The Case of the Mysterious Cornetist

From here, it's a long way to the jazz joints of New York and the art form's birthplaces, like New Orleans and Kansas City. Here at Concord College, on a damp, green ridge of the Appalachian Mountains, Gary Westbrook doesn't exactly resemble a ghost of Dixieland as he peers at a laptop computer. A sequence of contorted lines shudders across the screen. "It's all in the tone," he says.

Mr. Westbrook and a colleague, Tom Smith, say that in readouts like these are the solutions to mysteries that aficionados of early jazz, a proudly fanatical breed, have fixated on for decades. Was it, for example, really the cornetist Bix Beiderbecke on that 1929 recording of "Baby Won't You Please Come Home"?

Accounts of who played on what recordings were often incomplete, sometimes purposely. Louis Armstrong and Muggsy Spanier, among numerous early jazzmen, recorded at times under assumed names. Some -- Beiderbecke, famously -- had other musicians sit in for them while they, say, recovered from a drinking binge. Sometimes the producers colluded to keep unpleasantness concealed from the record-buying public.

To identify uncredited or miscredited performers, jazz experts have depended on their own ears, a practice that in almost all cases leaves room for disagreement -- and further speculation. So, aficionados who venerate 78-rpm platters and earlier wax-cylinder recordings will surely be disappointed if Mr. Westbrook and Mr. Smith are right in their claim that they can ascertain lineups by using modern sound-analysis software, backed by biographical research. They say that the software lets them differentiate players by measuring their tonal characteristics. The method's limitation, the two researchers say, is that it works only with instruments that produce tone, such as horns and woodwinds.

"At worst," says Mr. Westbrook, a percussionist who is an adjunct professor of music at Concord, "we can reduce it to, well, there may be four people it could be, but it couldn't be so-and-so because he was in Tokyo that day."

He stares at the patterns that syncopate across his computer screen. They do not readily surrender their secrets to a nonspecialist. In any case, one has the distinct impression that something is missing.

Sound!

No cornet or clarinet or trombone is heard. Mr. Westbrook does not analyze heard sound at all. He converts selected recordings to digital files, which his sound-analysis software works on in silence. An onlooker can only imagine, or recall, Armstrong with his blatting power and tireless invention; Spanier, his tone fat, his growl exuberant; Beiderbecke, sonorous and quicksilver in lyrical solos.

Mr. Westbrook and Mr. Smith, who is a jazz trombonist and a professor of music at Pfeiffer University, four hours south of here in North Carolina, depend for their detection on sound-wave-analysis software called SpectraPlus. It measures tone in terms of how comparatively loudly the musicians characteristically play in various parts of the spectrum of sound that their instruments produce. The two-dimensional spectrograms on Mr. Westbrook's screen register frequency on one axis and amplitude, or strength of sound, on the other.

In a three-dimensional mode, they can also show how frequency and amplitude are related over time. Mr. Westbrook switches to that mode and points to one region of the display, which resembles a topographical map of a mountain range. "See that ridge there? On most Bix solos, right around 11,000 hertz there seems to be this peak that's just way out on its own. I went back and looked at all my Bix examples, and that peak was present in all of them."

No humans, he notes, have ears so finely tuned that they can say, "Oh, he has a very high 11,000-hertz level." Rather, one hears the combination of such characteristics as the player's tone.

Mr. Westbrook and Mr. Smith say they have found that just as trumpets, for example, share a characteristic sound, each player produces characteristics of tone that the software program can register. In a sort of sonic fingerprinting, the researchers compare various recordings and determine whether they were made by the same player.

"We've found that we can take an artist in his 20s who to the naked ear sounds completely different at 60 years of age, but when we test him it is almost identical," says Mr. Westbrook. Certain emphases within the spectrum of the sounds that a player produces never change, he says. The researchers presume that this distinctiveness results from the physiology of a player's mouth, trachea, and lungs; the embouchure; and the diaphragm's strength and action.

Mr. Westbrook has analyzed several dozen recordings, using a test of statistical significance to compare them with recordings on which the lineups are known for certain. That produces likely matches. In the case of Beiderbecke, the two researchers have weighed in on which of his supposed recordings actually were played by a substitute, Andy Secrest. Throughout Beiderbecke's playing career -- which alcoholism brought to an early end -- the cornetist often missed recording dates and gigs. Secrest, a skilled imitator, often was called on to take his place.

Once a technical diagnosis is complete, Mr. Smith sets to work to try to substantiate the findings with biographical research. "All the mystery recordings have a story that's so fascinating," says the music historian, by telephone. Among early-jazz fans, he says, fact is often a matter of "the agreed-upon lie." His investigations are attempts to set the records straight. He is gathering them into a book he plans to call The Jazz Detectives, a sobriquet that colleagues have come to attach to him and Mr. Westbrook. The book will be full of "Sherlock Holmes-type scenarios," says Mr. Smith. "Musician X spits blood in balls, another guy has to fill in -- that sort of thing."

The jazz press has been fairly quiet about the two researchers' findings, although JazzTimes did report on their work in October. But some academics are skeptical. They doubt that the technology can do what the researchers say it can. The measurements are not sophisticated enough; factors like microphone placement affect qualities of the recorded sound; background noise make it nearly impossible to isolate a soloist's sound in order to be sure what SpectraPlus is analyzing.

Such objections frustrate the jazz detectives. "We have covered those factors as thoroughly as is possible with present technologies," says Mr. Smith. "We have argued ourselves silly with reputed experts everywhere, who are never satisfied."

He ascribes the skepticism to two causes: He and Mr. Westbrook are "treading on the sacred turf" of "reputed sound-technology experts," and "when we come up with this research, it kills their fun. I hate to be rude, but we really don't care. We're historians."

Doubters also wonder about the jazz detectives' claim that their research involves a proprietary method, one that no one else can use without their permission. Why, then, don't they file for a patent, so that the details of their testing could be made public? "We have," answers Mr. Smith.

Mr. Westbrook is not too bothered by these critics. He simply asserts that the method works. "Our system doesn't have anything to do with the rhythms or actual individual notes musicians are playing. It's just analyzing their sound," he says, as more patterns dance silently across his computer screen. It doesn't even matter if a musician is playing the same notes from. one sample to another. "It's still him playing that horn."