1990 - Greenwood - Cochlear Frequency-Position Function For Several Species - 29 Years Later

A cochlear frequency-position function for several
species29
years later
Donald D. Greenwood
SchoolofAudiology
andSpeech
Sciences,
University
ofBritishColumbia,Vancouver,
BritishColumbia
V6T1WS, Canada
( Received30 May 1989;accepted

for publication26 January1990)
Accuratecochlearfrequency-position
functionsbasedon physiological
datawouldfacilitate
theinterpretation
of physiological
andpsychoacoustic
datawithinandacross
species.
Such
functions
mightaidin developing
cochlear
models,andcochlear
coordinates
couldprovide
potentiallyusefulspectraltransforms
of speech
andotheracoustic
signals.In 1961,an almostexponential
functionwasdeveloped
(Greenwood,
196lb, 1974)by integrating
anexponential
functionfittedto a subsetof frequencyresolution-integration
estimates(criticalbandwidths).
The resultingfrequency-position
functionwasfoundto fit cochlearobservations
on human
cadaverearsquitewelland,with changes
of constants,
thoseonelephant,cow,guineapig,rat,
mouse,and chicken(Bksy, 1960), aswell as/n vivo(behavioral-anatomical)dataon cats
(Schucknecht,
1953). Since1961,newmechanical
andotherphysiological
datahaveappeared
on thehuman,cat,guineapig,chinchilla,monkey,andgerbil.It is shownherethat the newer
extended
dataonhumancadaverearsandfromlivinganimalpreparations
arequitewellfit by
thesamebasicfunction.
Thefunction
essentially
requires
onlyempirical
adjustment
of a single
parameterto setan upperfrequencylimit, while a "slope"parametercanbe left constantif
cochlear
partitionlengthisnormalized
to 1 or scaledif distance
is specified
in physical
units.
Constancy
of slopeandformin deadandlivingearsandacross
species
increases
the
probability
thatthefunctionfittinghumancadaverdatamayapplyaswellto thelivinghuman
ear.Thisprospect
increases
thefunction's
valuein plottingauditorydataandin modeling
concerned
withspeech
andotherbioacoustic
signals,
sinceit fitstheavailable
physiological
datawelland,consequently
(if thosedataarecorrect),remainsindependent
of, andan
appropriate
meansto examine,psychoacoustic
dataandassumptions.
PACS numbers: 43.64.Kc, 43.64.Bt
INTRODUCTION
brane. The latter hypothesishad beenadvancedand supportedinfluentiallyby Fletcher ( 1940, 1953) and Zwicker et
Sincethe late 1960s,more data on the frequency-posi- al. (1957). Correspondence
of critical,or other,bandwidths
tion coordinates of the cochlea have become available for a
to equal,althoughunknown,distanceson the basilarmemnumberof speciesand supplementearlier data, gatheredby
branewould imply proportionalityto the derivativeof a freBksy in the 1940s and Schucknechtin the 1950s. The
quency-positionfunction of the membrane.The paper of
newerdata-on man, cat, chinchilla,guineapig, gerbil,and
1961 simply integrated the suggestedcritical-bandwidth
monkeymhaveincluded additional species,and, in some function to obtain a frequency-positionfunction, in which
cases,achievedconsiderable
coverageof the cochlearpartipositionon the membranewas expressedin critical-band
tion. It may be usefulto comparesuchdata againwith the
units. The length of a critical-bandunit in physicalunits
simplefrequency-position
functionsdevelopedempirically
couldthenbe determinedby dividingthe lengthof the mem29 years ago from critical bandwidth data in man (Greenbrane by the number of critical bands end to end that subwood, 1961b, 1974). The newerdataincreaseempiricalsuptendedthe audiblefrequencyrange.
port for a family of suchalmost-exponential
frequency-posiBy coincidence,about 35 critical bands,accordingto
tion functions and for a scaling or normalization
this function,subtendedabout 35 physicalunits, millimerelationship,amongthesespecies,
that appearsto governthe
ters.The convenienteffectof this correspondence
wasthat
slopecoefficientof the function.This meansthat thesefuncthe frequency-position
functionthus obtainedcouldbe comtions differ essentiallyin only the other main constant.
pareddirectlyto B6k6sy'splot of frequencyversusposition
A review of this development(Greenwood, 1974) is
on the membranewithout a changeof slopeconstantto albrieflyrecapitulated.The originalfrequency-position
funclow distancex to be expressedin millimeters. The coincition wasderivedfrom a critical-bandfunctionproposedto fit
dencethusmeant that thesebandwidthswereproportional
my critical-band estimates in 1959-1960 (Greenwood,
to the derivative with a constant of 1 rather than with some
1961a,b). The developmentof the function assumedthat

critical bandwidths followed an exponential function:
other constant,as would have beenimplied by correspondenceto a differentdistancein millimeters.The comparison

had two major outcomes:The frequency-position
function
closelyagreedwith B6k6sy'scochlearcoordinates,and this
CB = 10(ax+ b), of distance

x alongthecochlear
partition,
and correspondedto a constantdistanceon the basilarmem2592
J. Acoust.Soc.Am.87 (6), June1990
0001-4966/90/062592-14500.80
@ 1990Acoustical
Societyof America
Downloaded 02 Oct 2012 to 128.151.164.140. Redistribution subject to ASA license or copyright; see http://asadl.org/terms
2592
providedsupportto the two ideasthat the initial criticalbandwidth estimates and some additional measurements of
also be related to the questionof what is the "effective"

