My days of rearing small children ended a long time ago, but I do have grandchildren and so have become acquainted with a little gem of a children’s story entitled The Puppy Who Chased the Sun. It’s about a little dog named Wilbur. One morning Wilbur woke up earlier than usual. He was hungry, so he barked. To his astonishment, at that very moment the sun began peering up over the horizon. Frightened, he barked again. Lo and behold, the sun kept right on coming up. So on that fateful morning, Wilbur came to a momentous conclusion: his barking caused the sun to rise. This made Wilbur feel very proud and powerful…so much so that he began shunning his lesser canine comrades in the neighborhood.

Then one morning Wilbur barked…but the sun didn’t come up. He barked again, and again and again, but still the sun didn’t rise. All that happened was that Wilbur got wet, because it started to rain. Wilbur became very depressed, so depressed that he overslept the next morning. When he did wake up, the rain had stopped, and there was the sun high in the sky. Thinking this over, Wilbur arrived at a revised conclusion: his barking didn’t cause the sun to rise after all; something else was responsible. Fortunately, Wilbur’s self-esteem suffered only briefly, and soon he was back to romping with his old doggy buddies.

Some time ago, after many years of making and repairing guitars, it began to dawn on me that despite my experience, reflection, reading and listening to others I really understood very little about what causes guitars to do what they do in producing musical sound. Listening to other guitar makers didn’t help much. On the rare occasions when the shroud of secrecy that often surrounds such exchanges among luthiers parted a little bit, I typically found myself wondering, “Does he really know that’s so?…Is it proven?”  The limited literature I am acquainted with in guitar acoustics wasn’t much assistance, either. As a practical, producing luthier concerned primarily with making guitars that real-world people actually want to own, I found this literature to be at best only remotely connected with pragmatic steps I could use to improve my instruments and at worst recondite technobabble and mental self-indulgence.

In the course of ruminating about this state of affairs over time, I came to realize that guitar makers, by no means excluding myself, are uniquely prone to two classic intellectual lapses. The first is superstition, a false cause-and-effect between events or circumstances, such as we see in our friend Wilbur. The second is known in philosophical terminology as reification, which is defined as regarding or treating an abstraction as if it had concrete or material existence. Wilbur, that smart (and honest) little superdog, eventually realized and faced up to his error and once again became a regular canine guy.

I also came in time to the sobering conclusion that my own guitar making (and, as near as I can tell, everybody else’s) is, if honestly regarded, largely a collection of practices and procedures with little or no attachment to proven tone consequences. Does that mean that ignorance is inevitable?  Not if we step back and take a fresh and honest look at what it means to really “know” something about the reality we live in and work with.

A major stumbling block to accumulating reliable, reality-based knowledge about how a guitar functions and the possibilities for changes which might improve its tone can be found in a confusion of language. For ordinary language purposes, we find the following definitions of the words hypothesis and theory (Source: The American Heritage Dictionary):

• Hypothesis:  1) A tentative explanation that accounts for a set of facts and can be tested by further investigation;  a theory. 2) Something taken to be true for the purpose of argument or investigation;  an assumption. (boldface & italics mine. )
• Theory:  1)a. Systematically organized knowledge applicable in a relatively wide variety of circumstances, especially a system of assumptions, accepted principles, and rules of procedure devised to analyze, predict, or otherwise explain the nature or behavior of a specified set of phenomena. b. Such knowledge or such a system. 2) Abstract reasoning;  speculation. 3) A belief that guides action or assists comprehension or judgment:  rose early, on the theory that morning efforts are best; the modern architectural theory that less is more. 4) An assumption based on limited information or knowledge;  a conjecture.

