any direction and arrive back where you came from without ever turning around. In 12-equal there's a circle of fifths, there are "circles of thirds," and "circles of minor thirds," there are "circles of whole steps," everything is circular. You can't do that in JI. Eventually, you have to turn around in order to get home. Classical music works great in JI. But in the Romantic era they started playing with the Escher Staircase Trick every once in a while, writing stuff that JI can't handle; and by the time you get to Wagner and Scriabin and Debussy, they were just rolling around in it like a dog in fresh non-Euclidian shit. Trying to play Debussy in JI would be missing the point entirely.
posted by nebulawindphone at 12:45 PM on July 23, 2012 [3 favorites]


Kraftwerk's absolute precision

That's because they used electronic drum kits/pads rather than a drum machine. Computer World was their first album to use a drum machine. You were hearing a person hit something.
posted by Potomac Avenue at 2:50 PM on July 23, 2012 [1 favorite]


Participatory discrepancies for err'body!
posted by LMGM at 3:59 PM on July 23, 2012 [1 favorite]


iotic: I think it's best to treat any musicologist or musician's claims about what they are doing or why it works, particularly as it pertains to tuning, as bullshit unless empirically verifiable evidence otherwise is presented.

String quartets and barbershop quartets do not gravitate to just intonation; irrational intervals do not always sound unpleasant, and are not the cause of dissonance; the harmonic series is not a property of most natural sounds, it is a special case of vibration where the vibrating object approximates a one dimensional column... just off the top of my head.
posted by idiopath at 4:16 PM on July 23, 2012 [2 favorites]


I wonder if the writer knows that musicians are perfectly aware of this, and are by no means even attempting, in most cases, to play exactly on the metronomic beat?

This is what occurred to me with the Ringo examples.

Somewhere online (I will find it) is an analysis of how the drumming on Dizzy Miss Lizzy speeds up in at the points of highest tension in the song - and given that 'tension' in music is a constantly shifting, moving-target mixture of surface and structural elements (for example, a moment like the conjunction of the singer screaming just before the guitar solo plus the harmonic structure reaching an anticipated point, plus an instrument like the bass springing out of its usual register), it is pretty impossible for a machine to know where it falls. So, machines can't drum.

I also read somewhere an interesting speculation that because in the womb we hear our own heartbeat randomly overlaid with that of our mother's, it makes us develop a liking for irregular pulses (and how regular are our hearts anyway?). Probably nonsense, but I found myself thinking about it.
posted by colie at 12:03 AM on July 24, 2012


I wonder if the writer knows that musicians are perfectly aware of this, and are by no means even attempting, in most cases, to play exactly on the metronomic beat?

I'd say most musicians are indeed perfectly aware of this. And playing drums, for example to a click track really kinda sucks. Some people get really pretty good at it, cause they do session work all the time that requires them to play to a click. Myself, I really hate playing to a click (though on occasion I have to, if I wanna get paid).
posted by flapjax at midnite at 3:53 AM on July 24, 2012


And playing drums, for example to a click track really kinda sucks.

Here's a good link where a guy breaks out the bar graphs to show where songs speed up and slow down if they're played by real humans (the Kraftwerk example indicates that at least some of their drum tracks were in fact played by machines).

The graphs are great, especially Sympathy for the Devil - and Won't Get Fooled Again, which is much-loved by drummers but was famously recorded to a click track. Also has the link to the Dizzy Miss Lizzy micro-tempo changes, and shows how Metallica don't use a click but Nickleback (officially the World's Worst Band) do.

Link.

And I know we've already had the Beatles-Godwin that is essential to all music threads, but it does appear that A Day in the Life might be the first song recorded (partially) with a click track.
posted by colie at 11:06 AM on July 24, 2012 [1 favorite]


Hi, this is kind of my field of expertise -- one of the professors I've worked for in the past few years has done a lot of work in quantifying expressive timing. I have mixed feelings about this article; on the one hand, it's completely bizarre from a musician's perspective to call these deviations "errors." They're clearly not mistakes, and they're clearly not random, and every musician I know plays around with rhythm in a very conscious way. On the other hand, I think this kind of research is really interesting and potentially revealing and I hope it continues.

One of my favorite papers on the subject is Friberg and Sundström's Swing Ratios and Ensemble Timing in Jazz Performance. It's one of the rare times where a scientific study of music has come up with a non-trivial, surprising, insightful and completely counterintuitive result. I'll try and explain it succinctly.

So, if you listen to a lot of classic jazz and bebop recordings -- John Coltrane, Miles Davis, Keith Jarrett, etc. -- you might begin to notice that while everyone in the ensemble is playing with swing, they're all playing really differently. Drummers tend to play with really heavy swing, while the soloists tend to play with the least swing, sometimes with almost no swing at all. If you think about the technical demands of the instrument this kind of makes sense; at the blazingly fast tempos they're playing at, it would be hard to keep a consistent heavy swing, especially while improvising.

So then how can they possibly play together, if they're all deviating wildly from each other? You'd think that they'd fall apart almost instantly. Well, you might guess that they sync up on the downbeats despite playing different upbeats, and this would be a reasonable guess. But in fact the opposite is true. Their downbeats are all over the place, and it's their UPBEATS that sync up. (!!!)

If you think about the simultaneous precision and looseness that this kind of thing requires, it's pretty mindbending, but for those players it's internalized to a point that it's practically effortless. In fact it's so internalized that no experienced jazz performer would describe it the way I just did -- it would be impossible to teach that way. Some players will talk about playing "behind the beat" as a particular sort of desired feel, and if you think about it, that's exactly what they're doing; playing behind on the downbeats and catching up on the upbeats. But "catching up on the upbeats" is a totally foreign concept for jazz players.
posted by speicus at 6:34 PM on July 24, 2012 [7 favorites]


This blog post I bookmarked a while back (full of broken links I'm afraid) is a great example of “yes, musicians consciously and repeatably deviate from a metronomic beat”: http://rumbaclave.blogspot.com/

I would dispute whether they’ve really corrected for human rhythmic feel, as far as I can understand it from the Physics Today piece. They say “one can be forgiven for asking whether LRCs persist when musicians play more complex rhythms such as those in contemporary pop and rock music. It turns out that LRCs are inherent in all sorts of complex rhythms...”, but there’s no discussion of how to extend the analysis to irregular rhythms.

One possible explanation of a long-range correlation in errors is that you’re measuring the error relative to the wrong thing. For instance, the unnamed drummer from Ghana could be extremely sensitive to the local timing between beats, but due to his or her experience playing in ensembles, comfortable with anticipating or lagging behind the metronome. So maybe a better model here is a slowly varying phase relative to the metronome, with additional random errors. (Or what speicus just said: if you asked a sax player to solo in time with a drummer who’s really swinging, there might be a hell of a lot of long-range correlations in the deviation from the drummer.)

Or maybe that is all to say, non-gaussian errors are usually just data crying out for a better model.
posted by mubba at 7:25 PM on July 24, 2012


« Older Bohemian Rhapsody. Piano. Drum. Bass. Vocal Choir...  |  Yesterday I had the dubious pl... Newer »


This thread has been archived and is closed to new comments