lengthof the cochlea,and what is its effectiveapical endpoint, sofar asthe physicsof the cochleais concerned.
)
The constanta (essentiallythe slopeof the straightportion of the frequency-position
function,whenlog frequency
isplottedagainstcochlearposition)wasfoundnot onlyto be
scalableamongthe otherspeciesstudiedby Btktsy but in at
leastsomeof thosespecies,
to agreereasonablywith the logarithmic slopeof the volume-compliance
gradientsmeasured
by Btktsy along their cochlear partitions (Greenwood,
196lb). To say that a is scalerelated acrosstwo speciesis
simplyto saythat, if knownfor one,it canbeobtainedfor the
otherby mulitiplicationby a scalefactor,determinedin this
caseby theratio of their cochlearpartitionlengths.Or, to say
the samething, a timesbasilarlength would be a constant
among cochleasif they were scalerelated in this respect.
Thus this constantproductcanitselfbe takenasthe valueof
a amongsuchspeciesif cochlearpositionor distanceis expressedas a proportionof total partitionlength (apex = 0,
stapes= 1).
From both Btktsy's frequency-position
data and his
compliancedata,especiallyfor the humanspecies,
the product of a timesbasilarlengthappearedto be about2.1 in 1961.
As then noted, 10 raisedto this power is about 126, or the
frequencyseparations
in earlyexperimentson consonantintervals (Mayer, 1894)--later solidlyconfirmedby Plomp
and Levelt (1965) and Plomp and Steeneken(1968)might correspondto a constantdistanceand obeyan exponential function.As agreementwith the cochlearmap was
evident, these conclusionswere simply contingent on
whetherthephysiological
datawereaccurateandcharacterized equallyboth the quick and the dead.
However,our chiefinteresthereis in the frequency-positionfunctionitself,independentof the two hypotheses
underlyingits origin,whichmightor mightnot be true or generallyapplicableto anygivensetof bandwidthestimates.We
focusfirst on the mostimportantoutcome,the closeagreement of the frequency-position
functionwith Bksy'smeasurementsof cochlearcoordinates,the only physicalmeasurementsthen available,and secondon the degreeto which
the frequency-position
function may successfullyapply to
physicaland physiologicaldata from other species.
In 1961,the functionprovideda convenientmathematical expressionfor a cochlearfrequency-position
map that
was a rational scalefor plotting resultsand might assistin
their interpretationwithin and acrossspecies.The function
fitted not only Bksy's human cadaverdata and his similar
factor of about 100 with which Btktsy characterized the
resultsfrom six other speciesbut alsoSchuknecht'sin vivo
variation in stiffnessof the cochlearpartition from end to
cat data, essentiallyby the changeof only a singleconstant end.As describedabove,thisnormalizedslopeconstantwas
that determinedthe upperfrequencylimit. The other main
quiteadequatefor functionsfittingfrequency-position
data
constant,which governedslope,couldbe scaledor normalavailablein 1961. As will be seen,the new frequency-posiized acrossdead and living preparations,which hastended
tion data from man, cat, chinchilla,guineapig, monkey,and
to supportthe map'sapplicabilityto the living humancochperhapsgerbiltendto confirmthis valueof about2.1.
lea. Sincewe necessarily
rely still on physiologicaldata from
living preparationsof other speciesasthe mostnearlydirect
I. MORE RECENT FREQUENCY-POSITION
DATA FOR
sourceof inferenceto man, the relation of the 1961 function,
especiallyits slopeconstant,to the augmenteddataof recent

and future years will be important, among other reasons,
becauseof its bearingon the applicabilityof cadaverdata
and the functionto livinghumancochleas.
The frequency-position
functionobtainedas described
above is
F=A(lOa"--k),
(1)
wheresuitableconstants(for man) are:A = 165.4 (to yield

frequencyin Hz) and a = 0.06 (if x is expressed
in millimeters), or 2.1 (if x is expressedas a proportion of basilar
length). The latter constantis an empiricalconstantarising
in the critical-bandfunction.The integrationconstantk was
originallyleft at the value 1, but it may sometimesbe better
replacedby a numberfrom about0.8 to 0.9, to set a lower
frequencylimit dictatedby conventionor by the bestfit to
data. Thus the value k = 0.88 would yield the conventional
lower frequencylimit of 20 Hz for man, and this value was
THE HUMAN
SPECIES
We considerfirstdata from humantemporalbones,obtainedwith the Mtssbauertechniqueby Skarstein(Kringlebotn et al., 1979). The sevennew data pointswere obtained
from sevenfreshtemporalbones,8 to 24 h after death, at
recordingsitesextendingfrom about the 2.2- to 6.2-kHz
pointson the cochlearpartition.The pointscloselyextrapolatedBtktsy data, whichhad endedat the 2- kHz point, and
alsofit the frequency-position
functionproposedby Greenwood in 1961. The top panel of Fig. 1 is reprintedfrom
Kringlebotnet al. (1979). The samedata,replottedin Btktsy's 1942format, reappearin the bottompanelof Fig. 1,
with additionalpointsrepresented
byx's andcrosses
derived
from two plottingsby Btktsy in 1943and 1947of displace-
ment envelopesalong the basilar membrane [see also

Fletcher's(1953) and Zwislocki's(1965) plottingsof Btktsy'sthreeseriesof visualobservations
in relationto theoreticalcurves].l
Although death could reasonablybe expectedto have
almost used in 1961, but we will continue to use 1.0 for man
exerted some influence on both setsof data, that effect in the
andotherwise0.85 throughoutthispaperastwo valuesthat

seemadequatefor the momentfor mostof the speciesconsideredhere,althoughLibermanhasfoundthat 0.8 bestadjuststhisfunctionto hislow-frequency
datapointsin thecat
and (with appropriateA) sets90 Hz aslowerfrequencylimit. (However,thequestionof theappropriatevaluefor k may
observations
presentedmay havebeenrelativelylimited. Btktsy'saccountsof proceduresindicatethat, observinga given turn, he not only made observationsof amplitudemillimeter by millimeter along the partition away from the
positionof maximumamplitude,but that "phasemeasurements provided a sharp definition of position, especially
2593
Donald D. Greenwood:Cochlearfrequency-position
function
J. Acoust.Soc. Am., Vol. 87, No. 6, June 1990
2593
vo.
.rgs
2O
,9, 10
2O
2O0
100
2O0O
2OOO0tZ
1000
10000
35t ......., ........ , ........ ' /

c30
.--
x25
<20
010
o
100
1000
Frequency
10000
in
Hz
FIG. 1. Top panel:Takenfrom Kringlebotnet al. (Fig. 1, 1979). Distance

fromstapesto theplaceof maximumdisplacement,
asa functionof frequency. Solidpoints:B6k6sy(1960). Opensquares:
Skarstein,datapointsfrom
9.9 to 16.5 mm from the oval window. Curve: Greenwood [ 1961b, Function
( 1) of presentpaper]. Bottompanel'B6k6syandSkarsteindatareplottedin
the standardformat for this paper,with the additionof additionalpoints
obtainedfrom two later seriesof B6k6sy'sobservations.
Solid circles'B6k6sy( 1960, 1942series).Crosses:B6k6sy( 1960, 1943series);laterallydisplacedcrossat 54 Hz actuallyfallsat 50 Hz. X's' B6k6sy(1960, 1947series);laterallydisplacedX at 215 Hz actuallyfallsat 200Hz. Solidsquares:
Skarstein(Kringlebotnet al., 1979). Curve:A -- 165,a -- 0.06, k - 1.
whenthe vibratingpoint wasobservedasit passedthrough

