THE JOURNAL

OF THE

ACOUSTICAL SOCIETY OF AMERICA

Volume
17
/• Number
3
JANUARY- 1946

The Mechanical Action of Instruments of the ViolhaFamily

F. A. SAU,•OEaS
ProfessorEmeritus,Han•rd University*
. . (ReceivedAugust17, 1945)

INTRODUCTION
responsecurves of several famous violins were
comparedamong themselvesand with those of
EARLY
tenyears
agoa group
inthe
Harvard laboratoriesbegan to study the new violins; the filtering action of the bridge
mechanicalbehavior of violins, in the hope of was discussed;a suggestionwas made in regard
finding out how the best ones differ in their to the meaning of the term "carrying power" so
vibration from those which are recognizedas often used by violinists; and an account was
inferior. It was hoped that the results might given of the resultsof a test of the ability of an
lead to improvementsin new violins. Sincethis audienceto pick out the tone of a Stradivarius,
work hasspreadover severalyears,and noteson when it and two new violins were played in
succession behind a screen.
it have appeared in various places,we should
begin with a brief summaryof it. The next publication swasa shortonein which
the excellent old instruments of the Curtis
The first publication• includedan accountof
an analyzer developedfor this work; the em- String Quartet were shown to have response
barrassingvariety of harmonicpatterns disclosed curves very like those produced by copies re-
in violin tones; the response (or frequency) cently made of them by a skilled craftsman in
curves obtained by means of the analyzer from Philadelphia. An account was included of tests
many new and old violins; the curves of total conducted by the Quartet on the judgment of
intensity against frequency;the mechanicaleffi- eight audiencesin trying to discover which set
ciency of violins; the effects of variations of of instruments was being played on, behind a
bowing,and of moisture,string tension,etc.; the screen.A brief discussionof the psychological
friction betweenbowand strings;etc. This article effects involved was added in the attempt to
raisedmany unansweredquestions. explain why thesejudgmentsdependedmainly
on the order in which the two sets of instruments
It was followed by a papers in which the
were played.
* Now Visiting Lecturer, Mount Holyoke College,South A note in the Year Book of the American
Hadley, Massachusetts.
• J. Actus. Soc. Am. 9, 81 (1937); referred to as A below. a In Overtones (April 1940), the journal published by
• J. Frank. Inst. (Jan., 1940); hereafter referred to as B. the Curtis Institute of Music in Philadelphia.
169
170 F. A. SAUNDERS

Philosophical Society (1940) mentioned some somewhatuncertainby experimentaldifficulties,

experiments in which Jamha Heifetz took part, which tended to raise the values. This error
which were designed to measure and compare, could have been removed by increasing the
from high speed records of actual playing, the duration of each record, but this might have
periodsof growth and of decay,whenan excellent introducedother errorsdue to unevenbowing.
and a bad violin were used in succession; hence In the new method the violin body is excited at
to securean indication of the value of the damp- all frequenciesthrough the foot of the bridge
ing in the wood of theseviolins.The resultsof which is under the G string, through forces
these experimentsare given later on in this actingon a wire which takesthe placeof the G
article. string.If the responsein thismethodisintegrated
The last publicationwasby Watson, Cunning- by semitones,a curve is obtainedwhich agrees
ham, and Saunders4 and gave an account of a well with the curve of the first method in the
new and superior method of obtaining response middle rangesbut is lower at both extremesof
curves directly, without tone analyses, by ex- frequency. These differencesare not very im-
citing the violin electromagneticallywith a force portant, since the object of the experimentswas
oscillatingwith a single frequency, which could to comparegood violins with bad, and this can
be varied over the wholerangeof the instrument. be done well enough if both are measuredby
All responsecurvesrecently obtained have been the same method. It is still possibleto compare
doneby this method,but this part of the experi- the responsecurve of a violin measuredby the
mental work came to an end in 1941. The new
secondmethod with one measuredby the first,
method included direct measurements of the
by making allowancesfor the differencesdis-
decay of pure tones, from which the damping covered by measuring several violins by both
at a few selectedfrequenciescould be obtained. methods.
One of these frequencieswas chosento coincide
All the responsecurves have been integrated
with that of the main vibration of the air inside
by planimeter to find the average intensiW of
the body; the others coincided usually with
response in decibels(db) abovean arbitrary level
natural vibrations of the body, which were un-
in differentregionsof frequency.If one chooses
affectedby the insideair.
small ranges of frequency the results vary so
A preliminary report of the latest work was
much as not to be useful; if one uses a large
made at the meeting of the AcousticalSocieW
range one may misssomeimportant peculiarity
in May, 1943. The presentarticle coversthis
work to date. of the violin. A fair compromisewould be to
chooseoctave intervals. This procedurehas been
RESULTS FROM RESPONSE CURVES OF STRAD$ adopted exceptat the two extremesof frequency.
BY THE NEW METHOD
In the region from 200 to 350 c.p.s. the emission
The responsecurves obtained by the old and of sound in all good violins is largely due to the
vibrations of the included air. It seemed wise to
new methodsdiffer in certain respects.The old
curves are based on observations taken a semi- treat this interval of somewhat less than an

tone apart, while the new curvesare continuous,
and includeall frequencies.Thus a narrow peak
of responsewhich might be missed by the first
method is caught by the second. This is im-
portant in the high frequencyregion, where the
overtones of the violin body are so crowded that
two or three may occur inside the interval of a
semitone.
In the old method, the intensitiesobtained for 349 c.p.s.; II from 349 to 784; Ill from 784 to
two or three of the lowest tones were made
4 R. B. X,
Vatson, W. J. Cunningham,and F. A. Saunders,
J. Acous. Soc. Am. 12, 399 (1941); referred to as D.
octave separately. The interval next above was
made a little more than an octave, to make the
first two equal to two octaves. Above 4000 c.p.s.
the emissionusually drops sharply. In order to
trace this drop more in detail it seemedxviseto
split the upper octave into two shorter ranges.
The ranges used in the violin tables to follow
are denoted by I, II, etc.; I covers from 196 to
1568; IV from 1568 to 3136; V from 3136 to
4186; VI from 4186 to 6272. The intensity above
6300 is relatively negligible in good violins.
MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIN FAMILY 171

TABLE I. Stradivarius violins. Average intensities (in eachviolin was the same.This is approximately
db) in different frequencyrangesmeasuredby the new
method. the same as reducing each violin to the same
loudness. Most of the sounds measured were
loud enoughso that the loudnesswas approxi-
Violin I I1 Ill IV V VI
mately proportionalto the intensityin db. The
Strad B
"Halir" (1694) 15.2 25.0 24.6 26.5 20.8 7.1
purpose of the measurementswas to compare
Strad D the distribution of loudness in one violin with
"Darnley" (1712) 16.0 23.9 23.8 24.2 26.3 9.0 that in another. The errors of our methods are
Strad E (Joachim,
Kneisel, 1715) 15.2 25.1 24.6 26.2 21.7 6.5 not likely to affect this comparisonseriously.
Strad G "Tom Table I discloses considerable variations in
Taylor"(1732) 11.6 23.0 26.9 25.9 23.7 10.6
StradJ ",M,
arquis de 11.9 strength in different ranges,especiallyat the
Riviere (1718) 22.3 24.7 25.3 24.5 16.4 extremes.Some of these variations may be due
Strad S "Spanish"
(1723) 14.3 23.3 30.1 28.6 25.5 14.0 to differencesin the spatial emissionpattern for
Strad W "Titian"
(1715) 14.5 23.2 25.0 24.6 22.7 12.7
.differentwave-lengths.The valuesare given to
Averages 14.1 23.1 25.7 25.9 23.6 10.9 tenths of one decibel, but the error (due to all
causes)may amount to as much as 2 db; the
relative accuracyof the values for one violin is
Different ranges are used below in considering probablyhigher.The table showsthat all Strads
violas and cellos. do not sound alike, as experts already know,
Many violins, including two or three of the Probably this was true evenwhen they were new.
seven Strads for which we have curves by the but somedifferencesmay have arisen from the
new method, showa weaknessin the range 1300 changessincemadein theseviolins.Someplayers
to 1800 or 2000 c.p.s., amounting to a drop of prefer a "full" or "round" tone, with more
4 to 8 db. This.appearsnot to have beennoticed strengthin low frequencies than in high;others
by violin experts, and to have no important prefer a more biting tone, with the strengths
effect on the reputation of the violins concerned. reversed. Table I cannot settle which Strad is
Our range ending at 1568 almost bisects this best, or which should have the highest price.
"hole," and this accident hides it from view in These data apply to steadytonesonly, and the
our tables. In such a violin there is a weakness quality of a tone (meaningby this the distribu-
in the fundamental of the tone in the first few tion of strength among its partials) is not the
semitonesof the secondoctave on the E string; only item of importanceabout a violin. In fact
but the upper overtonesmay be so strongas to it is so variable from one tone to another that it
make up the deficiency to the ear. On this may be rather unimportantcomparedwith other
accountit has not been thought worth while to effects,suchas thoseoccurringnearthe beginning
chooseany differentrangesin this region. and end of each tone. All these Strads must be
Table I showsthe average intensity emitted consideredas excellentsincethey commandhigh
in the chosenfrequency ranges by seven well- prices,and sincea large number (if not all) of
known Stradivarius violins. The values were the experts have agreed upon them. We have
obtainedwith somewhatdifferentamplifications. assumedthat the averagedistributionof strength
The distance of the violin from the microphone in our chosen ranges furnished by these seven
was about 40 cm but it could not be kept quite Strads representsthe closest approach to a
constant.For these,or perhapsother reasons,it standard of excellence that we cad attain.
is not possibleto compare the total intensity It shouldbe noted that the completehistory
emitted by one violin with that from another. of these(and of mostother) Stradsis not known.
Instead, all records were reduced to the same Some are in Hill's great book on Stradivarius,
total area, first by integrating the audiograph but even these historiesusually begin a century
curves with a planimeter, then reducing the or so after the date of birth of the violin. Data
frequency scale to a uniform logarithmic one, about alterations made in famous violins are not
and then adjusting the base level from which usually made public. None of these instruments
the areas were measured until the total area for can be in their original condition.The standard
172 F. A. SAUNDERS

