The Smith/ Westbrook Mystery Recording Method
Discovery of the Musical Fingerprint/ Tom Smith and Gary Westbrook Creators.
Jazz Times, October, 2002
The Jazz Detectives
Identifying Unknown or Mislabeled Musicians
Jonah Berman
If you're a collector of old jazz recordings you're familiar with these scenes:
While reading through personnel list you come across "bass player unknown," or"drums unidentified." Somehow, somewhere along the line, the
name of the musician who was playing that particular instrument on that particular date got lost.
Or how about the times when you're listening to a record that is allegedly by your favorite jazz musician, but deep down you know that the playing doesn't sound like his style.
While many might not be bothered by such inconsistencies, dedicated jazz aficionados often are. Fortunately, the jazz detectives have arrived.
Tom Smith and Gary Westbrook, both professors at Pfeiffer University in North Carolina, are attacking the problem headon. Using SpectraPlus
sound wave technology, Smith and Westbrook analyze old recordings in an attempt to name unknown or mislabeled musicians.
"It concerned me that when you review these old recordings you see 'personnel unidentified," Smith says. "It could be this guy, it could be that guy.
And then when you investigate, you discover the guess came from someone who used to hang around the clubs and had no idea what he was
talking about, or was lying altogether." Or someone would say, 'Well, I used to play with this guy, so he thinks he knows what someone sounds like
. or, It had to be this other guy, because I was on the session too."
"It bothered me that jazz history was being written for the next 1,000 years, and it was turning into what Donald Byrd called 'the lie that's agreed
upon. So why not open the door and use some technology as opposed to this guesswork."
This is how their detective process works: Smith selects a recording that is under question and brings it to Westbrook, who is the statistician. They isolate the instrument being analyzed and compare the sine wave reading to other recordings until they find a match. Each musician has what
Smith calls a "musical fingerprint," based on the amount of breath support, the diaphragm and facial bone structure, among other things. Smith
and Westbrook work until they find an exact match before confidently concluding who the musician is. "This fingerprint remains with you
throughout your career," Smith says. "What's fascinating is that some will change their style, the way they play, or sound different when they're
sick, but they register on the method same as before. When Sonny Rollins went under the bridge, his tone changed a bit, supposedly to adjust to Coltrane. But still, the 'fingerprint' comes out the same as before he changed."
While much of their work deals with naming unidentified lesser-known musicians on recordings, they have uncovered some fraudulent personnel listings as well. "For some reason, the ones that seem to be the most contentious are Bix Beiderbecke recordings," Smith says. "He sent so many substitutes to gigs because of his alcohol problems. If he got sick and had a session, he would just send a player who sounded like him, and that happened constantly. Then the producer would have to sell it as a Beiderbecke session, because nobody would buy the sub's record. There was a
guy named Andy Secrest who was regularly sent to Beiderbecke sessions, and he would get pretty close. But he wasn't Beiderbecke, and our
analysis tells us that. We've also looked at a couple of Coltrane imitators who came pretty close, but not close enough."
When they were first released, no one thought these jazz recordings would be listened to in a hundred years,"Smith says. "And now we're working through this cloud that was perhaps created intentionally, just to sell a few records. With jazz, there's so much folklore, and so many musicians
were taken at their word, we need to double check to make sure we have the facts right".
The professors have licensed their methods and presented their research at the 2001 IAJE conference in New York. They also plan to release a
book documenting their findings.
A tale of two clarinetists.
Left: Frank Teschemacher, Right: Benny Goodman
Unlike Beiderbecke historians, who have fought to maintain guesswork and
assumption, Teschemacher/Austin High Gang scholars have widely accepted
Smith/Westbrook and have (since 2001) altered discographies accordingly.
The clarinet solo discussed in the above research begins at 1:37.
The Beiderbecke Mystery Recordings
No recorded twentieth century brass musician has elicited a greater need for accurate identification than jazz cornetist Leon Bismark (Bix) Beiderbecke. Due to erratic behavior caused in part by chronic alcoholism, his attendance or lack thereof at as many as thirty speculative
recording sessions has fueled a musical legend already elevated by martyrdom derived from the unfortunate happenstance of dying young. Beiderbecke’s alcoholic episodes were at a peak at or around the first five months of 1929. This period, which includes a brief time spent in the employ of bandleader Paul Whiteman, coincides with extended interludes of paranoid insecurities, accompanied by a complete physical and
mental breakdown, and a mysterious beating that may have resulted in permanent injury. After a brief recuperation at his home in Davenport,
Iowa, Beiderbecke was back in New York performing with Whiteman in March, and engaged in a number of freelance recording sessions, that
were ill advised, due to the nature of his rapidly deteriorating condition.
