[sdiy] harmonics & vibrato - drums

[Chris Stecker]

Grant (and interested list),

Although "masking" is not the term I would necessarily use (it has a more 
specific, but related, meaning), there are many examples of adaptation in the 
nervous system, including (as Mr. Gravenhorst noted) fatigue of sensory 
receptors, and also lateral inhibition (stimulated regions of the retina 
suppress activity in neighboring cells, producing "Mach bands" or bright 
lines next to black/white boundaries). There are higher-order adaptive 
phenomena at many later stages of neural processing, that adapt to more 
complex or abstract aspects of stimuli. Some of these produce observable 
effects like afterimages, and some are more subtle or immediate. In general, 
adaptation helps to reduce the response to steady-state (continuous) 
stimulation, so that dynamic or transient stimuli are easier to detect. It's 
a bit like high-pass filtering or taking the derivative of any parameter 
you're interested in (across time, frequency, space, or whatever); changes 
become more noticeable, and the system's "dynamic range" is increased.

Lateral inhibition in the auditory system occurs between neighboring 
frequencies (as evidenced by "two-tone suppression" as well as various 
simultaneous masking phenomena), and also over time, as you suggested. 
Forward-masking (the reduced detectability of sounds following exposure to a 
masking--generally more intense--sound) is one example, and so is the 
precedence or Haas effect (reduced ability to localize or spatially segregate 
sounds following onsets). These phenomena and others are characterized by a 
limited duration of suppression (usually they have some measurable "recovery 
function").  [My specific research interest in this area is adaptation to the 
acoustics of rooms.]

What you're suggesting for drumming, however, is difficult for me to 
understand, primarily because we (I, especially) don't yet understand how 
rhythm, timing, and repetition are represented or processed in the brain. 
This theory would require a representation of intervals, whereby "units" 
representing particular intervals would fatique if repetitively stimulated. 
Jittering the intervals would spread activation around a small group of such 
units, causing less fatigue. It's not a totally unreasonable idea, and in 
fact at much shorter time scales (intervals around 5-10 ms) randomly varying 
the intervals (by a very large "jitter") reduces the strength of the 
precedence effect...but that's really a totally different time scale, a 
different perceptual dimension (space), and presumably a totally different 
type of mechanism (there is no precedence effect for intervals beyond 10 ms 
[clicks] to 50-100 ms [speech or music]). I don't know of any data suggesting 
that the perceptibility of sound is actually diminished by highly regular 
repetition, but it's possible that part of the "appeal" of less-quantized 
rhythm is in the "spreading around" of stimulation to more than one 
exact-interval processing "unit."

Here are two other possibilities I have in mind:

a) slight randomization sounds better because it sounds "human," and it sounds 
human because that's what we've been exposed to, listening to human drummers. 
(not that children raised by robots would think hard-quantization sounds 
"human," but...) Or maybe its a bit like the difference between 
photo-realistic painting and more expressive styles. (Or is it more like the 
difference between photorealistic painting and photography?)

b) slight randomization reduces the predictability of the beat. we like 
surprises, (because we pay attention to the derivatives?) so totally 
predictable rhythms sound "flat."  This is really just another way to frame 
your "crackpot" adaptation theory, but it does not depend on a low-level 
process that suppresses sound perception, per se. Rather, it hypothesizes the 
existence of fairly high-level perceptual adaptative mechanisms that track 
regularity in an analytic-synthetic, predictive, or change-oriented manner.

My money's on "a," as it seems simpler and in keeping with variability between 
listeners; some prefer "tight" (versus "sloppy") drummers, while others 
prefer "expressive" (versus "flat" or "robotic") ones. "b" appeals more to my 
views of neural computation, however, and is more aesthetically 
appealing/interesting, but that doesn't make it right (it's probably more 
testable though). 

-Chris


On Friday 20 December 2002 01:43 pm, Grant Richter wrote:
> > Ok, this makes more sense.  But the question remains, is there a random
> > factor or not?  If not, then why does doing a *slight* randomization
> > (weighted in my case) sound better (um more "human") than a quantized
> > track?
>
> VERY Crackpot theory,
>
> Neural systems have a masking factor. This prevents a stuck input from
> tying up the whole net. Like when you stare at something and it leaves a
> negative image in your eye. The negative image is the mask generated by the
> neural system.
>
> That may apply to time related items also, like when you are in a noisy
> room and your ears adapt by "tuning" out the noise so you can hear better.
>
> It may be that drummers instinctively play around the outside of the neural
> "mask" time. If they hit perfectly on the beat every time, they eventually
> couldn't hear themselves.