of pitch has risen,requiringincreasedtensionin even the author, who was permitted to be

the strings of old violins; hence longer bass-bars present, was soon able to detect them. But so
have been put into them to increase their delicate were these differences that a change in
strength. These changes serve to satisfy the the music played, or any conversationbetween
demand for louder violins to fit our concert halls.the tests of A and B, tended to destroy the
If the changeshave beenwell made they may not accuracyof judgment. When the relative merits
lower the value of the violins. The longer bass- of violins A and B were settled on all four strings,
bars may be valuable; and the changesin the violins A and C were treated in the same way,
wood itself that have come with age or long then A and D, B and C, etc., taking all possible
playing have probably improved the quality. It combinations.Thus the relative positionsof the
is not surprisingthat Strads are found to differ four violins were settled. If A represents the
among themselves,even if they were made on winner, B the one in second place, and so on,
the model of the "golden age." it is interestingto note that when A and D were
comparedit was agreed that D had the best E
THE HEIFETZ TEST
stringof the four, and nearly as gooda G string
In April, 19•,0 an interesting test was con- as A. This showshow closewere the qualitiesof
ducted by Jascha Heifetz, with the help of these four instruments.
SaschaJacobson.American violin makers had This test can be criticized on the ground that
been invited to sendin samplesof their work for theviolinswereplal•edastheyhadbeenadjusted
comparison, and nearly 100 violins had been by their makers. Their order might have been
received. The first tests eliminated a large different if they had all been adjusted by the
number of these. The final tests on the 'four best same person,or if they had been supplied with
violins should be described. Two violins to be different strings, or bridges, or bass-bars.It was
compared,A and B, were played by each of not practicable to do any of these things. In
the artists, H and J. With his back turned, J spite of this possibledefect, the test had a
listenedat one end of a large room, while H at particular value in connectionwith our research.
the other end played a certain passagefrom a Mr. Heifetz had the idea from the beginning
concerto on one string of violin A and then that the winners should be tested in the Harvard
changedto violin B, repeating the same passage Laboratories, and later on this was done with
on the same string of B. The one acting as three of the four at the top. But we profired in
judge did not know which of the two violinswas other ways than by securingthe opportunity of
being played, but decided which one sounded measuring someinteresting violins.Aftera few
better on this string. The same process was judgments had been made durihg the test, it
repeated on the other strings of violins A and B seemnedobvious, if one was at all accustomed to
in turn. Then J becameplayer and H judge with making tone analysesby ear, that the artists
the sameviolins. In casethe judgments did not wanted two qualities in a violin; first, great
agree (which was unusual) more testswere made power; second,an even distribution of strength
until agreement was reached. The differences among all rangesof frequency, the lowest octave
between two violins on the same string were being of special importance. It was said after
sometimes very small, but usually quite definite; the test that the winner gave the impression
that the limit of its power was never reached.
TABLE II. Modern violins in the Heifetz test. Average One should add that even a poor violin can be
intensities in different frequency ranges.
made to emit a great deal of sound,such as it is.
Ranges The power referred to here is limited by the
Violin• I II III IV V VI
condition that the tone must remain good; with
Yurkevitch 12.7 24.4 27.1 25.3 22.0 8.6 mostviolinsa very.loud tone becomesunpleasant.
Phillips 13.1 23.5 25.2 26.8 22.9 10.0
SangsterNo. 35 12.4 23.5 23.8 25.6 28.9 11.0 Since the power output could not readily be
measured,and probably dependssomewhat on
Average 12.9 23.8 25.4 25.9 24.6 9.9
Strad average 14.1 23.7 25.7 25.9 23.6 10.9 the player, we put the winning violins through
the same tests as already outlined for the seven
MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIX; F,\MII. Y 173

3O

29O

50

ZO

I
•TRJ•OB •

a 0

20O0 400O • 200 5O0

I I , l
20O0 4000
FREQUENCY FREQUENCY

Fla. 1. Responsecurvesby new methodof three Strads, and three new violins which led in the Heifetz test.
174
Strads. The violins tested were (A) by M. V.
Yurkevitch (1938) of New York City; (B) by
B. F. Phillips of Pittsburgh, Pennsylvania;(D)
by E. H. Sangster, now of Worcester, Massa-
chusetts. Table II shows the results. The distri-
bution of strength in different rangesis unusually
uniform (the rangeswere the sameas in Table I),
and its average is the sameas the Strad average,
within the allowable experimental error.
The responsecurves of these three violins
(particularly the Yurkevitch) showan evenness
of response whichis unusual.They still danceup
and down in a way that no proper loudspeaker
curve does,but the generallevel is steady,more
so than is usual even amongfamousold violins.
Figure 1 showsa seriesof responsecurves,all
taken by our new method. Three Strads are
shown, and the three violins of the Heifetz test.
There is a "hole" in the middle of the curve for
Strad D (and others), amounting to a drop of
8 or 10 db in the region from 1300 to 2000 c.p.s.
The only effect of this hole seemsto be to pro-
ducedifferences
in tone qualityof whichthe
player is not aware, or which he may even enjoy
for their variety.
We can find no evidence in such response
curvesfor the existenceof any special"formant,"
or region of strong response,which is supposed
to give the characteristicquality to the soundof
a violin. In fact, it seemslikely that we recognize
a violin tone as different from that of an oboe,
for instance,by the way the tonestartsand stops.
There is no marked difference
analyses of the two instruments. It would not
be difficult to test this by experiment, permitting
the listener to hear the tones only after they
had started, and not allowing him to hear their
natural endings.
TABLE III. Average values of intensity in the same
frequencyranges,by old and new methods.
I II III
9 violins, averages, old method 37.0 42.4 38.0
Same by new method 13.7 23.3 25.3
Difference due to method 23.3 19.1 12.7
7 Strads, new method only 14,.1 23.7 25.7
5 Strads, old method only, but
corrected to new 14.6 24,.4, 24.0
10 Strads, both methods 14.2 23.8 24.7
F. A. SAUNDERS
RESPONSE CURVES BY BOTH METHODS
The intensity distribution for a number of
other newviollns, measuredby the new method,
is given in Table IV, along with the results
obtainedby the old method,whichwe must
next consider.
In our article (D) on the new method it was
stated that the results differed from those by
the old method at both ends of the frequency
range. In the old methodordinary human bowing
is used,and ordinary strings.In the new method
only one "string" (a phosphor-bronzewire) is
used, and only from the G-string position. Its
tensionis equal to that of the ordinary G string,
but its linear density is less. It could not be
expectedto shakethe whole violin exactly as the
G string does; the loudnessis appreciably less,
though this may be mainly due to the absenceof
overtones.The observedreduction in the highest
frequenciesby the new method may be due to
the fact that the excitation is from the G-string
position.Some writers state that if the E and G
strings are interchanged, the high notes are
weakened. These differencesraise a difficulty in
comparingthe resultsof the two methods.
A method of reducing the values obtained by
one method to those by the other was found by
studying the violins which were measured by
both. These were five Strads and four new violins•
The average value in each of the six frequency
rangeswas obtained for each of thesenine violins
by both methods. The averages for the nine
in the tone were then obtained (Table III). If the new-
method values are subtracted from the corre-
spondingold-methodones,thesedifferencesmay
be used to convert any old-method results to
new, or vice versa. This assumes that the violins
were in the same condition in both tests, and
that no other variations occurred, such, for
instance, as might come from diffraction effects.
With nine violins to work with it seems reason-
able to hope that the averagesare .not seriously
IV V VI
affected by any such variations.
38.6 37.5 27.3 In all, 73 violins were studied. Responsecurves
26.3 23.8 11.6
were taken for 60 of these, 48 by the old method,
12.3 13.7 15.7
22 by the new, 10 by both. We have the response
25.9 23.6 10.9
curves of 38 violins by the old method only.
25.4 23.1 11.7
25.7 23.9 10.7 This massof observationalmaterial is of value by
itself, but it is better to be able to comparethe
MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIN FAMILY 175