In fact, the decline of Beiderbecke’s physical and mental health were judged so severe as to promote the budding career of a twenty year old cornetist, who for a time made his living performing the role of Bix imitator and “stand in”. Whiteman actually hired Andy Secrest (sometimes Seacrest) as a substitute during Beiderbecke’s recuperation, but kept him on later for the expressed purpose of performing Bix styled improvisations, for those occasions when Beiderbecke himself was indisposed. The eager Secrest became so expert at imitating Beiderbecke,
that he was able to extend his recording opportunities past Whiteman, and into other “Bix friendly” venues. In few places was Secrest’s impact
more felt than in the studio sessions of saxophonist Frank Trumbauer; a man intensely devoted to his friend Beiderbcke, yet practical enough to understand the necessity for insurance when the situation warranted it.
The “Baby Won’t You Please Come Home” Session
Few mysteriously identified recordings have generated more controversy than the April 17, 1929 Trumbauer rendition of the song Baby Won’t
You Please Come Home. In the liner notes for a 1947 reissue of the same recording, George Avakian states the following: “Baby Won’t You Please Come Home features a Trumbauer vocal and two solo choruses. Despite much speculation that Andy Secrest may have played one of the
choruses, the accepted decision among most musicians and Bixophiles (term used to describe a Beiderbecke researcher), is that Beiderbecke is responsible for both solos.” Despite Avakian’s reasonable assurances, legend continues to forward three possible scenarios. 1. Beiderbecke performed both cornet solos. 2.Secrest performed the first solo (open) and Beiderbecke performed the second solo (muted). 3.Secrest performed both solos. The source of the dispute derives from the intelligent observations of lifelong Beiderbecke researchers like Richard Sudhalter,
Randy Sandke, and Mark Richard, whose opinions must be weighed with due consideration.
Sudhalter in his book Bix: Man and Legend, provides the following information: “Solos: Secrest (16 verse); Bix (first fill); Secrest (other fills);
Bix (16); Secrest (lead last chorus); Bix (muted obbligato).” It is therefore the belief of Sudhalter that Secrest performed the first solo and that Beiderbecke performed the second solo (muted). This is the consensus of a vast majority of Beiderbecke researchers including Sandke, Red Hot Jazz Archives contributor Mike Donovan and M.J. Logsdon of the Wolverine Antique Music Society, whose comments do much to fuel the ongoing speculation. “The first solo in Baby Won’t You Please Come Home does resemble Bix, but its crispness, clarity, and lightly- brasher- sounding
–than- Bix sound are in distinct contrast to the muted second solo by Bix, which after several listenings, does in fact sound different from the
first solo, in spite of the mute.” Sandke’s observations are even more convincing, and shed greater light on the circumstances that may have led
to Beiderbecke’s abbreviated solo activity on Baby. (Regarding the April 17 session): “Instead of sounding stronger, he seems even more unsure
of himself. Secrest handles most of the lead. On both Louise and Wait Till You See Ma Cherie (previously recorded selections), Bix again uncharacteristically finishes with high notes and the results are again strained. On Baby Won’t You Please Come Home, he settles down and
plays a fine lyrical solo, but by this time is lip is spent and it almost refuses to vibrate on the last four bars.”
For years historians insisted that another musician played the second cornet
solo on this 1929 recording. Smith/Westbrook contends the matter is closed
within a .05% margin of error. Beiberbecke did indeed play both solos.
Mark Richard in his own liner notes for a “Masters of Jazz” compact disc collection derives at a third more radical conclusion. “Solos: Secrest, c (abbrev. for cornet) (16 verse)-Tram (abbrev. for Trumbauer) voc, with Bix, c in derby obbligato (16)-Tram, Cms (abbrev. for C melody saxophone)-Secrest, c in derby (16)-Bix, c (leads last 16).” 14 It is apparent from Richard’s outline that he disputes the notion of any Beiderbcke solo presence.
If one were to qualify the above data with conscientious reason, it would be surmised that a distinguished collection of talented observers was
in obvious disagreement. Therefore Gary and I judged Baby Won’t You Please Come Home an excellent example of a mystery brass recording in need of scientific evaluation.
