Lutherie Myth/Science: The Tone
of an Instrument Improves the More it is Played

Every luthier is familiar with the phenomenon. You string up an instrument for the first time, start to play it, adjusting tuning as the new strings and new instrument change in response to string tension. And as you continue to play, and tune, and play, and re-tune, the tone of the instrument begins to change. Interestingly enough, the change is always for the better. An instrument that started out weak and tinny starts to develop volume and richness as it is played those first few minutes.

This may be one source of the premise that stringed instruments get better the more they are played. Another source may be the simple observation that older instruments sound better than newer ones, and that older instruments must have been played a lot. Therefore, again, the observation that instruments get better with playing. And of course it may be that instruments actually do improve with lots of playing.

I don't know of any formal experiments that have attempted to quantify the changing quality of tone in the first few minutes of an instruments life. I have informally attempted to quantify this on a number of occasions, stringing up a new instrument and then subjecting it to repeated digital recordings which were then processed into spectrograms. In these attempts I have never been able to document obvious differences in tone as the instrument is played past the point at which it can actually hold its tuning from minute to minute. It is possible that most or even all of the observed changes in tone during the first minutes of an instrument's life are due more to acclimation on the part of the player than to any changes occurring in the instrument. This is of course just speculation on my part. But in any case my own preliminary investigations lead me to conclude that there is little reason to pursue research along these lines, at least for me.

But what about the idea that instruments improve in tone if they are played for a long period of time? Or that they improve even more if they are played well? Although these views are widely held there are no scientific studies which support these premises in any definitive manner. There are however a number of studies that do suggest that this is at least an area that bears more formal research. Graham Caldersmith¹ wrote a nice article on a related subject for American Lutherie, and I would highly recommend this as the first place for someone interested in this subject to look. Dr. Caldersmith does not provide answers to the question but frames the issue nicely and provides some historical perspective and references.

One of the cornerstones of scientific inquiry is the concept that correlation in and of itself does not imply causation. In the case at hand, what this means is that, although it is possible that there is indeed a high correlation between instruments that sound very good and those that are played a lot, this correlation in no way implies that there is a causal relationship between an instrument being played a lot and it sounding good. Lacking evidence to actually show causation, it is just as likely (some might say more likely) that it is the other way around – that the instruments that sound good are played more because they sound good. This premise falls roughly into the Darwinian, survival of the fittest category and it is quite a compelling argument. Tim Olsen² addressed this very nicely with respect to the correlation between the age of an instrument and superiority of tone, expressing the possibility that the perceived correlation between the age of an instrument and the superiority of its tone is a consequence of natural selection preserving those instruments which sounded good and destroying those that didn't. The same reasoning could just as readily be applied to the issue of a perceived correlation between tone and the amount of playing an instrument sees.

As mentioned above, there is some good research which indicates that there is reason that the possible causal relationship between amount of playing and superiority of instrument tone should be the subject of more formal scientific research. Norman Pickering³ reported on an informal study performed at a concert of a string quartet. The quartet played first on their own old, well played instruments, and then on newer and not so well played instruments, and the audience judged the tone of the older well played instruments to be superior. As pointed out this and other such informal studies do show that more formal research in this area may be fruitful, but they hardly provide solid evidence to support that a causal relationship exists. One of the reasons this study does not provide compelling evidence to support the premise is that it was not blind. It this case what that means is that the audience knew when the quartet was playing the old well played instruments and when they were playing the alternate instruments because they could see them. And it is possible (actually, given research into just this aspect of scientific research methodology itself, it is more than just possible) that the audience was unduly influenced by what they saw as much (or more) than by what they actually heard. In fact, Mead Killion⁴ recorded this concert and then presented the recordings to individuals, asking them to evaluate the tonal characteristics of the instruments from listening to the recordings. In this case there was no clear correlation between the superiority of tone and the age and use of the instruments.

