68 1,06 Received 27 July 1970 Effect of Material on Flute Tone Quality Jows W. Conn 310 Seatllacke Road, Ptburgh, Peansslrania 15235 ‘Thos koylens fetes of identi! insenal dimensions wers made of thin silver, heavy copper, and wood, respectively. Ther were paved out of sight of musically experienced absevers, who were asked to determine nly whether cones were alike ar diferent, No satatallysgnieant correlation between the listeners forse and the material ofthe instrument body waa found, Putats who played the aca sing an arrange INTRODUCTION ‘The role that the wall material plays in determining the tone quality of wind instruments has long been a subject of argument. Laboratory messurements. of rustained tones in artificially blown instruments! generally show no evidence that the wall material has ‘in appreciable effect. In spite of these experiments, instrument makers, players, and listeners continue to insist dat the nature of the wall material does have an effect on the instruments! cound. Metal clarinets ace considered suitable only for use in schoo! bands. ‘The silver flute is the accepted standard in. most countries today, though symphony players insist on retaining wood for piccolos. Not only: the nature of the material, but the thickness of the wall is considered important. Many makers of quality Autes strive to snake the wall as thin as is consistent with mechfinical requirements, At the sime time, high density is con sidered desirable. Varese wrote a composition entitled “Density 21.05 tu celebrate te platiaune Gute played bby America’e most famous flutist, Georges Barrer ‘Thus, there is a marked discrepancy between the results of laboratory measurement and the opinions of users. IL has become apparent? that harmonic content of steady’ tones is not sufficent to characterize an instrument and that listenera generally depend on transitory effects for identification. Moreover, from the player's standpoint, the “responsiveness” ‘of the instrument, that is, the ease with which sounds ean be produced and modulated, is an important aspect that ray be iuependent of te chatatter of the resultant sound, This also brings into question the pertinence of laboratory measurement of steady tones. 520 Volume 49 Number 2 (Port 2) 1971 ‘again a previously slated insirument [In all of these discussiows (which, as Backus* points cot, probably: started in early Stone Age circles with fssertions that 9 flute made from a human thigh bone had a much better tone than one made from a stick of bamboo) there is an almost complete absence of reports of controlled objective experiments directed toward determining Uhe degree to which Fisteners atl pagers ould, in fat, dcriminate between instruments rade of various materials. It therefore seemed worthwhile to carry out the two experiments described below, which attempt to elicit evidence toward this questi both for listeners and players 1. CONSTRUCTION OF THE FLUTES. ‘Three heslens Mules were consicucteal of silver, ‘copper, and wood. These were all of internal diameter 1% em. Thesilver tube was piece of stock from sehich professional silver futes are made and had a. wall thickness of 0.036 cm. The copper-tube wall thickness ‘was 0.153 cm, making it more massive than the slver tube bya factor of 4.0. The third tube was of grenadilla, a seond frequently emplayed for wondwinds The wal thickness was 0.41 em, a typical dimension for wooden ‘utes. It weighed 1.7 times as much as the silver tube Each tube was $2./ cm long and had affixed to it a short head of Deliin plastic extending another 5.1 en to the center of the mouthhole. The three plastic heads were shaped to the same internal dimensions by a specially made tapered reamer. The diameter at the mouthhole was 1.75 em, The mouthhole itself was shaped by using a paraffin impression of the mouthbole of @ professional model fute and casting epoxy resin around this pig a fill a anger hale previously boredEFFECT OF ig A Flutes of sitver, commer, an sunpeste Tor sing in the head, The far ends of the silver and copper tubes swere enlarged in thickness by’ a short outer cvlinder to ake the outer diameter of the end equal te that af the wooden tube. The acoustically important dimen: sions of all three instruments were thus identical within 0.01 em, ‘The instruments’ est mode sounded 398 Hz, approsimately Gy, the iniddle of the orlesteal Mute's The ease of sounding and the power and tone judged excellent by Sutists who tried them, II, EXPERIMENT WITH LISTENERS ‘The frst experiment waa directed toward finding out whether listeners could discriminate among. the ine struments when they were phived by the same Twenty-seven observers participated simultaneousl) in the listening trials, OF these, 20 claimed to be pro: fescional or shilled sn 13 of th utists, Seven claimed little training or experience, The trials were conducted with the author playing the flutes behind a small screen made of aluminized mylar 750 wm thick, whose sound transmission extends Well hheyond 15 kHz, The tests were