Procedure
For comparison eight solos were selected. Solo one was the Baby Won't You Please Come Home open mystery recording and solo two was the
Baby Won't You Please Come Home muted mystery recording. The recordings in dispute were compared to six solos known to be either
Beiderbecke or Secrest. Examples of both open horn and muted selections were included. For the facilitation of this procedure they are identified
as solos three through eight. Solo three was Dardenella by Beiderbecke. Solo four was Singin' The Blues by Beiderbecke. Solo five was You Took Advantage of Me, a muted solo by Beiderbecke. Solo six was Alabamy Snow by Secrest. Solo seven was What A Day by Secrest. Solo eight was Remember Me? a muted solo by Secrest.
The first two pairs analyzed were solos three and four. There was a strong and positive relationship between solo three (Dardanella) and solo
four (Singin' The Blues) (r = .794). A critical t value of 1.5 was found at the p = .138 level. This result led the researchers to retain the null
hypothesis that there were no statistically significant differences between solos three and four. The next two pairs analyzed were solos six and seven. There was a moderate positive relationship between solo six (Alabamy Snow) and solo seven (What A Day) (r = .523). A critical t value of
.11 was found at the p = .916 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos six and seven. The next two pairs analyzed were solos six and eight. There was a moderate positive relationship
between solo six (Alabamy Snow) and solo eight (Remember Me?) (r = .59). A critical t value of -.24 was found at the p = .808 level. This result
led the researchers to retain the null hypothesis that there were no statistically significant differences between solos six and eight. The next two pairs analyzed were solos seven and eight. There was a moderate positive relationship between solo seven (What A Day) and solo eight
(Remember Me?) (r = .69). A critical t value of -.41 was found at the p = .682 level. This result led the researchers to retain the null hypothesis
that there were no statistically significant differences between solos six seven and eight. The next two pairs analyzed were solos three and five. There was a moderate positive relationship between solo three (Dardenella) and solo five (Took Advantage of Me) (r = .574). A critical t value of
3.16 was found at the p =.002 level. This result led us to reject the null hypothesis that there were no statistically significant differences between solos three and five and to accept the alternative hypothesis that there were significant differences beyond the p = .05 level. The next two pairs analyzed were solos four and five. There was a moderate positive relationship between solo four (Singin' The Blues) and solo five (Took
Advantage of Me) (r = .638). A critical t value of 2.51 was found at the p = .014 level. This result led the researchers to reject the null hypothesis
that there were no statistically significant differences between solos four and five and to accept the alternative hypothesis that there were significant differences beyond the p = .05 level. These first six tests were administered to compare known soloists to themselves. This gave the researchers the opportunity to test the procedure, again.
The next comparisons were tests constructed to identify the soloist on each mystery recording. The next two pairs analyzed were solos six and
one. There was a moderate positive relationship between solo six (Alabamy Snow) and solo one (Baby Won't You Please Come Home) (r = .555).
A critical t value of 1.75 was found at the p = .084 level. This result led the researchers to retain the null hypothesis that there were no statistically significant different between solos six and one. The next two pairs analyzed were solos seven and one. There was a moderate positive
relationship between solo seven (What A Day) and solo one (Baby Won't You Please Come Home) (r = .601). A critical t value of 1.94 was found at
the p = .056 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between
solos seven and one and to accept the alternate hypothesis that there were significant differences beyond the p = .05 level. The next two pairs analyzed were solos three and one. There was a moderate positive relationship between solo three (Dardenella) and solo one (Baby Won't You Please Come Home) (r = .629). A critical t value of -1.47 was found at the p = .145 level. This result led the researchers to retain the null
hypothesis that there were no statistically significant differences between solos three and one. The next two pairs analyzed were solos four and one. There was a moderate positive relationship between solo four (Singin' The Blues) and solo one (Baby Won't You Please Come Home) (r = .636). A critical t value of -.41 was found at the p = .686 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos four and one. The next two pairs analyzed were solos five and two. There was a moderate positive relationship between solo five (Took Advantage of Me) and solo two (Baby Won't You Please Come Home) (r = .474). A critical t value of 2.56 was found at the p = .012 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between solos five and two and accept the alternative hypothesis that there were statistically significant differences beyond the p = .05 level. The next two pairs analyzed were solos eight and two. There was a moderate positive relationship between solo eight (Remember Me?) and solo two (Baby Won't You Please Come Home) (r = .625). A critical t value of 2.72 was found at the p = .008 level. This result led the researchers to reject
the null hypothesis that there were no statistically significant differences between solos eight and two and accept the alternative hypothesis that there were statistically significant differences beyond the p = .05 level. The next two pairs analyzed were solos three and two. There was a moderate positive relationship between solo three (Dardenella) and solo two (Baby Won't You Please Come Home) (r = .515). A critical t value of
-.51 was found at the p = .615 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos three and two. The next two pairs analyzed were solos four and two. There was a moderate positive relationship
between solo four (Singin' The Blues) and solo two (Baby Won't You Please Come Home) (r = .601). A critical t value of .57 was found at the
p = .572 level. This result led us to retain the null hypothesis that there were no statistically significant differences between solos four and two.