So any experiment, the results of which are to be used to support the premise that well played instruments sound superior to those that receive less use would need to be blind. In fact, the identity of the instruments used would have to be kept not only from the listeners, but also from the players, as they may unconsciously (or consciously for that matter) play differently on one instrument than the other. Such studies are called double blind studies, and they protect the results from the kind of bias that could result from the listeners being unduly influenced in their evaluation by the folks administering the study. This may seem to be a little nit picky at first blush, but consider an evaluation performed in the mid 19^th century which compared Felix Savart's trapezoidal violin with a well regarded traditional instrument. In this case, the player played the two instruments from behind a screen, while a panel of critical listeners evaluated the tone. The listeners could not see which instrument was being played, so the evaluation was blind. But the musician doing the playing was a friend of Savart's (who, it could be argued, had something to prove) and so it is possible that he could have intentionally or unintentionally played the two instruments differently, such that the results of the test would be biased one way or the other. A double blind evaluation, where the musician didn't even know which instrument he was playing, would eliminate this bias. Note that it is generally well understood that it may be difficult or even impossible to keep the identity of an instrument from the player. But the fact that double blinding may be impractical or impossible in no way decreases its importance in producing unbiased results. While I'm discussing both the Pickering study and the Savart study, I should also point out that having study subjects (in these cases the listeners) in groups is a potential source of bias too, since each person's evaluation may be influenced by others in the group. It is not clear from either of the studies mentioned whether they were conducted in such a manner as to reduce this bias or not.

I would like to point out here that, although I am arguing that the study mentioned above by Pickering (as well as other studies to be mentioned later in this monograph) does not support the existence of a causal relationship between how much an instrument is played and how good it sounds, this is in no way a criticism of either the study or its author. If you read this paper you will find that the author himself makes no indication that the results of the study support the premise that we are discussing here. And it is certainly the case from this and other research by this author that he clearly understands the implications of the experimental methodology used in the study on any conclusions that could (or should) be made from it. Just as it would be unreasonable to criticize a builder who set out to build a house because the house didn't end up looking like the Taj Mahal, so it would be unreasonable to criticize a researcher who conducted a study with the intent of producing modest results because the results were modest. Most steps taken in research are baby steps – we learn a little at a time. I for one am grateful for this and all the research done in this area.

The recording of the string quartet and subsequent evaluation of the tone of the instruments in the performance is mentioned in a nice paper by Daniel Ling and Mead Killion⁴. The paper describes research performed by the authors where a small group of violins were played and recorded, and then these instruments were mechanically vibrated for 500 hours and then played and recorded again. The instruments were vibrated using a contact transducer connected to an audio amplifier through which the programming of a 24 hour classical music radio station was played. The before and after recordings were played to listeners for their evaluation, and the players themselves evaluated the tone of the instruments when they played them. Although no statistical analysis was performed on the data, the results appear to indicate that both players and listeners found the tone of the instruments to be superior following vibration.

If one can correctly assume that the mechanical vibration in the experiment effectively substituted for substantial playing, then at first blush the results of this experiment might seem to provide compelling evidence in support of existence of a causal relationship between how much an instrument is played and how good it sounds. But the methodology used in this experiment is such that such a conclusion is not warranted. As with the experiments described above, this one was not blind, at least as far as the players were concerned (it is not clear from the paper whether the identity of the recordings were provided to the listeners – it may be the case that there was no blinding in this study at all). Also, again at least as far as the players were concerned, this experiment did not provide a control population. As with blinding, controls are essential in an experiment if definitive conclusions are to be made from its results. Basically, what a controlled experiment does is compare two populations, one that has been subjected to some hypothetical influence, and the other of which is treated in every way the same as the first group except that it has not been subjected to the hypothetical influence. To use the experiment above as an example, that experiment would have been controlled if two groups of violins had been used, one which had been subjected to the vibration and the other of which had been otherwise treated exactly the same, except that they were not vibrated.

Why is a control group necessary? Because without it there is no way of knowing whether other factors, including in this case the simple factor of the passing of time, didn't produce the observed results (note that, although it may be unlikely that the simple passage of time could have made a group of instruments sound better, it is actually pretty likely that, in the context of an experiment to see if some hypothetically tone enhancing intervention does in fact enhance tone, the mere passage of time might increase the perception that the instruments improved). Again, no criticism of the study is made here. It is a good study, providing meaningful data, but for the reasons indicated it would not be appropriate to conclude that it provides significant support of existence of a causal relationship between how much an instrument is played and how good it sounds.