the restingposition.Phasemeasurements
thereforeare particularly suitablefor the exactdeterminationof smalldisplacementsof the patternof vibration,especiallyasthesecannot
be determinedfrom the pattern of amplitude [italics added]." The phase measurementswere made through very
small observationwindows.Theseand other commentssuggestthat identifications
of the positionof maximumamplitude and positionchangeswere aided or corroboratedby
observations
of phaseandthat the graphicalrepresentations
of amplitudeversusbasilardistancewere subjectto consistencycheckswith observations
of zero crossings,
determinationsthat may makethe observations
presentedmorecoherent and accuratethan amplitudemeasurements
alonemight
allow.If the relativepositionsof the patternmaximawere,in
part, determinedfrom sharplyobservable
shiftsin the zero
crossings,
it suggests
onereasonB(k6sy's
determinations
of
2594
curveshapes(hencezero-crossingspacing)are lessaffected
by death than are the tuning curve peak positionsand
shapes.In further supportof the possibilitythat deathmay
not have much affectedthe frequency-position
coordinates
of B(ksyand Skarstein,Kohl16ffel's(1972a,b) data, relating frequencyto positionof maximumsensitivityin the guinea pig, indicatedthat, in the first day after death,there is
rather little shift of the pointsof maximum sensitivityfor a
givensetof frequencies
towardthebase.The datapointsstill
maintainedthe sameslopeandextrapolatedupperfrequency limit. Moreover,the shiftsin positionof maximumamplitude toward the basethat occurredafter the first day maintainedthe samerelativespacingof the maxima,and hence
the sameslopevalue of the frequency-position
functionin
the bestpreparations,for asmany as 5 or moredays.
Thus anyshiftsbasalwardof the B(ksy-Skarstein
data
pointsmightbe expectedto havenot muchaffectedthe slope
of the data curve. It seems a reasonable inference that B(-
patternposition,at leastof relativepatternposition[and

hencethe slopeconstantin Function( 1) ], may havebeen
relativelylittle influencedby death.
Rhode's(1973) observations
haveindicatedthat phase
J. Acoust. Soc. Am., Vol. 87, No. 6, June 1990
k(sy'stemporalbonepreparationsmay havebeencomparably freshand that the compositedata may possess

the slope
that would be expectedin living speciments.The fact that
the data extrapolateto about20 kHz, a commonlyaccepted
upperfrequencylimit in man, suggests
alsothat any basal
shift has been relatively small. Such deviationsof the data
from the superimposedfunction as occur in the lower frequenciesaremainlyin the expectedbasaldirection.If it were
the casethat Bksy's and Skarstein'sdata pointswere in
generallocatedtoo far basally (upward on the ordinate) by
about1 to 1.4mm andthat the extrapolatedupperfrequency
limit of about20.6 kHz, via Function ( 1), wereconsequently too low, thena correctivedata shiftof 1 to 1.4mm apically
wouldrequirea changein the constantA from 165to about
190 or 200, respectively,which would yield upperlimits of
about23.7 or 25 kHz, respectively.
II. CAT: NEW FREQUENCY-POSITION
REVISED BASILAR LENGTH
DATA AND A
A. Primary fiber CF and cochlear site of innervation
Very completefrequency-position
data for the cat have
beenpublishedby Liberman(1982), togetherwith a revised
estimateof basilarlengthat the pillar headsand free of the
errors inherent in the reconstructiontechniquesused by
Guild ( 1921) and Schuknecht(1953). In agreementwith
the scalingrelationproposedby Greenwood( 1961b,1974),
he confirmedthat the slopeconstanta agreedwith the expectedvalueof 0.084 if x is expressed
in mm--or, to saythe
samething, with the expectedvalue of 2.1 if distanceis expressedasa proportionof basilarlengthratherthan in mm.
(However, sinceLiberman preferredto expressnormalized
distanceas percentagerather than proportion, he used
0.021.) The completeness
of hisdata and the accuracyof the
fit made a striking confirmationof the scaledslopea and
henceof the value2.1 that remainedapplicablein 1982despitechangesin the experimentaldata, methods,and estiDonald D. Greenwood:Cochlear frequency-positionfunction
2594
matesof cochlearlength [from thoseof Schuknechtand

Neff ( 1952); Schuknechtand Sutton, (1953) ].
But, in addition, given his revisedestimateof 25 mm
(instead of 22 mm) for the average basilar membrane
length,whichalteredthe scalefactor,hisdata alsoindicated
that the constantA, which determinesthe upperfrequency
limit, wasnot scalablefrom the corresponding
constantapplicableto humandata.SuchanA, if scaledfromman,ledto
a functionparallel to, but displacedfrom, the new cat data.
In short, the earlier relation (suggestedby Schuknecht's
lengthestimateand parallelbut shifteddata), in which the
cat upper frequencylimit indicatedby the displaceddata
equaledthe upperfrequencylimit in man multipliedby the
squareof the scalefactor,wasshownby the correctedlength
andnewdata to havebeenadventitious,asit may well bealso
for the elephant,which,amongthe eightspeciesconsidered
in 1961, was the only other whoseupper frequencylimit
suggested
this relation (Greenwood,1961b). However,the
parallelcoursesof the newcat dataandthe displacedcurve,
ofbasilarlengthor, if averagemillimetermeasureis desired,

while 2.1 is divided by 25 mm (instead of by 22 mm, as in
1961).
B. Other cat data: Spiral ganglion cell CF and

mechanical
measurement
Additionalfrequency-position
datahadearlierbeenobtained in the cat by Kohl16ffel(1974, 1975). The CFs of
singlespiralganglioncellswererelatedto celllocations,relativeto the baseof the cochlea,by radiallyprojectingfrom the
locusof penetrationin theganglionto thebasilarmembrane.
Thus thesedata alsowerefrom livingcochleasunaffectedby
errorsof reconstruction
andpresumablyalsolongeron averagethan the older 22-mm estimateof meanlength.All but
one of 105 cells were 2 to 4 mm from the basal end, and all
but three,when groupedfor analysis,fell in 2-dimensional

"bins" (0.5 mm X 5 kHz) whosecentral CFs were between
20 and 35 kHz. The coordinatesof the cellswere in quite

goodagreement,whencalculations
of positionarerelatedto
whose value of A was too small, demonstrated that, for an
the basal end, with both Liberman's data and Function ( 1)
appropriatelyselectedA, generatingan empiricallycorrect

upperfrequencylimit, Function( 1) mustfit the newdataas
in Fig. 2 and with the original 22-mm functionsincethe

functionsdifferrather little, relativeto the basalend,in plotting high CFs.
well, or better than, it had fit the old.
Hence, as Greenwoodhad done in 1961 for the eight

speciesconsideredthen, an appropriatevaluefor the constantA for thecatwassimplydeterminedfromthenewdata,
which Liberman found yielded a best-fittingvalue of 456
(for Hz), as comparedto the originalvalueof about418 in
1961.Figure 2 presentsthe cat data in relationto Function
( 1), when Liberman best-fittedall constantsto the data. In
the final analysis,the 1961 cat functionhasbecomea function well fitting the new data by revisingA upwardby about
9% andby reducingk from 1 to 0.8 to providea betterfit to
the new low-frequencydata points,while the expecteda remainsat 2.1 whendistanceis to beexpressed
asa proportion
...,
0.
0.6
1.0-
o.i
I '1 [ i ii1|
'
i.o
I I I Ill I
io
CHARACTERISTIC FREQUENCY
I I'l
60
(kHz)
FIG. 2. AdaptedfromLiberman(1982). CurveisFunction( 1) fittedto the