These definitions suggest that the words theory and hypothesis may be used virtually interchangeably. Science, however, uses these words in a much more specific way, and the Oxford English Dictionary, while it acknowledges the above common language confusion, helps us get closer to this:

• Hypothesis:  …esp. in the sciences, a provisional supposition from which to draw conclusions that shall be in accordance with known facts and which serves a starting point for further investigation by which it may be either proved or disproved and the true theory arrived at.
• Theory:  A scheme or system of ideas or statements held as an explanation or account of a group of facts or phenomena; a hypothesis which has been confirmed or established by observation or experiment, and is propounded or accepted as accounting for the known facts.

According to these definitions, Einstein’s Theory of General Relativity was not a true scientific theory at all but only a mathematical one until it was confirmed by the now-famous observation at the eclipse of the sun. Indeed, Einstein himself insisted that it be so.

By comparison, the luthiery craft routinely treats its information with embarrassing promiscuity, in accordance with common language definitions. Our craft is awash in undisciplined speculation, unconfirmed hypotheses, and even outright superstitions.

For example, it is widely believed among both guitar makers and players that playing a guitar regularly day after day will “break it in” and improve its tone over time. In reality there is no evidence, at least none that I am aware of, for a true cause-and-effect relationship between playing an instrument and improving its tone. That the guitar improves over time as it is played by the musician may be true, but the cause of the improvement could just as well be merely keeping the instrument constantly in tune, thus maintaining a consistent tension on the soundboard, and storing it in a room temperature environment. The belief that playing the guitar accounts for its improvement may, in other words, be a superstition just like Wilbur’s barking to get the sun to rise.

Much more farfetched superstitions have been indulged in the past. Some people have tried storing guitars overnight in front of speakers playing sweet classical music, hoping the instrument would get the hang of it. A few have even imagined that placing a rattlesnake’s rattle in the sound box would promote tonal improvement. No doubt some guitars which “listened” to classical music or provided a residence for a rattlesnake rattle did indeed improve.

At this point, I am convinced that a grand theoretical paradigm for guitar is unachievable, and I long ago abandoned the hope that I might ever come up with any such thing, but a more modest, limited and honest application of elementary scientific method is well within the reach of guitar makers and has the potential to bring about real improvements in a builder’s instruments as well as adding to sound knowledge about luthiery. In the following account, I will present a simple type of protocol which I have used to test specific hypotheses on my own guitars. This protocol resembles in some respects the one typically used in investigating the effects of prescription drugs, although obviously with a much smaller sample.

LIMITED VARIABLES – One of the first conditions which must be met in attempting to gather reliable scientific information on guitars is to limit the number of variables in an experiment. This can be frustrating to a builder because he is probably producing ideas at a much faster rate than guitars. The fact is, however, that it is not possible in science to test two or more related variables at one time. The single variable may well be a complex, global one like a major soundboard redesign, but then your test results can only tell you something about this variable taken as a whole; you will learn nothing in the experiment at hand about conceivable subvariables, such as the positioning of specific braces or variations in the way the soundboard may have been graduated. It may be possible to test some tone variables, such as brace wood removal, more easily on a single guitar with access devices like a trapdoor in the guitar body. However, a builder/experimenter must be careful that devices intended to facilitate experimentation do not end up constituting unaccounted-for variables in and of themselves. The reality is that most experimental situations will require that the builder make a guitar to test each variable.

But not just one guitar. A second requirement for a sound experiment is that it must make use of a control. Basically, the control in a luthiery experiment would be another guitar in which every feature which could conceivably bear on the experiment is exactly like the test instrument, or as much so as you can make it, except for the variable being tested. Without a control, it is impossible to learn anything about the cause-and-effect relationship between the observation results and the variable. The difficulty for the luthier in doing this is obvious and staggering; it is a monumental challenge in the control of materials and techniques, to say nothing of the enormous amount of resources required. Even with a best effort at constructing a control, it is hard to be certain it does not contain some unaccounted-for variable, especially if the results of the experiment turn out to be unexpected. For reasons that will be apparent below, the control guitar will more than likely need to be constructed at the same time as the test guitar, which can also make it easier to maintain constants in the control instrument.