The next two pairs analyzed were solos six and two. There was a moderate positive relationship between solo six (Alabamy Snow) and solo two (Baby Won't You Please Come Home) (r = .64). A critical t value of 2.93 was found at the p = .004 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between solos six and two and to accept the alternative hypothesis that
there were statistically significant differences beyond the p = .05 level. The next two pairs analyzed were solos seven and two. There was a moderate positive relationship between solo seven (What A Day) and solo two (Baby Won't You Please Come Home) (r = .622). A critical t value of 2.95 was found at the p = .004 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between solos seven and two and to accept the alternative hypothesis that there were statistically significant differences beyond the
p = .05 level. The last pairs analyzed were the two mystery recordings solos two and one. There was a moderate positive relationship between
solo two (Baby Won't You Please Come Home) and solo one (Baby Won't You Please Come Home) (r = .517). A critical t value of -.86 was found at
the p = .393 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between
solos two and one.
Conclusions
Testing analysis concluded that solos three (Dardenella), four (Singin' The Blues) and five (Took Advantage of Me) were performed by Bix Beiderbecke; and that solos six (Alabamy Snow) seven (What A Day) and eight (Remember Me?) were performed by Secrest. This was expected since these were recordings where the status of said personnel was never in question. The researchers hoped to identify the mystery
performers by finding no statistically significant differences between the mystery recordings (solos one and two) and either solos three, four, and five (Beiderbecke) or solos six, seven and eight (Secrest). The results indicated that there was no statistically significant differences (t = 1. 5, p = .138) between solo three (Dardenella) and solo four (Singin' The Blues). Therefore, the researchers concluded that both solos must be from the same population. The result was expected since the performer of each solo was definitely Beiderbecke. The comparison of solo three
(Dardenella) and solo five (Took Advantage of Me) revealed statistically significant differences (t = 3.16, p = .002). The results indicated that there were statistically significant differences (t = 2.51, p = .014) between solo four (Singin' The Blues) and solo five (Took Advantage of Me). Therefore,
the researchers concluded that the solos must be from different populations. This result was unexpected since the performer of each solo was definitely Beiderbecke. However, solo five was a muted solo where Beiderbecke traded fours with Trumbauer. Westbrook initially "red flagged"
this solo, since Trumbauer’s and Beiderbecke’s continuity in the exchange was so fluid as to have interfered with the Beiderbecke sample. This
undoubtedly accounts for the statistical error. This also verifies and confirms the need for clean and clear samples, possessing definite points of embarkation and departure. The results indicated that there was no statistically significant differences (t = .11, p = .916) between solo six
(Alabamy Snow) and solo seven (What A Day). Therefore, said research concluded that both solos were from the same population. The result
was expected since the solos had been positively identified as Secrest. The comparison of solo six (Alabamy Snow) and solo eight
(Remember Me?) revealed no statistically significant differences (t = -.24, p = .808). Therefore, the researchers concluded that the solos must
be from the same population. This result was also expected since the performer of each solo was definitely Secrest. However, the researchers
were initially unsure of testing that compared an open brass solo to a muted brass solo. These results revealed that a performer has a unique sound, which is not affected by the use of mutes. The results indicated that there were not statistically significant differences (t = -.41, p = .682) between solo eight (Remember Me?) and solo seven (What A Day). Therefore, the researchers concluded that both solos must be from the
same population. The result was expected since the performer of each solo was definitely Secrest. The comparison of solo seven (What A Day)
and solo one (Baby Won't You Please Come Home) revealed statistically significant differences (t = 1.94, p = .056). Therefore, the researchers concluded that the solos must be from different populations. The results indicated that there was no statistically significant differences
(t = 1.75, p = .084) between solo six (Alabamy Snow) and solo one (Baby Won't You Please Come Home). Therefore, the researchers concluded
that both solos must be from the same population. However, since the alpha level (p = .084) was so close to the p = .05 level, we decided to give
this analysis a closer look. After all, Secrest was regarded as perhaps the most celebrated imitator of Beiderbecke, and had even possibly
fooled some of the world’s foremost Bixophiles. The comparison of solo three (Dardenella) and solo one (Baby Won't You Please Come Home) revealed no statistically significant differences (t = -1.47, p = .145). Therefore, the researchers concluded that the solos must be from the same population. The results indicated that there was no statistical difference (t = -.41, p = .686) between solo four (Singin' The Blues) and solo one (
Baby Won't You Please Come Home). Therefore, said research concluded that both solos were from the same population. Said research concluded that the soloist on the open solo of Baby Won't You Please Come Home was Beiderbecke: a conclusion that stands in disagreement with a large number of Beiderbecke researchers, but one that the researchers stand by nonetheless, based upon strong scientific principals, and an inconsequential margin of error. The comparison of solo five (Took Advantage of Me) and solo two (Baby Won't You Please Come Home) revealed statistically significant differences (t = 2.56, p = .012). Therefore, the researchers concluded that the solos must be from different populations.