TnBt•E IV. Intensitiesin different frequencyranges.

Ranges
Violins Method I II III IV V VI Violins Method I II III IV V VI

Strads, by both methods New violins

Strad K (1684) old 15.9 22.9 22.3 24.5 24.7 11.4 J. A. Gould (1882) old 13.5 22.9 27.3 25.7 25.5 9.7

Strad A (1691) old 15.5 23.7 24.1 21.9 25.7 15.7 W. S. Goss (1912) old 10.1 24.1 27.7 24.3 21.5 15.1

Strad B (1694) old 17.3 19.9 28.1 25.9 25.7 7.3 Sangster's No. 30 old 10.3 26.1 23.5 27.7 23.7 11.3
new 15.2 25.0 24.6 26.5 20.8 7.1 new 13.4 22.1 24.6 26.2 24.6 13.3

Strad M (1698) old 11.7 24.5 25.3 26.7 24.1 10.5 Sangster's No. 33 old 12.9 26.1 26.1 25.3 21.3 9.0

Strad J (17097) old 15.1 24.9 25.1 26.1 22.9 6.9 Sangster's No. 34 new 13.0 23.2 21.0 27.4 27.1 13.9
new 11.9 22.3 24.7 25.3 24.5 16.4
Sangster's No. 35 new 12.4 23.5 23.8 25.6 28.9 I 1.0
Strad D (1712) new 16.0 23.9 23.8 24.2 26.3 9.0
Sangster's No. 37 new 11.7 24.9 23.3 25.8 26.2 11.9
Strad E (1715) old 15.3 24.1 25.7 24.3 24.7 9.1
new 15.2 25.l 24.6 26.2 21.7 6.5 Sangster's No. 38 new 12.6 22.5 24.4 26.6 23.2 13.8

Strad W (1715) old 12.5 23,3 23.9 27.3 24.5 12.9 Moennig,
copyof
StradB old 12.1 24.0 24.3 24.7 28.3 15.3
new 14.5 23.2 25,0 24.6 22.7 12.5 new 11.3 21.3 27.8 26.2 23.2 13.4

Strad S (1723) old 13.1 25.9 24.6 26.3 25.3 12.3 Moennig, copy of $trad J old 11.1 24.5 26.3 23.l 21.9 16.9
new 14.3 23.3 30.1 28.6 25.5 14.0
Moennig, copyofaStrad old 14.5 21.3 22.6 23.6 28.8 17.8
Strad H (1731) old 14.9 25.1 23.7 26.5 19.7 10.0
Moennig, copy of a old 15.2 20.8 26.4 22.7 22.3 17.9
Strad G (1732) new 11.6 23.0 26.9 25.9 23.7 10.6 Guadagnini
Strad N (1737) old 10.1 25.7 24.7 27.3 21.5 11.3 ¾urkevitch new 12.7 24.4 27.1 25.3 22.0 8.6

Strad average 14.1 23.8 25.1 25.8 23.8 10.8 B. F. Phillips new 13.1 23.5 25.2 26.8 22.9 10.0

Guarnerius d. Gesu violins; old method Koch old 15.3 21.9 25.9 26.7 18.9 14.1
Guarnerius R (1743) old 10.3 25.7 24.5 26.3 25.9 10.3 new 14.0 23.8 23.2 26.2 24.9 11.5
Guarnerius H (1742) old 13.6 24.9 27.6 25.5 21.9 5.9
Guarnerius B (1728) old 12.9 23.7 25.1 27.9 24.1 10.1 Stanley old 12.5 20.9 26.5 28.1 20.7 15.5
Guarnerius F (1738) old 11.9 22.3 27.9 25.3 22.1 13.5 new 13.5 23.6 23.1 26.9 25.9 10.7

Guarnerius average 12.2 24.1 26,3 26.2 23.5 10,0 Moglie (1930) new 11.5 23.3 29.3 26.2 22.6 6.9

Old Italian violins; makers probably as indicated Averages of good new violins 12.7 23.3 25.1 25.8 24.0 12.7
Maggini old 15.7 23.7 26.3 23.3 22.5 12.3 Two bad violins
P. Guarnerius (Cremona) old 14.6 19.4 22.8 28.7 27.6 15.7
Stainer old 14.5 21.3 24.1 27.5 24.5 14.5 Violin X ($5.00) new 13.9 18.5 29.5 23.4 22.6 17.6
A. Guarnerius old 13.5 24.5 25.7 25.1 20.5 11.O Violin ¾ new 2.6 20.5 '20.9 42.2 17.0 16.4
$tradivarius old 12.7 22.7 24.9 25.7 23.5 15.1
Guarnerius, J old 18.1 23.3 24.7 24.1 22.9 11.3
Gagllano old 18.1 27.1 28.3 22.5 14.7 5.7
Guadagnini old 15.1 23.5 24.1 25.l 21.1 14.1
Pressenda old 12.5 22.3 24.3 27.1 22.l 16.7

These old violins, average 15.0 23.1 25.0 25.5 22.2 13.0

resultsof the two methods, as we canby the "long"model,andone(K) is of thehighmodel

aboveprocedure. TableIII includestheaverage (1684).The results
donotagreewitha common
valuesfor 7 Stradsby the newmethod,for 10 beliefthat the differences in shapeaffectthe
Stradsby bothold and newmethods (reduceddistributionof strengththroughout the range.
to new),andfor 5 StradswhichweremeasuredThe two long-model violinsdo not agreewith
by the old methodonly. Theseresultsshowonly each other. There is a rather uncertaintrend in
smalldifferences,
indicatingthat the process
of thedirection
of diminishing lowfrequencyemis-
comparisonis successful. sionwithincreasein ageof themastercraftsman,
We nextconsider the resultsof measurementsbut thesedifferencesare usuallylessthan the
on the mostinterestingof the violinsmeasured possible
errorof the measurements.
by the new method, and also those by the old In Table IV thereare four famousviolinsmade
methodtranslated
to thenew-method
scale.
The by J. Guarnerius
(del Gesu)whoseaverageis
12 Stradsincludedin Table IV are all pedigreed fairly closeto that of the Strads.Thereseemsto
thoroughbreds of unquestionedstanding,though be slightlylessemission in the lowestfrequency
one naturally does not know just how their range,thoughthis differenceis too small to be
present condition differs from that of their sureof; there are alsosomeindividual variations.
infancy.Two of these(A and B) are of the The second
in thelistis the "David,"playedby
176 F. A. SAUNDERS