TESTING PROCEDURES. – Assuming you have your experiment prepared in accordance with the above considerations, the next issue that must be dealt with is testing procedures. Testing for tone presents numerous problems and conflicting possibilities and offers few clear-cut options and conditions which are obviously and decisively superior to others. The basic problem is that, when you set a test condition to obtain a particular form of information, you may well be sacrificing another kind of information. An analogy is the classic one from particle physics known the Heisenberg uncertainty principle, which says that the more you know about a particle’s position, the less you know about its momentum.

Testing procedures can be roughly divided into technological information, usually from electronic measurements, and reports from human listening. Both types have their own advantages and disadvantages. Technological information comes in various forms, from graphs and screen displays to patterns of particles on guitar tonewood components created by exciting the component with an audible signal generator, or even a voice. Technological information is typically very precise and exact, much more so than listening reports, and the results can easily be permanently stored in raw form for later analysis.

The biggest disadvantage to technological information is that it is useful only at very low levels of complexity, compared to listening reports. An electronic device, lacking true cognitive capabilities to say nothing of emotional capacity, cannot “listen” to “music” in anything like a human sense. Technological data is never anything more or less than simple raw data; the ability of machinery to engage in summary and generalization on any but the most basic level is still a technological fiction. Moreover, what an electronic device “hears” may be very different from what the human ear perceives because of peculiarities of the human sense.

Another problem with technological measurements is that they may not be measuring anything related to variables which directly affect a guitar’s tone. I have yet to see, after more than one attempt to investigate this technology, how Chladni patterns, sometimes referred to as “glitter patterns” because of the material used to make them, created on tonewood components can be employed usefully to modify guitars so as to actually improve their tone.

The human senses, by comparison, and not hearing alone, are relatively much less exact and precise in an absolute sense. Moreover, the human memory can be very fickle and undependable in its retention of sense data as absolute information. Auditory memory, in particular, is notoriously short-term. This presents storage problems for listening information and makes it necessary to do summarizing and analysis tasks on the spot as listening takes place and to record these results promptly lest significant information be lost.

On the other hand, the human senses, all of them apparently, are capable of very high levels of precision in tasks which make use of comparative discrimination. Hand a person a piece of wood about an inch thick and ask him to use his thumb and forefinger to tell you exactly how thick it is, chances are his answer will be quite far off, but give him two pieces of wood very close in thickness, and he will be able to easily tell you which is thickest. Likewise, the human ear can detect very small differences in volume and/or timbre between two guitars if they are played serially, both single notes and short segments of music. On the other hand, human prejudices, often less than predictable, can influence listening reports. Some have observed, for instance, that if two guitars are played serially, the second one will probably be reported to be louder, maybe even “better. ”

It is also worth mentioning that loudness as perceived by a listener may be quite different from a decibel reading recorded electronically. This is because partials other than prime in certain arrangements in a tone are additive as perceived by the human ear. Thus it is possible for a human listener to perceive one guitar as louder than another, even though both produce the same decibel reading. Sometimes different listening environments accentuate this and cause listeners to report, for example, that a guitar “projects” well in a concert hall, even if it is not notably loud close up.

Both electronic measurement and human listening reports also share a common problem, and a serious one, which has to do with the translation of information. There is no symbolic/representational language of tone equivalent to the written languages for verbal speech or music. Taking the symbolic representations of speech and music in verbal writing and musical notation, we can imagine with great accuracy, depending on our degree of mastery, the speech and music being represented and use these representations to imagine and reproduce audible speech and music, but try imagining a tone by looking at an oscillogram. A person with lots of training and experience with this form of information might be able to muster in his “mind’s ear” a very crude approximation of a tone so represented, but for most of us the translation is essentially impossible.

The situation is only marginally better with attempts to represent tone with words. Serious guitarists are acquainted with terms like “warm,” “dark,” “bright,” or even “fat” or “thin” as they are used to describe guitar sound. What all these words have in common is that they are metaphors based on words originally meant to apply to other senses than hearing. This increases the distance, mentally speaking, between an audible or imagined tone and the representation thereof.