We again referred to the solo in You Took Advantage of Me as problematic because of the interference of the other soloist. The results indicated
that there was statistically significant differences (t = 2.72, p = .008) between solo eight (Remember Me?) and solo two (Baby Won't You Please Come Home). Therefore, we concluded that both solos must be from different populations. The comparison of solo three (Dardenella) and solo
two (Baby Won't You Please Come Home) revealed no statistically significant differences (t = -.51, p = .615). Therefore, the researchers concluded that the solos must be from the same population. The results indicated that there was no statistically significant differences (t = .57, p = .572) between solo four (Singin' The Blues) and solo two (Baby Won't You Please Come Home). Therefore, the researchers concluded that both solos
must be from the same population. The comparison of solo six (Alabamy Snow) and solo two (Baby Won't You Please Come Home) revealed statistically significant differences (t = 2.93, p = .004). Therefore, the researchers concluded that the solos must be from different populations.
The results indicated that there were statistically significant differences (t = 2.95, p = .004) between solo seven (What A Day) and solo two (Baby Won't You Please Come Home). Therefore, said research concluded that both solos must be from different populations. Said research concluded that the soloist on the muted solo of Baby Won't You Please Come Home was also Beiderbecke. A comparison of solo one (Baby Won't You Please Come Home) and solo two (Baby Won't You Please Come Home) found no statistically significant differences (t = -.86, p = .393). Therefore, the researchers concluded that both solos must be from the same population. This comparison shows that the soloist on solo one must be the same
as the soloist on solo two.
FINAL RESULTS: BEIDERBECKE PLAYED BOTH SOLOS.
"The Waiting at the End of the Road" Session
The Paul Whiteman Orchestra with both Beiderbecke and Secrest in attendance recorded Waiting at the End of the Road and When You’re
Counting the Stars Alone on September 13, 1929. In most accredited discographies, the cornet solos included are recognized as Beiderbecke’s
last with a Whiteman led ensemble, with a few notable exceptions. Still, because of Beiderbecke’s rapidly deteriorating condition and an
abundance of conflicting anecdotal information, there are appears to be enough just cause to assert three possible scenarios of what actually occurred: 1. Beiderbecke was the cornet soloist on both selections. 2. Secrest was the cornet soloist on both selections. 3. Beiderbecke and
Secrest shared solo responsibilities on either or both selections. Due to an abundant quantity of conflicting anecdotal information regarding what actually transpired at that Waiting at the End of the Road session, compounded by the apparent inability for discographers to reach an audial consensus, we concluded with ample justification that the recordings in question were an appropriate demonstration for our mystery
identification procedure.