conducted in « clase room that was acoustically treated and often wsed for chamber music performan In each trial three identical musical phrases were played, Two of these were played on the sume instru- nient, The observers were asked to mark the position ised. This tech nique was chosen hecanse it avoided any reference to ‘better or “worse” tone quality, and did not re ny. presumptions as to how a wood or silver fute should sound. Thirey-sh where the different inatrumen ch ials were made. In the first six the phrase consisted cimply of « single tongued MATERTAL ON FLUTE TONE and sustained note, the first mode of the instrument, Only the silver and wood flutes were vsed, the order of playing in a single teial being varied randomly. In three cases the wood flute was unique, in three enses the silver. A similar set of six trials compared the silver and copper tIutes. ‘The observers were told which pir wvas Leiny compared in each set of sis trials, The third and fourth groups were similar except that the second mode (approximately 800 Hz) was sonnded, while the phrase for the fifth and sixth groups consisted of a legato transition from Gy to the octave above, and Feturn, Additional information collected from the ‘observer was his musical backgwouind ats & Muti, other msician, or listener and his own rating of himeelf as fessional, skilled amateur or student, or person with litte training or experienc. ‘The results obtained may be examined statistically iu many wags, For completeness the results are given in two tables, Table let the individual scores (number of trials out of the entire 36 in which the observer identified correctly the unique instrument) in. two groups, one for those who claimed to be skilled or trained muscilly: and one for those of litte taining On @ purely random hypothesis, one expects success on one-third of the trials, so the expected score is 12. Scores of 27 observers forthe entire se of 36 tial Mean se Sage wel TB TG i 129405 {asi obser itu ‘The Journal of the Acoustical Society of America sarow ‘Taste TE. Distribution of votes on 36 tale WS, and © indicate the postion of the fue played ance ducing’ tal and imites i nas wood, silver, oF eOpDeE Siver va wood Silver vs copper Fundamental Ere ncR ca ee rhs inte anol Tei0 th Saino tt tMirst-cxs ttl Ese Win! oes obara tele) Teo HH eoh HnS Ha CID octave ce crente ty Se convo st eit etitin nt cto eM aise eee 6 omy 8 5 ou 8 2 6 Combination ey ee ce ee) Ronis 2 isiicls PSS VA i | Lge 7h et “The standard deviation of single score should be 28 and the mean for 27 observations 12-40.8. We see from Table 1 that the unskilled observers scored slightly (but insignificantly) higher than the skilled observers. The oversall mean score has the expected standard dlviation, implying thut no outstandingly successful chcervers were precent, while the mein score of 121 is slightly higher than the expected value of 12, The nature of the trials, however, was such that statistical independence of the 972 observations is not sssured, Fath of the 27 observers was listening ¢@ the fame eet of trial, If the performer, for example, were accidentally to play one note poorly. all 27 observers might be expected to score this tril the same. The statistical expectation that a playing variation acct dentally corresponds with the unique ‘ite is again fone thied, but now the standard deviation is. much higher, since there are only 36 esther than 972 triale ‘Thus we need to look ia more detail at the correlations between markings before deciding whether the slight departure from 12 is significant. Table II displays the results of the tests, Each ow of theee numbers represents atrial. The leter preceding 4 number designates the position of the unique ute for this tral and its material, The numbers are the umber of petsons in the 27 ubscrvers voting for the Corresponding poston Hf we total the columina, we find that obsccvers cast 275 votes for the frst insieument played, 880 for the second, and 367 for the third. The spread in these values more than five standard deviations fron the expected 324. Since the unique tute appeared an equal number of times in each eolans, this fs not a result of the material Tt strongly pointe to the fact often ob- served that the order of playing is important in deter- mining an observer's choice; Saunders, for example, found in listener tests of preference for Str 522 Volume 49 Number 2 (Port 2) 1971 COLTMAN Violins that the jury voted most often for the sevonl Instrument played, regardless of its type. By applying a chi-squared test to the entire set of dats, using a value of nine as the expected value, and assigning 2 degrees of freedom (df) to each tral, a strong rejection (p<0.0003) of the random hypothesis is obtained, Thus the votes are grouped more strongly than one could expect from pure chance. However, if awe take the majority vote at each tral asthe observer's selection (counting ties as oncshalf}, we find the ob- servers as a Jury voted correctly in Just 12.8 out of 36 trials, Thus the