TABLE
V. Intensity
distribution
in violins not pleased,and since Mr. Heifetz has a great
chosenby one artist.
many listeners,we might assumethat his type
Ranges is generally preferred; but it is to be remembered
Violins I II III IV V VI that a great violinist can play a rather ordinary
Strad H 14.9 25.1 23.7 26.5 19.7 10.0 violin in sucha way that the averagelisteneris
Guarnerius H 13.6 24.9 27.6 25.5 21.9 5.9
Yurkevitch 12.7 24.4 27.1 25.3 22.0 8.6 quite content, and thinks that he is hearing a
famous instrument.
The data in Table IV on Sangster'sviolins
Heifetz; the third the "Verviers," played by enable us to test the variability of the product
Z. Balokovic;the fourth the "Fontaine," played of one modern maker, using the same model
by .Miss Thelma Given. throughout.Actually,as the resultssho{v,his
A group of Italian violins follows, arranged violins vary lessthan the Strads do, though the
approximately in the order of their ages,which differencesare uncertain on account of experi-
run from 370 to 100 years. While these instru- mental error. This table also includes an inter-
ments are all valuable and undoubtedly old esting violin made by Moennig in Philadelphia,
Italian, someof them lack the highestcertificates, a handsomecopy of Strad B. Comparing the
so that onecan say only that the makersare values for this copy with those of the original,
probably as indicated. This group again shows we see that the Strad is stronger in the low fre-
marked individual variations, with an average quenciesand weaker in the high than its copy,
not far from that of the Strads, though the though the differencesare not greater than those
emission at both ends of the scale is a little that occur among Strads. Smaller differencesare
higher. Note how the Gagllano emphasizesthe found betweenStrad J and its copy by Moennig.
The Koch violin is of some interest since two
lower tones,while the highly arched P. Guarnerius
lecture audiences thought it was more surely
puts more strength into the higher ones.
Italian (and old) than a Strad with which it was
The group of new violins in Table IV shows
being compared (behind a screen). Its distribu-
the same variations as the other groups. If one
tion of strengthwith frequencyfollowsthe Strad
were certain that any particular Strad was ideal
pattern rather closely.The Stanley (Boston) is
in its distribution, and should be copied by all
not very different from the Koch.
makers, then one could find in this table at least
There are some unpleasant irregularities in
one modern violin which approximately dupli-
Table IV. For instance, the values taken by old
cates it. This matching can be done for high
and new methods for Strad B differ unreasonably.
model violins, or for low. One cannot say that In this case the violin had been treated to a new
one model is good and another bad. Artists who bass-bar in the interval between tests, and the
are much in the public eye use different models, changewas probably real. The very large differ-.
and are presumably pleased with what they encein the upper range for Strad J may again be
have. It happens that we are in a position to doe to a changein the instrument. Similarly, in
gauge the preferencesof Mr. Heifetz. We have the Koch and Stanley violins (rangeV) there are
measureda Strad and a Guarnerius, both chosen differenceswhich are larger than the experi-
for his own use, and we have the winner of the mental error. In all these cases there was a gap
new-violin test, which also represents.what he of a year or two between the two tests, and in
prefers.Table V presentsthe distribution in these this time all the instruments had been played a
three violins; the consistencyof judgment shown gooddeal, and their bridgesand stringshad been
in their selection is indicated by the values changed. Unfortunately, we took no tests by
presented here. The Heifetz type is evidently both methods on the same instrument at the
one in which the two upperranges(whichto- sametime, sothat we cannotsay positively,when
gether cover the octave from 3136 to 6300 c.p.s.) differencesare found, whether these are due to
are relativelyweakcompared with the middle the peculiaritiesof the methodsor to changesin
o.ne.s•
a.nd the lowestrangeis fairly .st.rong.the violins. Unpublished.testswhich we have
Sinc• fro artist 'can flou?ishif his listenersare made
onbridges
andons{rings
haveshown
[hat
MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIN FAMILY 177

a violin is sensitiveto changesin these items; by the two methodswere found from data on
even the changefrom a thin string to a thicker viola 6, and were used (lacking a more accurate
one affectsthe distribution of strength over the way) to convert from one method to the other.
frequency ranges. Hence we believe that the The values given are in terms of the old method,
above irregularitiesare probably real. because more of the violas were measured this
At the bottom of Table IV two violins are way. The total areas of the responsecurveswere
entered. Violin X is one of the cheapesttype, equalized, as with violins, to make the results
madein Czechoslovakia;it could be bought.with comparable.The first (lowest) frequencyrange is
case,bow, and a set of directionsfor playing it, taken as an octave and includes the region
for lessthan $15.00. It had, however, been some- affected by the air resonance.The secondrange
what "doctored" by a new bridge arranged to (another octave) is relatively strong in most
filter out someof the excessivehigh frequencies, violas and includes the first body resonance,
and by new strings to give more strength in the which is the sourceof the "wolf" tone.This body
low ranges. It sounded well when Mr. Heifetz resonanceis stronger in the larger instruments,
played it, but not otherwise.The figuresgiven and has the effect of strengthening the tones in
for it make it appear better than it was sinceit its region. It must on this account be regarded
had strongemissionabove 6300 c.p.s., which our as actually beneficial, and it is of course un-
tablesdo not include.Violin Y was a flat-topped avoidable. The actual wolf is found only at the
model made out of common lumber by an un- exact point of maximum resonance but the
skilledworkmanfor hisownpleasure.It possessed beneficialeffectsof this resonancespreadover a
to his ear a beautiful tone, though its low considerablerange on either side. The wolf can
frequencyemissionwas almost absent, and its be reducedby any device which tends to check
high frequencystrengthextendedfar beyondthe the productionof the particular pattern of sub-
proper range. divisionof the top plate of the instrumentwhich
RESPONSE CURVES OF VIOLAS AND CELLOS
'is responsible
for the excessive
resonance.
The
other (higher) body-resonances
are well scattered
Nine violas were measured, six old and three and numerous. The values for the three short
new. No. 1 was the "McDonald" Stradivarius frequency ranges at the top of the scale vary
(1701) in the Warburg collection; No. 2 the somewhat erratically in instruments of equal
Amati (1677) used by M. Aronoff of the Curtis merit. If a strong resonancehappensto fall in
Quartet; No. 3 a Gaspar da Salo (about 1570) the middle of one of these rangesits effect is
owned by Miss Eunice Wheeler, an especially fully evident in the measurementfor this range;
largeand responsiveinstrument; No. 4 a Storioni if, however, it falls at a boundary, its effect is
belonging to Marcel Dick; No. 5 a probable spread between two ranges,and is then partly
Gaspar da Salo; No. 6 possiblya Klotz (made hidden in our numerical values.
about 1770); No. 7 a new viola made by Haenel Figure 2 shows the response curves (old
of Toronto; No. 8 a copy of No. 2 made by method) for four of the most interestingviolas in
Moennig; No. 9 by Dieudonn• (1939). The this group.
rangesof frequencyusedwith violas were seven,
four octaves and three shorter intervals; I from Ta.nLEVI. Intensitiesin differentfrequency
rangesfor violas.
131 to 262 c.p.s.; II from 262 to 523; III from
523 to 1046; IV from 1046 to 2093; V from 2093 Rallge8
Violas I II III IV V VI VII
to 3136; VI from 3136 to 4186; VII from 4186
to 6272. Viola I 26.3 37.7 35.3 29.9 29.1 24.3 18.1
Viola 2 29.0 38.0 33.2 32.8 29.0 19.6 15.2
Table VI was obtainedby a proceduresimilar Viola 3 27.3 35.9 33.1 31.7 32.1 23.7 17.5
Viola 4 28.2 38.2 28.8 35.0 35.3 23.8 18.8
to that usedfor violins. Viola 6 was the only one Viola 5 27.3 38.9 31.1 34.6 36.9 22.0 16.0
of the nine which was measured by bot]q old and Viola 6 28.2 39.6 30.6 35.6 30.8 22.0 18.6
new methods. Violas 4, 5, and 9 were measured Viola 7 24.0 35.8 29.2 33.8 37.0 26.2 19.4
Viola 8 27.8 36.8 33.2 28.2 30.4 28.0 19.0
by the new method only; violas 1, 2, 3, 7, and 8 Viola 9 27.8 37.1 30.2 33.6 33.7 26.0 23.1
by the old method only. The differencesin values
178 F. A. SAUNDERS

and 4. They resembleoneanotherin the lower

frequencies.
The dataat highfrequencies
suffer
from the fact that the observations were taken
onlyat semitone
intervals,
therebymissing
much
interestingdetail. The two oldestcellosshow
sharppeaksat the two lowerresonances, while
the new one showswider peaks. Sharpnessin
resonance may be a signof very freevibration,
andmaywellincrease with age;but a widepeak
spreads
its helpfulinfluence
overseveralselni-
tones,which is distinctly better.
CONCLUSIONS FROM RESPONSE CURVES OF
VIOLINS, VIOLAS, AND CELLOS