Procedure
For comparison nine solos were selected. Solo one was the Waiting At The End of The Road open mystery recording. Solo two was the Waiting At The End of The Road muted mystery recording. Solo three was the mystery recording When You're Counting The Stars Alone. Solo four was Dardenella by Bix Beiderbecke. Solo five was Singin' The Blues by Bix Beiderbecke. Solo six was Alabamy Snow by Secrest. Solo seven was What
A Day by Secrest. Solo eight was Remember Me? which is a muted solo by Secrest. Solo nine was Here Comes The Showboat by Secrest. The
first two pairs analyzed were solos four and five. There was a strong and positive relationship between solo four (Dardanella) and solo five
(Singin' The Blues) (r = .794). A critical t value of 1.5 was found at the p = .138 level. This result led the researchers to retain the null hypothesis
that there were no statistically significant differences between solos four and five. The next two pairs analyzed were solos six and seven. There
was a moderate positive relationship between solo six (Alabamy Snow) and solo seven (What A Day) (r = .523). A critical t value of .11 was found
at the p = .916 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos six and seven. The next two pairs analyzed were solos six and eight. There was a moderate positive relationship between solo six (Alabamy Snow) and solo eight (Remember Me?) (r = .59). A critical t value of -.24 was found at the p = .808 level. This result led the researchers to retain
the null hypothesis that there were no statistically significant differences between solos six and eight. The next two pairs analyzed were solos seven and eight. There was a moderate positive relationship between solo seven (What A Day) and solo eight (Remember Me?) (r = .69). A
critical t value of -.41 was found at the p = .682 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos seven and eight. The next two pairs analyzed were solos six and nine. There was a moderate positive relationship between solo six (Alabamy Snow) and solo nine (Here Comes The Showboat) (r = .692). A critical t value of .71 was found at the
p = .48 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos
seven and eight. These first five tests were use to compare known soloists to themselves. This gave the researchers the opportunity to test the procedure, again. The next comparisons were test, constructed to identify the soloist on each mystery recording. The next two pairs analyzed
were solos six and one. There was a moderate positive relationship between solo four (Dardenella) and solo one (Waiting At The End of The Road)
(r = .582). A critical t value of -3.11 was found at the p = .003 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between solos four and one and accept the alternative hypothesis that there were statistically significant differences beyond the p = .05 level. The next two pairs analyzed were solos seven and one. There was a moderate positive relationship between solo nine (Here Comes The Showboat) and solo one (Waiting At The End of The Road) (r = .603). A critical t value of 1.18 was found at the p = .243 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos nine and
one. The next two pairs analyzed were solos three and one. There was a moderate positive relationship between solo nine (Here Comes The Showboat) and solo three (When You're Counting The Stars Alone) (r = .701). A critical t value of 1.75 was found at the p = .083 level. This result
led the researchers to retain the null hypothesis that there were no statistically significant differences between solos nine and three. The next
two pairs analyzed were solos four and three. There was a moderate positive relationship between solo four (Dardenella) and solo three (When You're Counting The Stars Alone) (r = .567). A critical t value of -2.68 was found at the p = .009 level. This result led the researchers to reject the
null hypothesis that there were no statistically significant differences between solos four and three and accept the alternative hypothesis that
there were statistically significant differences beyond the p = .05 level.
The next two pairs analyzed were solos one and three. There was a strong positive relationship between solo one (Waiting At The End of The
Road) and solo three (When You're Counting The Stars Alone) (r = .772). A critical t value of .72 was found at the p = .474 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos one and three. The next two pairs analyzed were solos four and two. There was a moderate positive relationship between solo four (Dardenella) and solo two (Waiting At The End
of The Road) (r = .624). A critical t value of -.84> > was found at the p = .403 level. This result led the researchers to retain the null hypothesis
that there were no statistically significant differences between solos four and two.
The next two pairs analyzed were solos one and two. There was a moderate positive relationship between solo one (Waiting At The End of The
Road) and solo two (Waiting At The End of The Road) (r = .727). A critical t value of 2.97 was found at the p = .004 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between solos one and two and accept the alternative hypothesis that there were statistically significant differences beyond the p = .05 level. The next two pairs analyzed were solos eight
and two. There was a moderate positive relationship between solo eight (Remember Me?) and solo two (Waiting At The End of The Road) (r = .721).
A critical t value of 2.84 was found at the p = .006 level. This result led the researchers to reject the null hypothesis that there were no statistically significant differences between solos eight and two and accept the alternative hypothesis that there were statistically significant differences beyond the p = .05 level.
The next two pairs analyzed were solos three and two. There was a strong positive relationship between solo three (When You're Counting The Stars Alone) and solo two (Waiting At The End of The Road) (r = .788). A critical t value of 2.79 was found at the p = .007 level. This result led the researchers to reject the null hypothesis that there were no significant differences between solos three and two and to accept the alternative hypothesis that there were statistically significant differences beyond the p = .05 level. The last pairs analyzed were solos five and two. There was
a moderate positive relationship between solo five (Singin' The Blues) and solo two (Waiting At The End of The Road) (r = .664). A critical t value
of .34 was found at the p = .732 level. This result led the researchers to retain the null hypothesis that there were no statistically significant differences between solos five and two.