Another experiment along the same lines was done by Carleen Hutchins⁵. In this one a small population of instruments was subjected to actual measurements of response peaks, played for a considerable amount of time, and then tested again. In all cases the amplitudes of the response peaks increased after the intensive playing. If one can assume that the increase in the amplitudes of these modes would correlate with the perception of an improvement of the tone of the instrument (something of a stretch lacking other evidence, but for the sake of argument worth considering) then this study might certainly be a candidate for supporting a causal relationship between how much an instrument is played and how good it sounds. But here again there is no control population, that is, no group of similarly treated instruments that were not heavily played over the same period, and therefore no way of knowing whether or not these instruments exhibited the observed changes as a result of the playing. Again, correlation does not imply causation – just because they changed after being played in no way indicates that they changed because they were played.

Yet another similar experiment was done by Sobue⁶. Here the Q factor was measured for vibrational modes for a population of wood samples, then the samples were subjected to continuous low level vibration for a relatively long period of time. Afterwards Q was measured again, and found to be higher. Assuming that higher Q correlated with better sound, this could indicate that “playing in” could be beneficial to sound quality. There are some indications that higher Q correlates with higher sound quality. Work by Schelleng⁷ found a correlation between Q and the desirability of the species used for instrument tops, for example. But plenty of counter examples also exist. Excellent violins have recently been built of balsa, a wood that generally exhibits very low Q. And hardwoods generally exhibit higher Q than do softwoods, but they are not often used for instrument tops.

In the context of this discussion I'd also like to point out a former commercial enterprise which subjected instruments to vibration in an attempt to improve their tone. The principals of this enterprise have not published any papers in peer reviewed journals to my knowledge, but a pretty good description of what it was they were doing is provided by Rick Turner in an article he wrote for Acoustic Guitar magazine. Basically, instruments were secured to a shaker table and then subjected to intense vibration for a relatively short period of time. Instrumentation indicated that frequency response at some points of some instruments changed after the vibration, but there appears to have been no formal blind, controlled study that would indicate that tone of these instruments improved. Anecdotal reports of tonal improvements were reported but unfortunately even a vast quantity of anecdotal evidence is still anecdotal evidence and in no way substitutes for the results of a formal study with good experimental methodology. Other commercial products intended to "play in" instruments pop up from time to time but their efficacy is never supported by evidence from good well-controlled double blind studies. A group led by B.M. Clemens did in fact conduct the only double blind study of the efficacy of one such vibration device and found that treated steel string guitars did no better than untreated controls in human playing and listening evaluations.⁸ By the way it is interesting to look at the analyses of the controls in that study, which inticate a great deal of variation over the course of the experiment. If this is not due to variability inherent in the measurement techniques used, it could very well indicate that changes noted in other non-controlled studies are simply the result of inherent variability over time in the instruments.

The research referenced here and other efforts do indicate that there are reasons to believe that there may be a causal relationship between how much an instrument is played and how good it sounds. But to date (and to my knowledge) there does not exist a study with enough methodological rigor to support this conclusion. More research is clearly needed in this area. For those interested in reading more about this I highly recommend a survey article written by Weldert.⁹

References and Suggestions for Further Reading

1. Caldersmith, G. “Why are Old Violins Superior?”
   American Lutherie #14, p. 12.

2. Olsen, T. “An Introduction to the Stadivarian Mystique”
   American Lutherie #17, p. 6.

3. Pickering, N. “Old Versus New Instruments at Cambridge”
   Catgut Acoustical Society Journal Vol.2, #6 (Series II), p. 39

4. Ling, D. and Killion, M. “New Versus Old: Playing-in Instruments through Vibratory Transmission of Music to the Bridge”
   Catgut Acoustical Society Journal Vol.3, #3 (Series II), p. 42

5. Hutchins, C. “A Measurable Effect of Long-term Playing on Violin Family Instruments”
   Catgut Acoustical Society Journal Vol.3, #5 (Series II), p. 38

6. Sobue, N. “Effect of Continuous Vibration s on Dynamic Viscoelasticity of Wood”
   Procedures of International Symposium of Musical Acoustics, 1995, p. 326

7. Schelleng, J.C. “The Violin as a Circuit”
   Journal of the Acoustical Society of America, #35, p. 326

8. Clemens, B.M. et al “Effect of Vibration Treatment on Guitar Tone: A Comparative Study”
   Savart Journal, Vol. 1, #4, http://savartjournal.org/index.php/sj/article/view/22

9. Weldert, G. “Sound Enhancement of Musical Instruments by 'Playing them in': Fact or Fiction?”
   Europiano, 41-43, https://www.researchgate.net/publication/320041676_Sound_Enhancement_of_Musical_Instruments_by_%27Playing_them_in%27_Fact_or_Fiction