Two positiveconclusions
canbe drawnfronl
thispartof thework.Thesearethat thereis no
correlationbetween the price of an instrument
FIG.2.Responsecurves
offourviolas
(Nos.1,2,3,and8) and its distribution of strengthwith frequency;
takenby theoldmethod. Thescale belowislikea piano and that, whateveris the bestdistribution,it is
keyboard, 12divisions
to theoctave;frequenciesof C's not exclusively
the propertyof old instrmnents.
given.Thevertical scale
shows thelogarithmof thein-
tensity,in decibels.
The numbered scalerefersto the There is about the samevariation in distribution
Strad;theothercurvesareshiftedeach20 db upward in
succession,
to saveconfusion.
Thus,thelinemarked 40 is amongthe old as alnonggoodnewones,and
the zero for the Gaspar. whatever
typetheplayerprefers,
hemayfindit
either in old or new instruments.This beingso,
Fourcelloswerestudiedbs;the old method. we must examine other properties of these
The resultsare shownin Table VII. instrumentsto seekthe reasonfor the very high
Cello 1 is a Pressenda,
playedby Maurice regard
in whicholdItalianinstruments
areheld.
Stad;No.2 isa Montagnana
playedbyOrlando While it is true that the responsecurvesof
Coleof the CurtisQuartet;No. 3 is a Strad,the violas and cellos strongly resemblethose of
"Vaslin,"playedby GeraldWarburg; No.4 isa violins,thereis an interestingdifferencein the
copy
byNloennig
ofNo.2.TableVII shows
that position
of themainair resonance
point,aswell
thesefourdifferremarkablylittle amongthem- as in the strengthof the main body resonance.
selvesin their intensitydistribution.They all Counting by semitones fromthelowesttoneof
showa largeemission in themainbody-resonant the instrument,the air resonance
usuallyoccurs
region VI. Theonenew in violinsnearNo. 6, in violasnearNo. 9, and in
(II), andagainin range
instrumentin the grouphas high emission
in cellos near 7 or 8. The main body resonance
thelowestrange,andrelatively
lowin thehigh comessomewherenear 17 in all these instru-
ranges;
in otherwords,
it shows
the'emission
ments,with variationsof two semitones
either
characteristics
whichwe havecometo associateway. Thereis a regionof weakness nearthe
with the best old instruments.
As before,in bottomof the scaleand anotherbetweenthe two
treating
themeasured
results theareas main resonances.
oncellos, The positionof the air reso-
of thecurveswereequalized.
Thusthe relative
TABLEVII. Cellos.Intensitiesin differentfrequency
strength
indifferent
ranges
isgiven,
butnotthe ranges,
eachan octavelowerthanforviolas.
actual intensities.
It is unfortunatethat no celloscouldbe Rangea
Celloa I II III IV V VI VII
measuredby the new method.Our apparatus
wasnotlargeenough, andit wasnotthoughtCello I
Cello2
35.2
35.0
46.5
44.4
35.9
36.8
32.1
32.8
29.9
27.4
36.3
37.4
27.3
29.2
likelythatresults
of anynew•-pewouldbe Cello3 36.8 45.2 35.4 31.2 28.8 35.6 30.2
obtained the Cello4
froma largerone.Figure3 shows 39.1 45.9 36.5 31.3 27.7 31.9 26.9
response
curves
(oldmethod)
of Cellos
2, 3,
MECHANICAL ACTION.OF INSTRUMENTS OF THE VIOLIN FAMILY 179

nance can be lowered permanently only by

making the volume of the body greater, or
by making the total area of the openingsof
the f-holes smaller;but it may be loweredex-
perimentally by filling the body with carbon
dioxidegas througha small rubber tube, or by
covering a considerablepart of the f-hole
openingsxvith "Scotch"tape (which peelsoff
againwithout injury to the varnish).The great
mastersof the craft placedit about lnidway be-
tween the bottom and the first body resonance,
whereit woulddo the mostgoodin strengthening
the otherwiseweak lower toues. Why doesit not
occurat the samepositionin all instrumentsof Fro. 3. Responsecurvesof three cellos('qos. 2, 3, and 4).
the violin family? The scalesare as iu Fig. 2. The note C• has a frequency
131 c.p.s.
The ansxverto this questionis not certain, but
two suggestions canbe made.Onewhichapplies
Curt two curves run approximately parallel. The
particularlyto violasis that, as Professor
Sachshas pointedout, violaswere made longer difference between the loud and soft values at
and deeperbefore1700than they are now.This each tone is nearly constant, though there are
mightwell havebroughtthe air resonancedown small erratic errors due probably to the player.
to where it is in violins, but such instruments These differencesgive the range of loudnessof
were so deepas to be clumsyto play, and they the instrument, which for a violin varies little
havebeensuperseded by smallerones.Many old from 30 (lb. This number is of no specialinterest
ones were then altered to be less deep, thus becauseit is nearly the samefor old and new, and
raisingtheir air resonantpoints.This has made for goodand bad violins. lf, however,we average
the tones on the lowest string in many small the two values at each tone, we get a set of
violas rather weak, and sometimes mean in numbers yielding an average loudness curve,
quality;but thereare manylongviolasof recent which is significant. It differs from the response
make which seem satisfactory, even with rather curves already considered.The latter show the
high air resonances.
This anomalycan be ex- responseof the violin for one frequency at a
plainedby a suggestion
madein thenextsection. time, dealing with only one partial tone (e.g., the
fundamental). Loudness curves show the re-
CURVES OF INTENSITY (OR LOUDNESS)
sponseproducedby the bow for one tone at a
We must now consider other types o[ tests time, but each tone contains all its partials. The
loudness curve does not discriminate between a
which throw light on the peculiaritiesjust men-
tioned.The simplestof theseare curvesgiving tone which has a strong fundamental and weak
the total intensity (or approximatelythe loud- partials, and one having a weak fundamental and
ness) of tones over the whole range o[ the strongpartials. Every peak on the responsecurve
instrument.Thesemay be obtainedby meanso[ appearson the lc•u(lnesscurve, but the latter may
a sound-meter. The player keeps a constant show additional peaks not present in the former.
distance from the sound-meter in a deadened The most important case is that in which the
room, and first plays as softly as he can on one fundamental is not strong, but the secondpartial
tone while the loudnessis read on the meter. This is so loud that it more than makes up for the
is repeated [or eachtone(by semitones) overthe weaknessof the fundamental. This happensone
rangeof the instrument.Then the wholeexperi- octave below each of the most important body
ment is repeated,but this time to measurethe resonances.Thus a violin may have its air reso-
maximum loudnessthat can be produced.Each nanceat the open D (294 c.p.s.) and a body peak
set of observationsyields a curve in which the near B flat on the A string (466 c.p.s.). The
mainresonantfrequencies showaspeaks,and the result is that the loudnesscurve showsa peak at
18(I SAUNDERS