The "Waiting at the End of the Road session is deemed important because it
was reputedly Beiderbecke's last, meaning even short solos like the one
beginning at 2:35 have historical significance.
Conclusions
The purpose of this study was to identify the performers on the two Waiting At The End of The Road mystery recordings, and the When You're Counting The Stars Alone mystery recording. Solos four (Dardenella) and five (Singin' The Blues) were performed by Bix Beiderbecke. Solos six (Alabamy Snow), seven (What A Day), eight (Remember Me?) and nine (Here Comes The Showboat) were performed by Secrest. Solos one (open), two (muted) and three were mystery recordings. The researchers hoped to identify the mystery performers by finding no statistically significant differences between the mystery recordings (solos one, two, and three) and either solos four and five or solos six, seven, eight, and nine. The
results indicated that there were no statistically significant differences (t = 1. 5, p = .138) between solo four (Dardenella) and solo five (Singin'
The Blues). Therefore, the researchers concluded that both solos must be from the same population. The result was expected since the
performer of each solo was definitely Beiderbecke. The comparison of solo six (Alabamy Snow) and solo eight (Remember Me?) revealed no statistically significant differences (t> = -.24, p = .808). Therefore, the researchers concluded that the solos must be from the same population.
This result was expected since the performer of each solo was definitely Secrest. However, the researchers were unsure what the comparison
of an open solo to a muted solo would identify. These results revealed that a performer has a unique sound, which is not affected by the use of mutes. The results indicated that there were no statistically significant differences (t = .11, p = .916) between solo six (Alabamy Snow) and solo seven
(What A Day). Therefore, the researchers concluded that both solos must be from the same population. The result was expected since the performer of each solo was definitely Secrest. The comparison of solo six (Alabamy Snow) and solo nine (Here Comes The Showboat) revealed
no statistically significant differences (t = .71, p = .48). Therefore, the researchers concluded that the solos must be from the same population.
This result was expected since the performer of each solo was definitely Secrest. The results indicated that there were statistically significant differences (t = -3.11, p = .003) between solo four (Dardenella) and solo one (Waiting At The End of The Road). Therefore, the researchers
oncluded that both solos must be from different populations. The comparison of solo nine (Here Comes The Showboat) and solo one (Waiting At
The End of The Road) revealed no statistically significant differences (t = 1.18, p = .243). Therefore, the researchers concluded that the solos
must be from the same population. The researchers, therefore, concluded that Secrest must be the soloist on solo one.
The results indicated that there were statistically significant differences (t = -2.68, p = .009) between solo four (Dardenella) and solo three
(When You're Counting The Stars Alone). Therefore, the researchers concluded that both solos must be from different populations. The
comparison of solo nine (Here Comes The Showboat) and solo two (When You're Counting The Stars Alone) revealed no statistically significant differences (t = 1.75, p = .083). Therefore, the researchers concluded that the solos must be from the same population. The researchers,
therefore, concluded that Secrest must be the soloist on solo three. A comparison of solo one (Waiting At The End of The Road) and solo three (When You're Counting The Stars Alone) revealed no statistically significant differences (t = .72, p = .474). Therefore, the researchers concluded
that both solos must be from the same population. Individual t tests revealed that Secrest was the performer on solos one and three. A t test analyzing the differences between solos one and three indicated solos one and three were performed by the same soloist, which confirmed
the results of the initial findings. The results indicated that there were no statistically significant differences (t = .34, p = .732) between solo five (Singin' The Blues) and solo two (Waiting At The End of The Road). Therefore, the researchers concluded that the solos must be from the same population. The comparison of solo four (Dardenella) and solo two (Waiting At The End of The Road) revealed no statistically significant
differences (t = -.84, p = .403). Therefore, the researchers concluded that the solos must be from the same population. The results indicated
that there were statistically significant differences (t = 2.84, p = .006) between solo eight (Remember Me?) and solo two (Waiting At The End of
The Road). Therefore, the researchers concluded that both solos must be from different populations. Therefore, the researchers concluded that
the solos must be from the same population. The researchers concluded that the soloist on the muted solo of Waiting At The End of The Road
was Beiderbecke. A comparison of solo one (Waiting At The End of The Road) and solo two (Waiting At The End of The Road) revealed statistically significant differences (t = 2.97, p = .004). Therefore, the researchers concluded that the solos must be from different populations. The results indicated that there were statistically significant differences (t = 2.79, p = .007) between solo two (Waiting At The End of The Road) and solo three (When You're Counting The Stars Alone). Therefore, the researchers concluded that the solos must be from different populations. These results were expected since t tests confirmed that Secrest was the performer on solos one and three, and Bix was the performer on solo two.