data,withx expressed
asa proportionof totalnormalizedlengthfromthe
apex.Data displayrelationbetweenprimaryfiberCF andcochlearlocation
in the cat; eachof the 52 fiber locationswasnormalizedto its respective
cochlea.Circle, X, andfilledtrianglereferto high,medium,and low rates
of spontaneous
discharge.
Meanlengthof cochleaandbest-fitting
constants
areprovided
by Libermanasfollows:Meanlengthis25 mm;A (for frequencyin Hz) becomes
456 (a 9% increase
fromthe 1961functionfittingSchuknecht'sdata); a remains2.1 asin man,or becomes
2.1/25 = 0.084 (rather
Two bodies of mechanical data had been obtained over
the basal 9 mm of the cat cochlea,by Wilson and Evans

(1977) and in a smallbasalregionby Khanna and Leonard
(1982). The data of Wilson and Evans had demonstrated
goodagreementin slopewith Schuknecht's(1953) basal

dataandthe 1961function.However,the slopeof thesedata,
judgedvisually,may be slightlymoregradualthan the 1961
function(and Schuknecht's
data), both of whichwereplotted on a 22-mm membrane, which would indicate that the
Wilson and Evans slopeis in still closeragreementwith

Function( 1) andLiberman'sdatain Fig. 2, whichpertainto
a 25-mm membrane.This closeness
of slopewas showndirectly by Liberman'sreplotin his Fig. 8 of the Wilson and
Evansdata in comparisonto his data in Fig. 2. Although
theirslopeisparallelto Liberman'sdatacurve,themechanical data are displacedtoward lower frequencieson the abscissa(basally on the ordinate). On the samegraph, the
Khanna and Leonarddata agreein displacementwith the
replottedWilson and Evansdata, while Kohl16ffel'sneural
data agree,asnotedabove,with Liberman'sneuraldata.
In summary,the frequency-position
functionfor the cat
appearsto bewell andconsistently
determinedby fivebodies
of data.Liberman'sdata,by coveringthe wholecochleaand
establishingthe most reliableestimateof cochlearlength,
well supportsthe function'sform whileit enhances
the consistencyof the slopeconstanta, whoseapparentscalability
acrossspeciesis our secondfocusof interest.However,the
paralleldisplacements
notedabovebetweensomebodiesof
datarequirecommentandare relevantto the interpretation
of someof the frequency-place
data reportedfor other species.
C. Why might mechanical peak frequencies be

displaced basally from primary-fiber CF data?
cochlearlength;k = 0.8 (rather than 1).
Althoughthe slopesarethe same,thepeakfrequencies

in the mechanicaldata are displacedfrom Liberman'sdata
2595
Donald D. Greenwood:Cochlear frequency-positionfunction
than 2.1/22 as in 1961) ifx is to be scaledin mm and referencedto mean
2595
relating unit CF to position.Liberman listedtwo possible

explanations:
Cochlearmechanicsare disruptedby the proceduresrequiredto make the mechanicalmeasurements;
or,
alternatively,that the frequencyrequiredto producemaximum amplitudeat a givenspotis simplylowerthan the CF
of a primary fiberinnervatingthe samespot.However,the
displacement
is about6% of cochlearlength,or 1.5 mm on
the ordinate.This displacement
wouldseemmoreeasilyunderstoodon the first basisthan the second,for the following
reasons.
On the hypothesisof cochleardisruption,a shift of the

maximumamplitudeof a displacementenvelopetowardthe
basewith cochlearinjury wouldaccountfor the requirement
of a lower frequencythan normal to placethat maximumat
any givenpoint.Suchshiftsdueto injury, anoxia,and death
have beenseenin severalstudies(Kohll/Sffel, 1972b;Rhode,
1973; LePageand Johnstone,1980; LePage 1981;Khanna
and Leonard, 1982; Sellick et al., 1982;Robleset al., 1986),
and a shift of 1.5 mm may not be too large.
As for the secondhypothesis,note that an increaseof
stimulus frequencysuch that the displacementenvelope
shifted1.5mmbasally(between
andoctin thecat) before
the stimulusfrequencyreachedthe CF of the primary neuron innervatingthat pointwouldbe largerelativeto the apical segmentof the envelope.Sincethe apicalsegmentof the
envelopein the basalhalf of the cochleais probablyonly
about0.71 mm in the squirrelmonkey (if cochlearlengthis
20 mm, but see later comment) and about 0.66 mm in the
The same basal shift of the displacementmaximum,

owing to cochlearinjury, that would accountfor the displacementof the mechanicaldatafrom Liberman'sprimaryCF data and the curvein Fig. 2 shouldconstituteprobably
the most important factor causing(a) lower fiber CF with
elevationof fiberthreshold,and (b) a lowercutofffrequency
for hearinglossthan the normal CF for the point at which
damagebegins.
Thus, if outer hair-cell (OHC) damageor deathlinearizesand reducesbasilarmotion at a givenpoint innervated
by a primaryfiberunderstudyandcausesa basalshiftof the
positionof maximumamplitudethat would normallyoccur
at this point for a tone at the point'snormal CF, then these
changesalsoshouldraisethe studiedfiber'sthresholdto a
tone at its normal (and now ex-) CF. Thresholds will be less
affected,however,for tone frequencieslower than the former CF, frequencies

at whichthe point'sresponse
isnormally linear or more nearly linear. Moreover, tonesof theselower frequencieswill now be required in order to place
maximum displacementamplitudeat this point. Thus, despite higher thresholdsat the lower frequencies,the now
greaterrelativesensitivityof the basilarpoint and its innervatingfiberto toneslower than the former CF will establish
a new and lower CF for both point and fiber. As a result,
also, behavioral threshold to tones at the former CF of units
innervatingthe damagedpointsshouldbe higher,making

hearinglossbeginat frequencieslower than the normal CF
for the point at which damagebegins.
guineapig ( Greenwood,1974), it isprobablynomorethana

scaled distance of about 0.9 mm in the cat. A 1.5-mm shift,
largerthan the lengthof the apicalsegmentof the envelope,

arguesthat the secondhypothesisabove,takensingly,would
requirethat the envelopebecomea maximallyeffectivestimulusfor a neuroninnervatingthis pointwhenthe envelopeis
shiftedsofar basallythat eventhe apicalfootof the envelope
is about 0.6 mm basalto the point in question.A primary
neuron excited most effectively (at lowest threshold) by a
tone whoselargestamplitudeeffectsdo not reachthe neuron'spoint of innervationwould not easilybe accountedfor
with current conceptions.It is conceivable,of course,that
both alternativehypotheses
togethercouldeachaccountfor
a part of the displacement
of the datacurves,but the second
for only a presumablylimited part.
III. CHINCHILLA:
FREQUENCY
OF HEARING
VERSUS POSITION OF DAMAGE
LOSS
sameinterval noted above).
A map of frequencyversuspositionfor the chinchilla