Table IV, fivehavea goodarrangement

strength-
eningthelowesttones,threedefinitelylackthis
advantage,and four are intermediatefor one
reason or another. New bass-barswhich have
beenput into all theseviolinshave probably
shifted the body resonances,so that we cannot
saythat thisrecorddimsthehalooverthehead
of Stradivarius.
In •iolas,if the onestestedare representative,
there is somevariability. Many have weak C
strings,
butmorehavetheirlowertonesstrength-
ened by the help of body resonances
an octave
higher.
Sincetheair resonance
lieshigherthanit
doesin violins(relatively),the chancethat the
airandbodyresonances
willbeanoctaveapartis
smaller. Moreover, the body resonances are
FiG.4. Curvesof averageintensitiesfor a viola(No. 6), nsuallystrong.Here may be the reasonfor
two old violinsand a new one.The horizontalscaleis placing theairresonance neartheninthsemitone,
thesameas thatin Fig.2 for theviola,but for theviolins rather than at the sixth,as in violins.This applies
the lowesttoneis G. The verticalscalesare as before.
to cellosalso. If there is no wolf tone there is no
233,aswellas thoseat 294and466.Sincethe strong
bodyresonance,
andhence
theC stringis
are not very narrow,snchan arrange- likelyto beweak;in suchcases
resonances a relativelylow
air resonancewonld be of some advantage.
mentyieldsa fairlysatisfactory
loudness
overthe
In all the four cellos tested the main body
lowerrangeof the violin.If, however,
the air
resonanceis at C sharp (277) and the body resonance is about an octaveabove the air reso-
resonance
is aboutan octavehigher,the loudness nance.The curvesin Fig. 3 showthis. It is likely
near277is needlesslyenhanced by the strength that the G stringsof thesecelloswouldbe ap-
of the secondpartial,and thereis nothingto preciablyimprovedif the air resonance were
strengthen theweaktonesbelow277,or in the movedupward,say by enlargingthe f-holes
intervalbetween277and 554.The lowertonesof (scandalous assucha suggestion maybe).The
sncha violinare weak,aswell as thosecentering mainbodyresonance is strongenoughto give
around410,midwaybetweenthe twopeaks.We ample loudness anoctave below,andtherewould
know of one famous violin in which strong be a moreevenresponse over the low tones.
resonancesoccur near all the C's in-four octaves, EASE AND QUICKNESSOF RESPONSE
to the detriment of the strengthof many other
tones in its range. We mustnowconsider
otherqualitiesof violins
curves to which playersfreqnentlyrefer. Theseare
Figure4 showsthreeaverageloudness
for violins,and one for a viola. The top curve vaguely included
in theirideaofthetonequality
showsa bodyresonance at A (440)whichin- of the instrument.Theseattributesare referred
creases the loudness
at 220.This makesa much to as the easeof playing,or the quicknessof
betterarrangement, especially for the G string, response, or "facilearticulation," or by many
than that shownin the curve belowfor an old otherphrases. Some of these have todowiththe
Italian violin,probablyAmati, wherethereare growth of thesotrod at the beginning ofa tone,or
weak hollows becausethe two main resonances itsdecay at the end. Players assume that if the
areaboutan octaveap.art.The nextviolinshows toneiseasily produced it muststart quickly and
thebodyresonance at B flat (466)helping to fill continueon for a long time after the bow ceases
out the tonein the regionof 233.Viola 6 at the to act on thestring,but physicists knowthat if
bottomhasa strongbodypeakbetweenF and F a vibrationis slowto decayit mustalsobe slow
sharp(about
360)which
exerts effect in starting.
anexcellent Thisistrueif theconditions
ofvibra-
an octavebelow.Of the twelveStradslistedin tionarethesamein bothcases,
whichisnotquite
MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIN FAMILY 181

the case with bowed violins, since the bow is in than those by J. Guarnerius. The $5.00 fiddle
contact with the string during the growth, but comesout very near one Guarnerius!This shows
not duringthe decay.Theseeffectsdeservesome rather conclusivelythat the easeof playing has
study. Violinsalsodiffer in efficiency,that is in little to do with the merits of a violin. Evidently
the amountof powerwhich must be exertedby theseforces,varying from eight to twenty grams
the player to make them sing.This may be con- or so, are so small that no violinist noticesthem,
nectedwith the quicknessof growthof the tones, or cares about them. These data refer to the
but it is not the same thing. Two lines of in- productionof soft sounds(piano on the musical
vestigation are suggestedby these thoughts: gradation);in fortissimopassages one can easily
(1) to measurethe powerneededto makevarious exerta forceof 150gramswith the part of the bow
violinssing;(2) to measurethe rate of growthof nearestto the hand, but even this (one-third of a
their tones, or the rate of decay. The decay is pound) is too small to fatigue the player, and it
perhapsmorecharacteristicof the violin than the is small comparedwith the force which he must
growth,and lessof the player; the decayis also exert to raise his own arm. The differences in ease
much easier to measure. of playing may not be detectableby the player
RAMAN CURVES
when they are small, but it seemslikely that the
difference could be felt when it amounts to 20
The measurement of poweris doneby means gramsor more, even thoughit doesnot bother
of a type of apparatusdevisedby Raman (see him. Certainly if the player wants a violin that
articleA). We call the resultingcurvethe Raman is easy to play he can find one without trouble
among either old or new instruments.It is also
curve of the violin. This yields the samedistribu-
tion of resonances possiblethat the violin that is hardest to play
that is given by the loudness
curve, and it is more difficult to measureaccu- may be the loudest,and may on this accountbe
rately;but it alsoyieldsnumericalvaluesfor the preferredby the player.
power nee(ledto make the instrumentspeak,
DAMPING
which are not obtainable otherwise. We have
restricted these measurements to the tones in the We now turn to the study of the rate of
first octave on eachof the two loweststrings,and damping of the vibration of a violin. Many
will save spacehere by presentingonly the expertsconsiderthat the soundof the bestviolins
averagevaluesfor the thirteentonesmeasured on rings on for a relatively long time after the
eachstring.Theseaveragenumbersare foundto excitation has ceased to act. One can give the
dependgreatlyonthetypeof stringused,that is, back of a violin a sharp rap with a knuckle and
on its tensionand linear density. Sincean artist hear a brief sound thereafter, and the duration of
cannot be asked to change the stringson his this soundmay be an indicationof the quality of
violin, we have had to contentourselveswith the instrument. To make this test properly one
what we foundthere, trustingthat his selection mustpreventthe stringsand the tail-piecefrom
was the best possible.The numbersgiven in
Table VIII are in gramsof forcewith whichour TABLEVIII.againstAverageforcein gramsneededto pressbow
stringto makeviolin speak.
rotatingwheelbowpressed on thestringwhileits
circumference wasgoingby at a speedof 38 cm Old violins New violins
G D G D
per sec.,a speedkept approximatelyconstant. string string string string
The distanceof the bow from the bridgewas held Strad B 13 11 X.rurkevitch 14 14
Strad J 12 13 Phillips 13 11
at 4 cm; the resultswould be altered if either of Strad S 12 13 $angster's No. 30 12 10
Strad W 14 11 Sangster's No. 33 12 10
these numbers were changed. Strad G 13 11 Sangster's No. 34 13 9
Strad D 14 15 Sangster's No. 35 1t 9
According
to thesenumbers
theeasiest
violin Strad average
Guarnerius R
12.9
17
12.4
12
Sangster's No.
$angster's No.
37
38
11
13
9
9
to playin thislist is the Balistrieri(1776,owned Guarnerius
Balistrieri
H 15
9
10
g
Stanley
Koch
11
12
9
10
by W. Wolfinsohn). The next"best"is oneof the Stainer
P. Guarnerius
20
16
14
17
Moennig, copy of
Strad J, first It 8
Moennigs,which was perhapstwo years old Strad $. second
Strad B
13
13
11
-11
x0henmeasured.The Sangstersaveragea trifle Violin X ($5.00) 16 12
easier'l•hafi"the'Strad•, hnd the latter are' easier
182 F. A. SAUNDERS