Nevertheless, it has long been my conviction that in most instances the use of the human listening report, with all its inadequacies, is the best option for tone testing aimed at obtaining information useful to a luthier. Listening reports can involve many evaluators, not just a few experts in the language of technical symbols. The metaphorical words we use may be distant approximations, but they are the closest thing we have to a common language to describe tone. Human listeners can also evaluate tone in the context of actual music, which is, after all, the goal of the enterprise. Even the subjectivity and prejudice which are an unavoidable risk in human listening reports can be accounted for in most instances and perhaps even turned to advantage if handled correctly.

TWO EXPERIMENTS – To illustrate the application of the principles outlined above, following are descriptions of two test situations which I conducted recently using instruments made in my workshop. Both of these tests show something about the possibilities opened up by more rigorous testing procedures and also the ongoing problems encountered when truly definitive results are sought. In both cases, my hypotheses were at least put in doubt, and perhaps shown to be incorrect; in other words, there was an element of inconclusiveness in both cases. Naturally, this was disconcerting, although more so with regard to the first test which required me to conclude, “I still don’t know the answer,” rather than just “I may have been wrong. ”   However, on the positive side, the results led me to look at another variable which I had not adequately accounted for. Testing this variable at some future date may help clarify the results of these experiments.

The first test was directed to the question, “What difference does scale length make?” For several years, I have been making instruments with scale lengths varying from 640mm to 665mm. Fifteen years ago all my guitars were made with scales in the long end of that range, and the scales have gotten generally shorter since then. In the past three years I have made several instruments with 640mm scales for my distributor in Japan and for a small number of other customers. Prior to conducting this test, my informal impression, my hunch, was that instruments with scales from 640mm to 650mm were pretty much the same in terms of volume. If there were timbre or sustain differences which might be attributed to scale length, these had not become apparent to me in the normal course of my guitar building. My hypothesis, in other words, was: Scale length makes no difference in the guitar’s tone within 640-650mm scale range.

Here are material specifications for the guitars in the first experiment:

Guitar serial#	149	150
Scale	650mm	640mm
Soundboard wood	Cedar	Cedar
Back & sides wood	Brazilian RW	Brazilian RW
Density factor/weight
...Soundboard	38df	37df
...Back & sides	89df/317g	86df/306g
...Neck	65df	68df
Bridge weight	16.3g	14.4g

The density factor, which is a measurement I use primarily to compare component materials for several purposes, is calculated as follows:

(weight/length x width x thickness) x 100.

The back sets were very closely matched to each other and probably cut from the same billet. The soundboards appeared to be consecutively cut pieces from the same billet. Density factor differences in these components can probably be accounted for by imprecision in measurement. The soundboards were graduated to a nearly identical finished thicknesses for both instruments. No attempt was made to intelligently graduate the backs; they were all scraped and sanded in the completion process in the same way. Basic construction features were identical for both instruments, including bracing patterns. Thus all component specs were held constant except for scale length…and bridge weight (unintentionally), but more on that later.

Four listeners, including me, with varying levels of music training and experience with classical guitars, listened to these guitars. The consensus was that both guitars seemed to have equal loudness but that guitar 150 had a slightly brighter, more “punchy” tone with a little less sustain. Listeners agreed that the differences were very small and that these differences required hearing short segments of music in a comparative discrimination setting. I accepted these conclusions at the time without further critique because they seemed pro-intuitive and “reasonable. ”

The second experimental setting had a rather more elaborate agenda and involved three guitars, again all made by me. However, the protocol actually constituted two separate experiments. One was designed to compare the effects of backs made of Brazilian rosewood and Indian rosewood. The second test was designed to compare two different soundboard bracing patterns: Pattern A, which I have used on all my guitars virtually unchanged for several years, and Pattern B, which is a modification of pattern A. The hypotheses were:

Experiment 2: If wood density and thickness are held constant, backs of Brazilian and Indian rosewood will produce the same tone results; guitars 165 and 166 compared.