(Eldredge et al., 1981) is basedon the relation betweenaudiometricfeaturessuchas notchesor abrupt transitionsin
sensitivityto correspondinglesionsof the organof Corti.
They discussedthe considerablevariability in data of this
kind in a review of the experimentalmaterials,methods,
problems,and sourcesof variability.Although the variability in suchdatareducestheir powerto indicatedifferences
in
goodness
of fit to functionsof differentform, the methodology and quantityof data allow confidencethat the coefficient
of the exponentof the simpleexponentialfunctionfitted to
the data (i.e., the slopeconstantwhen frequencyis logged)
mustcloselyapproximatethe averageslopeof anyotherempiricalor theoreticalcurveto be comparedeventuallyto the
data. Six correlationsby Ryan and Dallos (1975) of OHC
losswith the cornerfrequencyof hearingloss(rangingfrom
1-8 kHz) yield locationsthat the presentauthor finds fit
within therangeof variationfoundby Eldredgeetal. andare
on averageabout 0.75 mm aboveFunction ( 1) in Fig. 3.
Eldredgeet al. give the averagelength of the cochlear
partitionin the chinchillaas 18.4 mm and reportthe chinchilla hasnearlythe samefrequencyrangeasman. The section they consideredwasbasalto a point about5.6 mm from
the apex;the corresponding
point in man is about 10.6 mm
from the apex.This is a restrictionin both casesto the frequencyrange aboveabout 500 Hz, the lower limit of the
chinchilladata and the frequencyabovewhichBksy'shuman data arewell knownto approximatea straightline on a
2596
Donald D. Greenwood: Cochlear frequency-positionfunction
D. Why might frequency-place correlations of hearing

loss be displaced basally from primary-fiber CF
correlations with place?
Liberman (1982) also summarized data basedon corre-
lationsof the CFs of singleauditoryunitsshowingthreshold

shifts(Liberman and Kiang, 1978) with placeof noisedamage,in comparisonto the data of Schuknecht(1953), which
correlatedfrequencyof hearinglosswith cochleardamage.
Both setsof data showa systematicshift toward lower frequenciesawayfrom the curveshownin Fig. 2 basedon primnary fiber CF. Liberman notesthat Robertsonet al. (1980)
havesuggested
that, in casesof chronicsurgicallesionsto the
organof Corti, the CFs of fibersfrom damagedsitescanshift
to lowerfrequencies
thatareasmuchasto octaway,(the
2596
mm
ioo
I
6
8o
14
12
o60
i0
"40
I .'
Ol
I
0.2
0.5
I
1.0
1,1
2.0
.5.0
I
I0
ever, the chinchilla data, like the cat data of Schuknecht
(1953) and Liberman and Kiang (1978), are basedon a

correlationof hearinglosscutofffrequencies
andpositionsof
cochlearlesions.If the chinchilla curve correlatingcutoff
frequencyof hearinglossto positionturned out to be displacedupward on the ordinate (from one relatingunit CF,
or cochlear-pointCF, to position)by an averagedistanceof
aboutone-thirdof a millimeter, an equalcompensatoryshift
of Function (1) downward in order to representprimary
fiber-CF data would increasethe upper frequencylimit by
only about 2 kHz. The shifted curve would also still pass
closeto the mechanicalobservation,deviatingby only about
0.19 mm, this time passingbelowit. In short,the chinchilla
data, even if systematicallydisplaced somewhat from a
curve that might be basedon primary-CF data or from a
curvethat describedthe true positionsof displacementenvelope peaks,not only provide a well-definedslopebut probably are quitecloseto a curvethat mightbebasedon further
20 k Hz
Frequency in Kilohertz
0.1
Also shownin Fig. 3 (solid circle) is the determination

by Robleset al. (1985) ofa CF of 8.35kHz for a pointabout
3.5 mm from the baseof the cochlea.This point deviates
apicallyfrom Function (1) by only about0.13 mm. How-
, . I,,,,I
fice.
2O
theconstantA in Eq. ( 1) issetto about163.5,a valuenearly

the sameasthe constantfor man, whichwouldequallysuf-
10
100
' ........
' ........
'/
<X
80....
direct determinationslike those of Robles et al. (1985).

,,,,!
'
O. 1
, , ,,,,
Fneuency
......
..10
n K I ohentz
FIG. 3. Upper graph:Taken from Eldredgeet al. ( 1981). Relationof frequencyof hearinglossin chinchillato placeof cochlearlesion,basedon
audiometricfeaturessuchas notclesor abrupttransitionsin sensitivity.
Straightlineexpresses
thesimpleexponential
relationprovided
byEldredge
et al. asasbest-fitto theirdata.Dashedsection:
regionof no data--apical
30% of cochlea;frequencyrangerepresented
by dataliesbetween0.5 kHz
to about15kHz. Lowergraph:Straightlinerepeatsthesamesimpleexponentialfunctionaboveprovidedby Eldredgeet al. Dashedsectionasabove.
Curvedline represents
Function( 1) with samea applicableto man or cat
(or 2.1/18.4 ifx is to beexpressed
in mm of an averagecochlea,A = 163.5
asdetermined
by dataof Eldredgeet al. (to yieldthesameupperfrequency
limit astheirexponential),andk isleftat 0.85,whichseemsadequateto the
purpose.
Solidpointprovidedby RuggeroandRobles( 1984,personalcommunication;
Robleset al., 1985).A parallelshiftof thestraightpartof the
curvedownwardby about1/3 mm wouldincrease
upperfrequencylimit by
about2 kHz. For comparison,a similarshiftdownwardby about 1.25mm
wouldincreasethe upperfrequencylimit to about30 kHz.
log-linearplot. The coefiScient

of the exponential
that Eldredgeet al. fitted to the chinchilladata (basilardistance
expressed
as a proportionof total length) was about 5.1
(basee or 0.277, if distanceis expressed
in mm). In man,a
simpleexponentialfitted to the human data above500 Hz
wouldhaveaboutthe samevalueof 5.1; 5.1 convertsto 2.2
(base 10).
Thus an almost-exponentialfunction, Function (1),