FINAL RESULTS: Waiting at the End of the Road (open) = Secrest
Waiting at the End of the Road (muted) = Beiderbecke
When You’re Counting the Stars Alone = Secrest
When Benny Goodman (right) was very young, he and Frank Teschemacher (left) sounded much alike. At that same time, the clarinetist Pee Wee Russell (below) was also confused with Teschemacher. Smith/Westbrook determined that Teschemacher performed on more recordings than was originally given credit.
"Business in F" is another former Teschemacher mystery resolved by Smith/Westbrook, important because it represented the clarinetist's last and "possibly" best performed solo (2:08). It was this improvisation that prompted Smith and Westbrook to create their musical fingerprinting system.
From Tech TV, February 27, 2002
Computer software is helping to solve musical mysteries from the last century.
It is often difficult to tell who is performing on vintage jazz recordings, so two "jazz detectives" from North Carolina are trying to match
anonymous musicians with the old melodies.
Tom Smith, Pfeiffer University music professor and jazz trombonist, says it's time to set the record straight -- about a lot of records.
"There wasn't enough attention paid to detail... in jazz history, and often it became what one of my professors called the lie that was agreed
upon," Smith said.
During the Prohibition era, Smith says, record producers often tried to pass off the work of unknown imitators as the product of stars like jazz
great Bix Beiderbecke. And they largely got away with it.
"At that time, many people never believed that jazz recordings would be something cared about 40 or 50 years down the line," Smith said.
But recent advances in voice-recognition technology convinced Smith and research partner Gary Westbrook that there must be a way to
measure every horn player's unique voice, or tone.
"The things that are going to make you and I sound different on the same instrument are the makeup of our face, the makeup of the chambers
of our body, the diaphragm, the amount of breath support we are able to generate," Westbrook said. "All of those things that also make us
individual with our voice, the way that we breathe, the way that we talk."
In their research, Smith and Westbrook decided to use sound-wave analysis software by SpectraPlus. The software measures frequency -- low notes on the left, to high notes on the right -- and loudness. Tone is measured by how loud a sound is at certain frequencies.
Westbrook randomly samples each mystery soloist three times, then compares the data with a known soloist on another recording.
"[For example], if one of them was a known -- Bix Beiderbecke -- the other was an unknown artist. If they are identified as not statistically significant, then we say that it must be the same artist," Westbrook said.
But the colorful characters from the early age of jazz often seem to be trying to evade detection.
In one Beiderbecke recording Smith and Westbrook studied, the musician misses an obvious note in a solo in an otherwise polished recording. Would a great cornetist like Beiderbecke botch it that badly?
Smith says it's possible. "He was a chronic alcoholic, and had numerous emotional problems that would have created blatant inconsistency."
"I think Bix would probably would be very embarrassed by it all," Westbrook said. "He probably thought that 20 years after his death no one
would ever listen to him play again."
But jazz history must be corrected now, Smith says, not only to credit the mystery musicians for their work, but to solidify jazz's place in the
human cultural record.
"Otherwise, this will be guesswork forever, and the lie that is agreed upon will become a self-fulfilling prophecy," Smith said.
On the campus of Pfeiffer University, an hour north of Charlotte, North Carolina, the jazz detectives are preparing to publish their findings -- and preparing for the controversy that will surely follow. Smith and Westbrook say there are a lot of jazz lovers out there who believe software is no match for human ears, and statistical analysis is no match for informed opinions.
"Probably 10, 20 years down the line, the procedure will be modified, and it will be perfected to much greater heights than [Westbrook] or I could ever imagine," Smith said. "But the ball has to start somewhere."
Interview excerpts that discuss Tom Smith and Gary Westbrook's work with
the musical fingerprinting system (extracted from National Public Radio's
All Things Considered and Tech TV) are found on this video. (Below photo)
Andy Secrest (1907-1977), the frequent Beiderbecke substitute in question.
@ Copyright 2022 Tom Smith All Rights Reserved