vibrating, in order to be surethat what one hears total effect. After the bow made full contact, the
comesonly from the body, or the included air. tone grew in volume for about 0.012 sec., then
A gifted violinist may have only to hold the lasted at full volume for 0.014 sec. Then the bow
violin resting on his extended hand and feel its left the string, taking about 0.004 sec. to do so,
vibration while he talks to it. Such effectsdepend and an irregular motion occurred in this time
on the lightnessof the violin, and on the rate of while the string was taking on its free type of
growth or decay of vibrations in it. vibration. The decay then followed, lasting about
The growth of sound when a bow touches a 0.018 sec. Finally there was an interval of 0.02
string is a complex affair, and takes place at a sec. of almost complete silence.The period of
different rate from the decay of the sound after growth of the tone was lessthan that of decay,
the bow leavesthe string. Our first experiments and the form of the vibration was different, as
on growth and decay were carried out with the one might expectsincein the first casethe bow
help of Mr. Heifetz, who played his Guarnerius, was driving the string, while .in the secondthe
and then the $5.00 fiddle, in front of a micro- stringwas free. Thesefigureswere obtainedfrom
phone.With a speciallyquick-actingcircuit and a record of the cheap fiddle, which was the one
a high speedmoving-filmcamerawe were able to which reactedmore quickly. With a goodinstru-
record the motions of the spot of a cathode-ray ment the times of growth and of decay are both
oscilloscope,thus getting the wave-formsof the longer, as was shown from the records of the
vibrations in great detail. The film moved at a Guarnerius. In fact the decay was so slow that
rate from 200 to 500 cm/sec. A repeatingpattern there was no longer complete silencebetween
of eight toneswas played spiccatowith extraor- successivetones, though this differencewas not
dinary speedand cleanness,at the rate of about noticed by the listener.
14 tones per second.An interval of silence was Figure 5 gives a reproductionof part of the
found between each pair of tones except in the recordof Heifetz playingthe rapid fingerexercise
caseof one of them (A 880) which resonatedwith already mentionedon his Guarnerius. The end of
a harmonic of an idle string. The tone of the the decay of a tone of frequency 880 c.p.s. is
Guarnerius was louder than that of the cheap shownand the beginningof a tone of frequency
fiddle, and took longer to grow or to decay by 740 c.p.s. The open A string resonatedwith its
some 25 percent. But the form of the complex second partial tone in the interval between,
waves which were produced altered so much though this is mixed with other tones from the
during a changein loudnessthat measurements violin, whosevibrationswere dying out slowly.
of ratesof growth or of decaywerevery uncertain The film speedwas 240 cm/sec. Here the maxi-
(seeFig. 5). This alteration was due to the fact mum amplitude goes off the film. The figures
that the rate of growth or decay changeswith given in the previous paragraph apply to this
frequency. record also, approximately, as they are charac-
Some details of a single typical record are teristic of the player rather than the instrument.
worth giving. The tone had a frequencyof 569 The beginningof bow contact for the secondtone
c.p.s.; the film speed was 370 cm/sec., and the is very definitely shown in the figure, and the
record of this tone covered 26 cm of film. In high-frequencycomponentin the irregular mo-
spiccato bowing the bow bouncesclear of the tion at that stage has now a frequency near
string between one tone and the next. The initial 10,000 c.p.s.
contact of the bow with the string showsclearly Sincethe experimentsjust describedcouldnot
and can be measured to about 0.0001 sec. For yield accuratemeasurementsof the rate of decay
about 0.005 sec. the contact was being made, of a vibrating violin, we turned to the method of
more and more of the bow-hairscomingto grips pure-tone excitation (article D). By this method
with the string. During this interval an irregular of shakinga violin at one frequencyat a time, it
motion occurredwith one or more strong high was easy to choose particular frequencies at
frequencycomponents(5000 to 7000 c.p.s.), but whichto take high speed.records of decay.This
this soundwas both shortand relatively weak, so was done for a considerable number of violins,
that it probably did not contributemuch to the both old and new. Records were also taken of
 MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIN FAMILY 183

O. Ol sec. with the body resonances

it was oftenas highas
usual. Assuming a low efficiency for the violin,
the damping must be due mainly to viscosityin
the wood,with onepossibleexception,namelyat
the main air-resonancefrequency.
> ? The frequencieswhich we choseas points at
Decay, Contact Drive which to take readingswere almost all resonant
peaks, and they are recordedin Table IX as air
or body resonances,with their approximate fre-
FIG. 5. A high speedrecordof the vibrations produce(
by Heifetz on his Guarnerius near the interval between quencies.The values of/i are given in this table
two short tones. for various violins and for a few violas. The
damping coefficientshere given may be in error
knuckle knockings, in which case the vibrations by as much as ten percent. The knocking times
disclosedwere made up of the main air resonance, measure the duration of the air resonance, since
which persistedlonger than any others, together this is the most persistentsoundafter the knock.
with two or three body-resonant vibrations. The
The amount of energygiven to the violin by the
duration of the main air vibration could be
knock varies somewhatfrom one trial to another,
measuredalmost equally well by both methodsof but that given by the electromagneticexcitation
excitation. is much more uniform. The variation in time of
The logarithmicdecrement0) of the amplitude decay from the knocking is not large, as the
on the recordsis found from the equation initial decay is very rapid. Unfortunately, this
•= (2.303 'n) log]0(At/A), tableis not completein all items,as the necessary
observationscould not always be made.
where n is the number of vibrations between the
When the electromagneticexcitation is cut off,
two whose amplitudes were measured,A0 is the there is a redistribution of energy among the
initial amplitude, and A the final amplitude natural modes of vibration of the violin, and this
selectedfor measurement.If the decay is expo- gives rise to beats. The beats have frequencies
nential, /i should be the same at all frequencies which are the differences of frequency of the
produced by the same type of vibration. The strongest adjacent natural modes. Such beats
damping of the air resonancemay possibly be were first observed by Knudsen in a small room
different from that of the body resonances,since in which pure tones of low frequency and con-
it may be caused largely by radiation of sound siderable energy were being started and then
through the/-holes, and lessby viscosit3'in the stopped. In our case this redistribution accounts
wood. Of course, the body resonant vibrations for the fact that the first part of the decay is not
also loseenergy by radiation, but the processis exponential. Observations taken of the decay of
likely to be lessefficient than in the caseof the air the first few vibrations led to higher values of the
resonance.
damping; the listed values were derived from
Thereare manyindicationsthat theenergyput measurements of some 15 to 20 vibrations, after
into a violin is largely dissipatedas heat in the most of the initial disturbance had subsided.
wooditself,sothat only a smallpart of it escapes The conclusions that can be drawn from a
as sound.The latest work on this was by Rohloffs study of thesedamping cocfficientsare again as
who measured tile motion of the wood in different
confusingas before. The average knocking time
placesover the violin when it was made to vibrate of the old instrumentsis slightly longer than for
over a considerablerange of frequenciesboth in the new, but the damping coefficientsaverage
ordinary air and in a partial vacuum of about about the same. Individual new instruments (e.g.,
1.7 cm of mercury pressure. In the best violins the Yurkevitch and one by Moennig) have as lmv
the energy at the air-resonant frequency was damping as tile lowest of the Strads (B and J).
ahnostentirelyabsentat the lowpressure,though The values for Strad S (a famousand beautiful
•E. Rohloff, Ann. d. Physik •8, 177 (1940); Zeits. f. violin) would, of themselves,place it among the
Physik 177, 64 0940). least promisingof the new instrmnents.On the
MECHANICAL ACTION OF INSTRUMENTS OF THE VIOLIN FAMILY 183