canalsoeasilybe superimposed
on the chinchilladata, with
the slopeconstanta scaledfrom man or cat. Its value remains2.1 if distanceis expressed
asa proportionof cochlear
length or is about 0.114 (0.263 base e), if distanceis expressedin mm. The upperfrequencylimit (20.5 Hz) given
by the exponentialfunctionof Eldredgeet al. ismaintainedif
2597
J. Acoust.$oc. Am., Vol. 87, No. 6, June 1990
IV. SUMMARY
COORDINATES
OF GUINEA
PIG FREQUENCY-POSITION
A. Frequency-position data
The two functionspublishedin relation to the early

guineapig data (Greenwood, 196lb) wereboth of the form
of Function (1), although B6k6sy'sdata curve, showing
somereversedcurvaturenear the apex, was not. One function took into accountonly B6k6sy'sdata from dead cochleas and extrapolatedthe straight upper end of the curve.
The other attemptedto fit B6k6sy'sdata aswell aspossible,
whilefittingalsothe approximately50-kHz upperfrequency
limit reportedfrom living cochleasby Pestalozzaand Davis
(1956). The slopeconstantsof the two functionsbracketed
the value of 0.1135 that would be obtainedby scalingfrom
the valueapplicableto man.
Later, Kohl16ffel( 1971) provideda precisedetermination of the point of maximum CM responseto tonesfrom
13.5 to 14.5 kHz and republishedcorrelationsof cochlear
damagewith exposuretone frequencyby Smith and Wever
(1949) and by Neubert and Wiistenfeld (1955), all of which
werein betteragreementwith the functionyieldingthe higher frequencylimit. Kohl16ffel's( 1972a,b) laserstudyof the
basilarmembranein deadguineapigpreparationsandtwo in
vivopreparations(Kohl16ffel,1972c), alsosupportedboth
the approximatecorrectness
of a slopescaledfrom the value
for man and the higherupperfrequencylimit. All theseresultsoverall are in still closeragreementwith the function
discussed
belowand appearwith it in Fig. 4.
Wilson and Johnstone ( 1972, 1975) summarized their
own data from the basal 4 mm of the cochlea and those of
others, including Kohl16ffel's(1972c) in vivo determinaDonaldD. Greenwood:Cochlearfrequency-position

function
2597
and the value of about 0.35 for the constant A. This function
0.1
yieldsan upper frequencylimit of about 43.8 kHz, and

agreesequallycloselywith mostof the data.
X
The determinationsof inner hair-cell (IHC) CFs by

Russelland Sellick ( 1978, 1988,personalcommunication),
at positions
fromabout1to 4.5 mm fromthebase,arealsoin
goodagreementwith the Wilson and Johnstone
data and
Function ( 1). Immediatelyon the low-frequencysideof the
main concentrationof IHC points,of which six are on the
ElO0
line, Kohl16ffel's measurementsof the 14- and 15-kHz
(J
pointsfromhisCM recordings
alsofall onthefunction.Two
points representpeak frequenciesof responsefunctions
basedon cochlearmicrophonicdata of Schmiedtand Zwislocki (1977). An additionalgroupof pointswaspublished
by RobertsonandManley ( 1974);the CF of a cellrecorded
in the spiralganglionwaspairedwith the cochlearpoint at
the endof a line radiallyprojectedabout700p to the cochlear partition.Thesepointsare nearlyparalleland slightly
5_
o
o. 1
FIG. 4. Frequencyversuscochlearpositionin the guineapig. Inset graph:

basalto the function. Robertsonet al. (1980) later plotted
Data as plottedby Kohll/Sffel(1971). Solid square:Kohll/Sffel'spoint of
the CFs versus basilar location of 102 neurons between
maximumresponseto 13.5- to 14.5-kHz tones,basedon CM responses
recordedby a 12-electrode
array ( 150-/zminterelectrodespacing).Solidcirabout 1.5 and 5 mm from the base.They agreequite well
cle represents
inferredlocationof 15 kHz maximumbasedon a 0.22-mm
with Wilson and Johnstone's(1975) simple exponential;
basalshift of CM minimum. Short verticalline representsdamageto the
about 62% of the pointsare apical to the function.The
cochleawith exposureto a 10-kHz tone (Smith and Wever, 1949). Long
verticallineindicatesregionof swollenhair cellsafterlongexposureto a 15pointsagreeaboutequallywellwiththeFunction( 1) in Fig.
kHz tone(NeubertandWiistenfeld,1955). Opensquare:Kohll/Sffel's
point
4; about58% of the pointsare basalto the function.Quite
of maximum mechanicalresponse(in vivo) to 28 kHz, repeatedin main
consistentwith functionslopeis the reportof Johnstoneand
graph.Main graph:Solidpointsrepresentmechanicaland inner hair-cell
data. Open points are basedon cochlearmicrophonicmeasurements. Taylor (1970) of a shift in peakfrequencyfrom 19 to 16.5
Crosses(+ ) representthe CFs of primaryneurons.Solidsquare:Kohl16fkHz, with a 0.5-mm shift of the Mfssbauer sourcefrom 1.5
fel ( 1972c)--28-kHz peakfrequencyof mechanical
response
at point 1.5to
to 2 mm from the base.Theselocationswould be displaced
1.7 mm from basein two in vivopreparations.Solidtriangles:Wilson and
basally
from the calculatedcurveby about1.67mm, but the
Johnstone( 1975)m13 cutofffrequenciesof mechanicalresponsecurves;
calculatedseparationof the 19- and 16.5-kHzpointsis 0.53
five symbolson or touchingline, with four below and four above.Solid
circles:
Russell
andSellick( 1978,1988tableofdatapointsmpersonal
communication)mCFsof 14 individualinnerhair cells;sixof thesepointsare
virtually on the line and, in the cluster, at about 15 mm, there are seven
points,fiveof whichareamongthe sixon the line.Openinvertedtriangels:
Schmiedtand Zwislocki ( 1978)--peak frequenciesof responsefunctions
basedon cochlearmicrophonicdata. Opencircles:Dallos' cochlearmicrophonicdata asplottedby Wilson and Johnstone(1975). Crosses:Robertsonand Manley (1974)m13 CFs of spiralganglioncellsversusposition,
pointsthat arenearlyparallelandabout0.5 mm abovethe line;onepointis
undera circle.Wilson'sand Johnstone's
peakfrequencies
alsofollow the
slopewell, displacedon averagesomewhatto lower frequencies[Wilson
(1972) ]. B6k6sy'sguineapigcurveisnearlyparallelto thelinefromabout5
to 12 mm from apexand displacedabout 1.9mm basally(upward) on the
ordinate. The curve is Function (1),
where ,4 =0.35,
a= 2.1/18.5,
k =0.85.
tionsof the 28-kHz pointand estimatestheybasedon Dallos' CM measurementsat a numberof loci. They fitted these
dataon a log-linearplot with a straightlinewhoseslopewas

0.1204commonlogunitsper mm (2.5 mm/oct) andwhose
basalinterceptwas45 kHz (Wilson and Johnstone,1975).
Thisslopeconstantconvertsto 2.22 (base10), or 5.13 (base
e) (if distance
isexpressed
asa proportionofbasilarlength),
closeto the valuesfor man and chinchillaif simpleexponen-
tials are used.Their own peak frequencies

alsofollow the
sameslopewell, somewhatapicallydisplacedon averageto
lower frequencypoints(Wilson and Johnstone,1972).
If, instead,an equationof the form of Eq. (1) is used,
thesedatacanbe quitewell fittedwith the scaledslopeconstant of 2.1/18.5 = 0.1135 (2.1 for proportionaldistance)
2598
mm, in goodagreementwith the shiftof their source.

