JW: The idea has precedent; acoustical presentation of data was attempted in the EEG domain, and Mayer-Kress, a guy who did a lot in non-linear dynamics of the brain in Santa Fe, made a lot of hype around this. Quite independently, a Czech neurologist few years ago tried the same general approach. Concerning transformation of Z-scores to sound, the following seems to work quite well: k = gamma * log( 2 ) f = f0 * exp( k * z ) where f0 is the frequency assigned to Z=0, and gamma is the steepness parameter; with gamma = 1, difference in Z of +-1 is translated to doubled or halved base frequency (1 SD = 1 octave).

RDN: A chord whose notes are determined by an algorithm applied to the current datastreams and various conversions of it. Background hum that is very soft when the composite across all sites is very little different from no structure, and increasing volume with indications of departures from expectation. The dominant notes are the clusters of Eggs that are most correlated, and the complexity of the chord in time is a function of the level of global resonance. Rhythms are determined by the occurrence of warm and cool localized areas or global patterns, where there is much more or much less coherence and resonance than usual.

The heart rhythms act as a carrier wave for life and emotion and intelligence, at least in an important metaphoric sense. For Gaia, we may expect an equivalent rhythm and carrier, with an expectation it will be vibrations from a different frequency domain and in a different medium. The heartbeats of Gaia have an interval of about 5 or 10 minutes, based on abstract calculations with results between 4.8 and 8.4 minutes, e.g., the ratio of temporal perception units apparently operating in human and global consciousness, and by dimensions of human and planet represented in a volumetric ratio. The random music will need a carrier wave, and the modulated excitement in the heart of Gaia can serve.

JW: I was playing a bit with the acoustic feedback I mentioned few weeks ago. There is no sophisticated synthesis, the current deviation of the random walk trajectory is simply converted to sound, using my old formula and the PC speaker as the output device. (I hate the idea to program SoundBlaster chip myself.) Anyway, it works and seems to provide usable feedback to the operator. I decided to play W[n] = 2S[n] - n (where S[n] is the bitsum at the n-th step of the walk) rather than Z[n] = W[n] / sqrt(n), for one obvious reason: if W is constant for a while, the corresponding Z decreases, and this is somewhat counter-intuitive for a statistics-naive operator, given the instruction `"push the curve as high/low as possible." However, the conversion factor is adjusted so that at the end Z=1 corresponds to one octave, as I proposed.

These preliminary experiments gave me to think about some details: (i) Frequency coding -- I realized that it's not always easy to get the direction of the change immediately, at least with gamma=1; with gamma=2, the feedback is much more convincing, but then the range of ``audible'' values is limited, since the frequency changes by factor 4 by 1 standard deviation.

RDN: How about a non-linear mapping, where differences close to the expectation yield large changes, but those closer to the end of the range produce progressively less change. I call this the Zeno scale in the case of our newest ArtREG version, where the best the operator can achieve at any given point is progress half-way to the goal -- so it is not possible actually ever to get there.

JW: Human hearing is non-linear too, so it may pay to think of perceptual harmonics. (ii) Base tone and tempo seem to play major role in the aesthetic perception of the resulting ``music''. I currently use 640 Hz for zero deviation, and the tempo is fixed by my CCD device (12.5 bits/second); however, experiments with bitstrings playback showed that the impression may change a lot with these parameters. The main purpose of all this was to develop a feedback strategy not involving eye movements, and thus facilitating EEG recordings. Anyway, I thought it might be of some relevance to your "music of noospheres" subproject.

RDN: Most relevant. I have not found time to dig into my old Amiga files, where the source of the taped music probably resides in APL functions. It may not be very important to do so, because the material for the MON (music of noospheres :) is going to be very rich and deserves fresh and creative thinking.

JW: BTW, I took a book by John Barrow, "The Artful Universe", for my weekend bed reading, and there was a section devoted to music synthesized by random sources with varying autocorrelation function (Chapter 5, "Natural History of Sound" ?? I read the book in German translation, so I'm not sure).

RDN: Did you have the impression it was beautiful music, and something we might well use? I think we will/do have varying autocorrelation function, but usually within an expected range. However, these variations can be "amplified" and themselves as the source of parameter control for the synthetic music.