grouping ia nota result of the material Tt is likely that, in addition to the effect af order, Perceptible variations in playing contribute to the grouping. This is supported by a comparison of pairs of trials in witch the position of the unique fhute was the saine, AIL observers questioned amitted ty great dificulty in making a selection in almost all eas that these playing variations. were not la prevent us, however, from applying a simple Bernoulli {est on the over-all scores, ‘An analysis of variations carried out with respect to order of playing, plrase used, copper versus worrd, al their interactions showed none of these tobe significant, at the 5% level, in determining whether an observer correctly ientiied the unique instrument. ML, EXPERIMENT WITH PLAYERS In order to obtain an objective measure of a player's ability 10 dlseriminate between instruments, it is necessary to remove any elues to ita identity other than those associated with the sotind produced. ‘This was accomplished by the arrangement shown in Fig. 1. The three tlutes described in Sec. I were mounted sym- retecally with their axes parallel to a central rod and Ueie heals projecting through & plastic shiekd affixed to the same rod, The player grasped the rod rather than the flutes, and when his head was in the normal blowing position, he could not see the tubes formin the bodies of the instruments, A counterweight inside the shield restored the imbalance due to the differing weights of the instruments, 20 that in all visual and tactile clues to ide In making a test, the subject was permitted to blow the instruments in’any order, rotating the fixture to change flutes, but otherwise not changing position, He was asked to select an instrument whose tone he pre ferred (or thought he could identify again), and the experimenter noted the selection. The subject wis then told to spin the fixture so as to lose the identity, snel then, by blowing at will on various instruments, find his original choice. When he was sisted he had found it, he notified the experimenter, who scored one point for success or zero for failure AA test consisted of five attempts to find the originally selected instrument, so that test scores can vary from zero to five. In the'first test, the flutist was restricted pects ty were remnvedEFFECT OF MATERIAL to sounding the fundamental, and in the second the octave, but he could tongue of attack the note at will, Nine auch teats were carried out with four different florists, all reasonably skilled performers. ‘The four Autists scored, as follows: 0,1) 07: 0,9: 21 (On a random hypothesis, the expected score Is SX}=L7, und the mean of nine such scores would have a standard deviation of 0.33. The experimental results give a mean score of 1.3, barely one standard deviation less than the expected mean. We can expect, a score of four or higher to occur in one test out of 22, its weusrence in a set of nine is not surprising. ‘The Hlatist who seared this four gor scares uf yore andl one fn the next wo tests, If we assume that not only’ the first selection, but also the flutes named as “identical” constitute “pre feed” Mates, we find silver nained 21 times, wood 19, cul copper 1 On a random hypothesis, we expr value of 18 in each category, A rhisquare test an the observed numbers produces «value for chi-square of 1, well below the 9.9¥ required for significance with 2k. These is thus no evidence heve of anything but random relection IV, CONCLUSION No evidence has heen found that experienced listeners or trained players can distinguish between flutes of like mouthpiece material whose only difference ON FLUTE TONE is the nature and thickness of the wall material of the body, even when the variations In the material and thickness are very marked. OF course, itis possible that individuals exist’ whose discriminatory senses are keen enough to find a distinction, but ifso, they are certainly not common, Moreover, the results suggest that even careful attempts to produce identical sounds on the ‘sume instrument produce variations that are more perceptihle than any that might he associated with the material, One player did, correctly, point out that one of the three instruments appeared at first to be slightly flat ‘Thy effect io due to the high thermal inass of the heavy copper tube, which causes it to warm up more slowly than the others. ‘This is an example of a reason to prefer certain materials fr flute construction, and there are many others. Tone quality or ease of response are hot, however, among them, ACKNOWLEDGMENT Thanks are due to Robert Hooke of the Westinghouse Research Laboratories for assistance in design of the experiments and in the statistical analysis of the results. 21. Racks, J. Aeatst Soe. Ammer. 36, 1881-1887 (1064) sodas and °C. Hundley, J Acoust Soe. Amer. 39, S12, Kinet and ML, Mathews, Phys: Today 22, 23-38 (1969) 14° Gaus, The Acsuaticl” Foundations) Macs (Norton, Nese Var, 149), 208 Sie Sounders Sound 1, 1S 1962) The Journal of the Acoustical Secety of America 23