Experiment 3: Bracing pattern B will produce a tone which is “darker” and “mellower” than Pattern A; guitars 164 and 165 compared.

Here are material specifications for the guitars in these experiments:

Guitar serial#	164	165	166
Scale	650mm	650mm	650mm
Soundboard wood	Cedar	Cedar	Cedar
Bracing pattern	B	A	A
Back & sides wood	Brazilian RW	Brazilian RW	Indian RW
Density factor/weight
...Soundboard	38df	38df	39df
...Back & sides	94df/295g	90df/283g	89df/279g
...Neck	58df	58df	57df
Bridge weight	16.4g	15.8g	14.9

The backs on 164 and 165 appeared to be from the same flitch; the difference in density factor is probably measurement deviation. The soundboards were all clearly from the same flitch, possibly consecutive slices.

All three guitars were listened to in one session in a home living room by six listeners, including me, with guitar listening experience which ranged from a musically educated but novice guitarist to one highly talented, expert guitarist, who did all the playing for the test. We listened to numerous short segments of music, each played very consistently on each instrument selected randomly and in some cases re-selected to confirm results with a second hearing. The music segments were chosen to emphasize various parts of the instruments’ tone registers and various tempos and loudness levels. I did all the questioning to elicit the responses as the instruments were being played and took notes on the responses. However, I did not inform the listeners of my hypotheses until the testing was concluded.

In Experiment 2 listeners were in unanimous agreement that 165 and 166 had a different timbre, and there was consensus that 165 was slightly warmer/darker than 166, although no one thought 166 was a notably bright guitar. Two of the listeners thought 166 sounded slightly louder; no one thought 165 was louder.

In Experiment 3 listeners were unanimous that there was a small difference in timbre, but no difference in apparent loudness, between 164 and 165. However, there was no consensus that terms like “dark” or “bright” described the difference, and no other suggested metaphors attracted a consensus. Opinion was also divided on which instrument listeners “liked better” on a global, subjective basis.

What conclusions can be drawn from these experiments? The results of Experiment 1 are clearly flawed by my inadvertent neglect of the bridge constant, for no better excuse on my part than inattention due to the normal distractions of production work. The results of Experiment 2 are probably the most definitive of the three. However, I would like to confirm these results, if for no other reason than that they were counter-intuitive, but also because the experiment has to do with a costly option for customers. I still haven’t decided what to do, if anything, with the results of Experiment 3 and the alternative soundboard design. A decision would obviously be easier if the results had been more conclusive, either favoring or disfavoring the alternative design. Before attempting to re-test the hypothesis in Experiment 1 or confirm Experiment 2, there is clearly a need to test the bridge weight variable more rigorously. This experiment would be valuable in itself because it involves a variable which could potentially be used to modify the tone of my guitars but also because more reliable results could indirectly shed more light on the results of Experiments 1 & 2.

After reading these accounts of guitar making experiments, the reader has no doubt come to appreciate the difficulty and high cost, in terms of luthiery resources, of this kind of inquiry. Such resources obviously cannot be wasted on frivolous questions. I would never conduct a rigorous experiment to try to find out what difference in makes to substitute a rubber band for the E-string. Another issue is the degree of certainty a luthier requires with regard to the various materials, procedures and techniques used in the course of building a guitar. Clearly not everything requires the certainty level of a rigorous experiment. This is why a luthier’s intuition will always play a key role in progress in the improvement of guitar tone.

However, I have found that my basic thought process about guitar tone issues, even in areas which will remain in the realm of my intuition and never come to be subjected to a formal experiment, has been changed permanently by my attempts to conduct formal experiments. I now think in terms of a hierarchy of certainties. I am also more alert to my own lapses of superstition and reification than before. Finally, I am sure my guitar making career will see more attempts to use basic scientific method to clarify important tone issues in the future.

(This article appeared in Soundboard, Spring 1998. )