05/02/2006
On April 21, our local National Public Radio outlet, WNYC, broadcast (and streamed over the Internet) an episode of the station's Radio Lab program. This slickly produced series combines aspects of NPR-style radio journalism with modern audio-studio production techniques that are the sonic equivalents of MTV-type visual effects. Much of the time I find that the high-tech machinations obscure the journalistic message. But the episode that aired on the 21st, and that is scheduled for a rebroadcast/restreaming at 2 and 3 PM EDT on May 16, should be heard by everybody interested in how the sounds hitting your eardrums are turned by your brain into music. You don't even have to wait if you don't want to, since MP3 streams of the program are available at www.radiolab.org. Look for the program titled "Musical Language."

The first segment of the three in the show quite rightly centers on Diana Deutsch, a psychology professor at the University of California in San Diego and one of the pioneering researchers in the fundamentals of music perception. Her website (http://psy.ucsd.edu/~ddeutsch/) is a great place to read, listen, and download, especially her famous work on musical illusions, one of which is illustrated here in musical notation:

You can even order a couple of CDs of musical illusions, nearly all of which are ear-openers. I recommended both of them.

Radio Lab's coverage of Deutsch's work focuses on her "discovery" that when listening to multiple repetitions of a spoken phrase — Deutsch saying "But they sometimes behave so strangely" in a recorded lecture, for example — you can actually pick out an internal musical melody in ordinary speech. This effect had already been exploited decades ago by composers like Steve Reich, whose 1966 Come Out is available from Apple's iTunes store. The melody Deutsch heard in her phrase, and which receives the full weight of Radio Lab's production values, is shown here in musical notation:

This is a good example of the hearing system's propensity to extract a "central pitch" and to create a perception of a melody from the extracted pitches even when the input to the ear is not intended or produced as music. Deutsch believes this power is partially suppressed in most Western listeners but operates, as it must, at full force in speakers of "tonal" languages, such as Chinese, where the meaning of a syllable or word can be changed by the melodic contour with which it is spoken. (Ancient Greek operated in this way, too.)

But when I first heard this demo, available for download in full at Deutsch's website, my ears heard the first note not as a D-sharp, as Deutsch would have it and as shown in the notation, but more as a C-sharp, a whole tone lower. The spectrum analyzer contained in Adobe Audition (an excellent pro-grade PC sound editing program that contains its own pitch-estimation algorithm) confirmed my impression. In order to convince you of my interpretation, and as an example of how "pre-biasing" can force you to reinterpret the sounds you hear, I've prepared my own demo containing an excerpt of Deutsch's original track (in the right channel) with my own "guide-pitch" track of pure-spectrum sine waves in the left channel. The effect of the slight initial melodic C-sharp/D-sharp curve, rather than Deutsch's repeated D-sharps, is actually more audible when heard in reverse, a ploy that disturbs the pitch-extraction/melody-formation process.

But you don't need the repeated pronouncements of a UCSD professor to hear the power of the hearing system to extract melodic information and impose it on the perception of the incoming sounds. You can instead go back a millennium to the formative years of Western music. Perhaps the best known trait of Gregorian chant is that it contains a single melodic line. But if it is recorded in an adequately reverberant church and you play it backwards, dissonant harmonies suddenly jump out at you and what was sung by a single singer can sound like it comes from a choir singing multiple melodic lines. The latter, polyphonic effect presumably comes from the hearing system "connecting" the each note change it hears to the nearest convenient continuous tone, with multiple notes in rapid succession being assigned to different (and illusory) melodic lines, the whole process being aided by the holdover from the reverberation. You can even faintly hear such effects in the reversed-chant moments mixed into Enigma's MCMXC a.d.

Warning: This is not your typical chill-out Gregorian chant recording. It is instead a time-capsule/travelogue of different chant styles, some of which include real, non-illusory harmony, and is sung with wide-awake, full-throated fervor (sample track 10 at the iTunes store or at Amazon). Some of the works receive their "first known performances since the Middle Ages." How about that for a "sleeper."

The second segment of Radio Lab's "Musical Language" program, covering the cross-cultural similarities in the melodic contours of baby talk as well as the characteristics of the output of the basilar membrane (the organ in the ear that changes sound into nerve impulses), was less interesting to me. I'd already heard about the cross-cultural similarities in baby-talk melodic shapes back in the 1970's from no less an eminence than Leonard Bernstein in his Norton Lectures delivered at Harvard. (My fans will already know that I appear, fleetingly, in the DVDs of those lectures, as well as in a still photo in the exciting, still-in-print book version from Harvard University Press.) And the portion of the second segment about the "chaotic" nature of the ear's neural output when confronted by harmonic dissonances I'd already read about in hearing/acoustics textbooks. But if you haven't encountered these ideas, I urge you to listen to this segment, too. It does make the really obvious error, however, of assigning Disney's use of Stravinsky's Rite of Spring to the dancing-mushroom sequence of the original Fantasia instead of the primeval-earth/dinosaur sequence. (Hey, this is the Dartboard!) And it favors those chaotic nerve impulses over the numerous more significant social and artistic factors surrounding the riot that developed at the Rite's 1913 premier in Paris, best elaborated in Thomas Forrest Kelly's book First Nights: Five Musical Premiers.

The Radio Lab program's last segment likewise would have been more interesting to me if the subject — composer David Cope's development of computer software that could generate music in imitation of various composer's styles — hadn't been anticipated in Kemal Ebcioglu's work at IBM on similarly automated harmonizations of hymn melodies "in the style of Johann Sebastian Bach" that I covered when the research was brand new (the late 1980s). I would have been more interested if Cope had come up with an automated "composer" with a recognizable style of its own and that produced melodies and harmonies with real emotional power of their own, not merely as a consequence of our hearing "into" the compositions our reactions to the human composers being imitated. P.D.Q. Bach has investigated that area already, with far more entertaining results.

A note on the Gregorian Chant excerpt: It comes from the second track of an outstanding Sony/BMG album called Chant Wars performed by Sequentia and Dialogos. It's available complete as an iTunes download, but I ripped the example from my disc. I highly recommend the latter as it is a hybrid SACD in surround sound, which always enhances Gregorian chant.

A note on the Radio Lab streaming replay: After hearing the live over-the-air broadcast of the Radio Lab show, I was astonished to find that the online MP3 streaming version is only in mono. All the show's crafty stereophonic tricks are eliminated in the mixdown! This does save on Internet transmission bandwidth, always a concern at a public radio station, but to me it is bit-wise and sound-foolish.

See the previous David's Dartboard entry.

Back to Homepage
What's New on S&V