Thusthevariousbodiesof dataplottedin Fig. 4, despite
any residualuncertainties,seemto be reasonablyfit by the
functionsuperimposed
on them, which possesses
the same
normalizedslopeconstantusedfor the precedingthreespecies.It isalsoverycloseto theaverageslopeof thebasaltwothirdsof B6k6sy'sdata,whichareshiftedbasally,awayfrom
the in vivo function, in the way describedby Kohllfffel
(1972b).2
B. Shape of the functionsdegree
of curvature
It may well be, of course,that the degreeof curvature

neartheapexmaydifferfromonespecies
to anotherdependingontheirevolutionary
specializations,
notto mentionthat
the form of the requiredfunctionmay differ somewhatas
well. Sincethe functionfitted here to the data is empirical,
thereis nothingto arguethat the samevalueof k is suitable
for everyspecies,
evenamongthosefor whichthe general
form may providean adequatefit to the data.
However,asto the suggestion
or possibilitythat a simple exponential,that is, a straightline on a log-frequency
versuspositionplot, may be adequatefor a givenspecies
everywherein the cochlea,a secondconsideration
mustbe
raised. If evidence exists that very-low-frequencytones
causemostof the cochleaof the givenspecies,
includingthe
apicalregion,to vibratealmostin phaseand if it is known

(viz., Andersonet al., 1971) or theoreticallypredictedthat
phasetravel time to a point is an exponentialfunctionof
Donald D. Greenwood'Cochlear frequency-positionfunction
2598
monkey might be about 2.1/23 = 0.091 for a and 0.370 for

A, yieldingan upperfrequencylimit of 46 kHz.
Estimatesof frequencyof hearinglossversusthe position of IHC lossfor four M. Nernestrinamonkeyshavebeen
reportedby Stebbinsand Moody (1979). Thesefour monkeyshad an averagelengthof cochleaof 25.6 mm (Stebbins
and Moody, 1988, respectivepersonalcommunications).
The cutofffrequenciesof hearinglosswere correlatedwith
the position of 50% IHC lossto the nearest0.5 mm and
suppliedto me with the individualcochlearlengths.These
distancetraveledto that point,thenfrequencycannotalsobe

laid out exponentiallyin the apical region (Greenwood,
1977). To make this more intuitively clear, considerthat
phaseat a pointissimplytheratioof phasetraveltimeto that
pointoverthe tone'speriod.If we knowhow two of these
variableschange,we know how the third changes.Flattening of the spatialphasecurveoccurs(e.g.,in man,cat, and
squirrelmonkey)because
phasetraveltimeat thepositionof
maximum amplitude, for example,obeysan exponential
functionof distanceoverthe apical three-fourthsof the partition, whereasthe frequency-position
functiondoesnot decreaseexponentiallyat progressivelymore apical points-where octavesbecome crowded. Hence, the period (frequency'sreciprocal)of the tonereachingmaximumat those
progressively
moreapicalpointsincreases
morethanexponentially, in the denominatorof our ratio. In effect, a lowfrequencytonedoesnot travelto pointsfar enoughdownthe
cochleafor its period (ratio's denominator) to constitutea
smallor constantproportionof the tone'sphasetravel times
to apicalpoints (ratio's numerator). Rather, the reverseis
true, leadingto smallervaluesof accumulatedphase (i.e.,
flattenedspatialcurves)for low-frequencytones.
Thus,in the guineapig, if spatialphasecurvesbecome
flatterin theapicalregionthan elsewhere,
whilethe relation
of phasetravel time to positionsof maximumamplitudein
that regionis knownor believedto be exponentiallyrelated
to cochlearposition,then frequencyin the guineapig will
not be laid out logarithmicallyover the whole cochleabut
will resembleinsteadthe patternseenin man, elephant,and
cat.The samereasoningmay be appliedto the chinchilla,in
the regionbelow500 Hz whereFig. 3 lacksdata,andto the
gerbil,where data are scant.
lossis comparableto the displacementof Liberman'sprimary-fiberCF data (in Fig. 2) from the hearinglossversus
placecorrelationscitedby Liberman (1982). As in the case
of the cathearing-loss
data,it is reasonable
to expectthat, as
a tone'sfrequencyis raisedand progressively
shiftsthe displacementmaximumtowardthe regionof missingouterand
v. MONKEY: FREQUENCY OF HEARING LOSS VERSUS

POSITION OF 50% IHC LOSS
four positionsare replottedin Fig. 5, afterfirstnormalizing

them with respectto individualcochlearlengthsand then
convertingthem to positionsrelativeto the mean cochlear
length among thesefour animals (0.63 mm lessthan the
meanof 52 monkeysof the samespecies).
Function (1), using the constants a = 0.082 and
A = 0.36 (to yieldkHz), calculates
a curvelyingon average
1.25 mm apical (lower on the ordinate) to the four data
pointsplottedasopensquares.
Whenthefourpointsarealso
plottedassolidcircles1.25mm moreapically,they demonstratea goodagreementin slopeand alsosimplyillustrate
the suggestion
herethat the frequency-position
functionfor
thesemonkeysprobablylies apical to the positionsof the
squares,to proceedto the 45-kHz upperfrequencylimit indicatedby Stebbins'other data cited above.The displacement of the curve 1.25 mm from the locations of 50% IHC
0.1
25
Data have become available on a number of monkey
species.
Stebbins
(1970) andStebbins
etal. ( 1973) reportan
upperfrequency
limit of about45kHz amongthemacaques.
Stebbins
andMoody (respectivepersonalcommunications,
1986,1988) reportmeancochlearlengthsfor sevenspecies
ofmacaques
rangingfrom23.05-26.26mm.If theslopeconstanta scalesfrom manto thesespecies,
thisfrequencylimit
and thesebasilarlengthswould suggesttheseconstants:
A = 0.36 for all and a = 0.09 to 0.08, respectively.
Beecher(1974a,b) has reportedthe upper frequency
limit of the squirreland owl monkeysto be about46 kHz.