with the body resonancesit was often as high as

usual. Assuminga low efficiencyfor the violin,
the damping must be due mainly to viseosiW in
the wood, with one possibleexception,namely at
the main air-resonancefrequency.
The frequencieswhich we chose as points at
which to take readingswere almost all resonant
peaks,and they are recordedin Table IX as air
or body resonances,with their approximate fre-
Fro. 5. A high speedrecord of the vibrations produced
by Heifetz on his Guarnerius near the interval between quencies.The values of •i are given in this table
two short tones. for various violins and for a few violas. The
dampingcoefficientshere given may be in error
knuckle knockings, in which case the vibrations by as much as ten percent. The knocking times
disclosedwere made up of the main air resonance, measure the duration of the air resonance, since
which persistedlonger than any others, together this is the most persistentsoundafter the knock.
with two or three body-resonantvibrations. The The amount of energygiven to the violin by the
duration of the main air vibration could be knock varies somewhat from one trial to another,
measuredalmostequallywell by both methodsof but that givenby the electromaghetic
excitation
excitation. is much more uniform. The variation in time of
The logarithmicdecrement(•i) of the amplitude decay from the knocking is not large, as the
on the recordsis found from the equation initial decay is very rapid. Unfortunately, this
/J= (2.303/n) Iog•0(Ao/A), table is not completein all items, as the necessary
observations could not always be made.
where n is the number of vibrations between the
When the electromagneticexcitation is cut off,
two whose amplitudes were measured,A0 is the there is a redistribution of energy among the
initial amplitude, and .4 the final amplitude natural modesof vibration of the violin, and this
selectedfor measurement.If the decay is expo- gives rise to beats. The beats have frequencies
nential, • shouldbe the sameat all frequencies which are the differencesof frequency of the
produced by the same type of vibration. The strongestadjacent natural modes. Such beats
damping of the air resonancemay possiblybe were first observedby Knudsen in a small room
different from that of the body resonances,since in which pure tonesof low frequencyand con-
it may be causedlargely by radiation of sound siderable energy were being started and then
through the f-holes, and lessby viseosiW in the stopped.In our casethis redistributionaccounts
wood. Of course, the body resonant vibrations for the fact that the first part of the decay is not
also loseenergy by radiation, but the processis exponential.Observationstaken of the decay of
likely to be lessefficientthan in the caseof the air the first few vibrationsled to highervaluesof the
resonance.
damping; the listed values were derived from
Therearemanyindications that theenergyput measurements of some 15 to 20 vibrations, after
into a violin is largely dissipatedas heat in the most of the initial disturbance had subsided.
wooditself, so that only a smallpart of it escapes The conclusions that can be drawn from a
as sound.The latest work on this was by Rohloffa study of these damping coefficientsare again as
who measured the motion of the wood in different confusingas before.The averageknockingtime
placesover the violin when it was made to vibrate of the old instrumentsis slightly longerthan for
over a considerablerange of frequenciesboth in the new, but the damping coefficientsaverage
ordinary air and in a partial vacuum of about about the sa•ne.Individual newinstruments (e.g.,
1.7 cm of mercury pressure. In the best violins the Yurkevitch and oneby Mocnnig) have as low
the energy at the air-resonant frequency was damping as the lowestof the Strads (B and J).
almostentirelyabsentat thelowpressure, though The values for Strad S (a famousand beautiful
•E. Rohioff, Ann. d. Physik /18, 177 (1940); Zeits. f. violin) would,of themselves,
placeit amongthe
Physik 177, 64 (1940). leastpromisingof the new instruments.On the
MECHANICAL ACTION OF INSTRUMENTS
have led to the same result. We know of one
conductedby Mr. Stokowskion the membersof
an orchestra,and another by the famousviolinist
Isaye on a groupof professionalmusicians;there
are several others recorded.
We have not made accurate measurements on
the maximum volume of tone that an artist can
produce on violins of different sorts, nor on the
efficiencyof theseinstrumentsas producersof
sound energy; these may be important to the
artist, but we are sure that they are not to the
listener, who neither knows nor cares how hard
the player is working, and whose impressionof
loudnessdependsas much on the acousticsof the
concert hall as on anything else. The crucial
quality which we seek doesnot lie in this field.
The extraneous noises producedby the scraping strong forces which the tension of the strings
of the bow, especiallyat momentsof contactwith imposeson the instrument.Modern makershave
the string, or releaseof the string, are so small as said that one cannotsafelymake a violin top as
not to enter into the listener'sjudgment. They
are not noticedat any considerabledistance,and
they are soweak in the caseof steady tonesthat
our analysesdo not showthem; that is, they are
at least 30 db down from the maximum of the
componentpartials. The "scratchy" tone of the
cheapestviolins is almost certainly due to the
presencein their steady tonesof an excessamount
of very high frequencyenergy.
There is one experiment on violins which is
very difficult to carry to completion; this is to
find the effect of centuries of time on them. It
seemsto be a general belief that a good violin
improves with age; whether a bad one also im-
proves no onh seems to know. There are two
effectsthat might be included in the processof
ageing;(1) chemicaland physicalchangesin the
wood, or in the varnish, due to the passageof
time, with its inevitableaccompanyingfluctua-
tions of temperature,atmosphericmoisture,im-
puritiesin the air, etc.; and (2) effectsdue to the
playing process.The latter consist mainly of
effects due to continued vibration, but the
influenceof the heatand moisturecontributedby
the player may greatly accelerate the chemical
and physical changesmentioned above. The air
near a player's body in warm weather is almost
saturated with water vapor, and a considerable
amount of this will be absorbed by the violin
during a long playing session.This will make the
wood expand across the grain, and thereby
OF THE VIOLIN FAMILY 185
changethe forcesin the violin body somewhat.
Daily playingfor yearsmight thusdo something
to a violin which would not happento one which
is resting in its case.
Our measurements indicate that old violins
weigh lesson the average than new ones.Seven
Stradsvaried from 373 to 394 grams(with chin-
restsremoved);average383 grams.Six other old
violins varied from 354 (Stainer) to 389 grams;
average 374. Thirteen new violins varied from
381 to 435 grams;average410 grams. The three
violins in the Heifetz test weighedfrom 391 to
413 grams; average 410 grams. The new violins
are 7 percent heavier than the Strads. Naturally,
the lighter a violin is, the easier it is to shake. It
ought to be as light as is safe, consideringthe
thin as thoseof Stradivariuswithout danger of
collapse.This can hardly be due to any changein
200 years in the nature of the wood available
from the tree whosewood is almost universally
preferred, the Norway spruce,Picea excelsa.It is
true that the tension of violin strings is now
greater than it was 200 years ago, becauseof the
rise of musical pitch; but the Strads, fitted with
strongerbass-bars,are capableof withstanding
this. So we concludethat the wood has gained
strength with age, or that it was treated by the
early makers in such a way as to increase the
strength at the time, or at least to make an in-
creasein strength more likely through age. It has
been stated that the few Strads that have sur-
vived without having beenplayed very much are
not as good as those which have been in active
use. It is also true that Strads have growngreatly
in reputation, and probably in merit, sincethey
were new. These facts point to an aging process
acceleratedby playing, which is responsiblefor
the present strength of the wood in Strads.
We have already mentioned (article A) that
with age and long playing a certain amount of
cracking of the glue is found in the purfling, and
perhaps elsewhere,so that the top of the violin
has becomelesstightly bound, and is better able
to vibrate. This may be an important part of the
ageing process, though it does not satisfy the
requirement of an increasein the strength of the
wood.
186 F. A. SAUNDERS

More researchis neededon the effectsof age, Our sincerethanksare due to many artists and
and of continued vibration, on wood, and this is collectorsfor permissionto test their precious
most convenientlydone on small strips of wood, instruments, and in some cases for active as-
rather than on wholeviolins,wherethe properties sistance in the tests. The list includes Jascha
of the wood (Young's coefficient of elasticity, Heifetz, Wolfe Wolfinsohn, JacquesGordon, Leo
density,velocityof soundin the Wood,damping Reisman, Z. Balokovic, the members of the
coefficient, etc.), can conveniently be measured. Curtis Quartet (Messrs•Brodsky, Jaffe, Aronoff,
At the same time the effects of various varnish and Cole), Marcel Dick, Gerald Warburg,
Bernard Robbins, William Lincer, Miss Thelma
treatments (or of the first "filler") can be
studied, and a searchbegun for a substancewith Given, Maurice Stad, Mrs. W. E. Ellcry, Mrs.
Olga Petzoff, Mrs. E. Ginn, Miss Lillian Shattuck,
which to impregnate the wood so as to add to its
Miss Sally Dodge, Henry Guerlac, Kerr Atkinson,
strength, protect it from moisture, decreaseits
H. S. Shaw, A. P. Saunders, Malcolm Holmes,
damping, etc. One might also searchfor better
F. C. Keyes, and many others, both amateur
woods,now that wood from all over the world can
musiciansand violin makers. We are especially
be obtained. Some of these studies have been
indebted to ProfessorCurt Sachs for advice, and
started,conductedby variousinterestedpeople, to E. H. Sangster, Wm. Moennig and Son, and
and methods of treatment involving heat or J. A. Gould and Sons for loans of instruments,
changesin pressureare being tried. As far as we adjustments, etc. Finally, the credit for the
know none of these methods have as yet led to operation andadjustm. ent of theapparatus, and
any very striking results. Future work must for the construction of much of it, goes to Dr.
includea longprogramon treatedstripsincluding Robert B. Watson without whose skillful help
further studiesof their damping. this research could not have been done.