For the squirrelmonkey,Igarashiet al. (1968) report a
cochlearpartitionlengthof 20 mm, but alsoreporta 22-mm
lengthfor thecat,andan 8-mmlengthfor therat. However,
, ,,ll
il i i i i [
10
, i i i i I
20C
0
oo
L15
E
10
o 5
._
0,
0.1
Frectuency
10
in
K I ohertz
constants
for thefrequency-position
functionfor thesquirrel
FIG. 5. Relation of cutoff frequencyof hearinglossin M. Nemestrinato

placeof 50% lossof innerhair cells(StebbinsandMoody, 1979,1986and
1988, personalcommunications).Open squares:cochlearlocationsexpressed
asproportions
of cochlearlengthin eachmonkey,beforeexpression
with respectto themeancochlearlength(25.6 mm) in thesefour monkeys.
Opencircles:samepointsshifteddownontheordinateby 1.25mm.Curveis
Function ( 1), whereA = 0.36, a = 2.1/25.6, k = 0.85. ConstantA is setto
yieldupperfrequencylimit of 45 kHz (Stebbins,1970). Shiftedcirclesare
to illustratea fairly closeagreementof data slopeto the function.Original
squaresshowa displacement
of correlations
of frequencylossversusplace
of 50% IHC lossfrom a functionthat yieldsthe upperfrequencylimit indicatedby behavioraldata and that might representpositionof maximum
cochlear-displacement
amplitudeversusfrequency(seetext).
2599
these estimatesuse the method of reconstructionused by

Guild ( 1921) and Schuknecht(1953), which underestimate
membranelength.As notedearlier,Libermanfound the

lengthfor the catto be25 mm whenthissourceof errorwas
avoided,andB6k6syreporteda lengthfor the rat of 9.7 mm.
Ifcochlearpartitionlengthfor thesquirrelmonkeyis underestimated
by a percentage
errorcomparable
to the errorfor
the cat,thenbasilarmembranelengthin thismonkeymaybe
almost23 mm, ratherthan 20, mm. If it were,the applicable
2599
inner hair cells,the frequencyat which lossbeginswill be

lower than the normal CF for the pointsat which damage
begins.
!oo
:: 12
.... I
looo
........
loooo
........
, ,
,_
Vl. FREQUENCY-POSITION
DATA FROM OTHER
SPECIES
For otherspecies
for whichwe possess
frequency-positiondata,thedata (elephant,cow,rat, andmouse)areolder
(B6k6sy,1960) and were consideredearlier (Greenwood,
1961b) or arenewbut sparse--Mongolian
gerbil(Sokolich
et al., 1976). Perhapsotherdataexistthat shouldbetreated
here,but this surveyis not intendedto be exhaustive,and
only thesespecieswill be considered.
To reviewtheolddata,elephantandcowwerequitewell
fit in 1961by functionsemployingscaledslopeconstants,a
equalto 0.035and0.055,respectively,
or 2.1 whenbasilar
distanceis normalizedasa proportion.B6k6sy'smouseand
rat datawerethensatisfactorily
fit by functionswith exactly
scaledslopes.However,thelatterdataallowedof considerable latitude in both main constants,chiefly constantA,
whichfor a givenslopeconstantwasempiricallydetermined
by approximate
conformance
of thecurveto B6k6sy's
data.
The datadid not permitverysecureextrapolationto an up-
O 6
O 2
._
r-I 0
:1O0
1000
Frequency
! 0000
in
Hertz
FIG. 6. Data points:frequencyversuspositionin theMongoliangerbil(Sokolichet al., 1976), basedon frequencyof maximumCM versuselectrode
position.
Meanlengthof gerbilcochleaisreportedas12.1mm.Solidcurves:
,4---0.400 to yield a 50-kHz upper frequencylimit; a= 2.1 (or 2.1/
12.1= 0.174 to scalex in mm: k = 0.85 or 0.35 to bringfunctioninto better
agreement
with the mostapicalpoint (but seeRyan andBone,1978).
perfrequency
limit.Pairsoffunctions
seeming
to delimitthe
considpermissible
frequency
limitsandto bracketthescale-related closelywith the existingdataasin the othercases
ered,althoughtheshortlengthof membrane
represented
by
slopevalueillustratedthesepoints.
Now, morerecentestimatesof theupperfrequencylimits in mouseof 120 kHz (Ehret, 1975) and in rat of 80 kHz
the ordinatemakesa millimeterdeviation(of a point from
0.3 for the mouseand 0.216 for the rat. The resultingfunc-
theline) in the gerbilappearabouttwiceaslargeasin the

cat.However,thepointof thiscomparison
is not that three
datapointsdictatea curveof theformof Function( 1), but
thatthepointsarereasonably
consistent
with it. The figure
tions are more securelydeterminedand remain in good
alsois intendedto illustratethat, for an increasinglyimpor-
agreement
with B6k6sy's
datacurvesasregardsslope.B6k6sy'smousedatacurveis nearlyparallelbut displacedba-
tant laboratoryspecies,more frequency-position

data are
(Kelly and Masterson,1977) yieldA constants

of about
0.960and0.640,respectively,
givenscaledslopeconstants
of
salward,consistentwith Kohl16ffel's(1972a,b) findings,by

about1.4mm. B6k6sy's
rat datacloselyfit thenewfunction,
coinciding
at 200and5000Hz, andarenevermorethan0.5
mm from it.
A tentativefrequency-position
functionfor theMongolian gerbilcanbecompared
with onlythreepairsof empirical frequency-position
coordinates
(Sokolichet al., 1976).
In thesedata, the maximumCM is plottedversuselectrode
position.Two of the threeCM peaks,at 0.5 and2 kHz, are
rather clearlyindicatedand would seemto warrant greatest
weightin fitting,but the mostapicalpointof thesetwo is
basedononlya singlegerbil.Theaveragelengthof thegerbil
cochleais reportedby Sokolichet al. as 12.1 mm, which
would indicate a scaled constant a of about 0.174. This con-
stantandanA constantof 0.4 yieldan upperfrequencylimit

of about50 khz, whichis in fairly reasonable
agreementwith
their plot of their data.To bringthisfunctioninto a better
agreement
with themostapicalpointa subtractive
constant
needed.
VII. RECENT
DEVELOPMENT:
A PSYCHOACOUSTIC
FREQUENCY-POSITION
FUNCTION IN EQUIVALENT
RECTANGULAR BANDWIDTH (ERB) UNITS FOR

HUMANS
A recentfrequency-position
functionfor man hasbeen
developedby Moore and Glasberg (1983) and Moore
(1986), alsoby integratinga critical-band(ERB) function,
the samemeansemployedby Greenwoodin 1961,but based
on morerecentbodiesof data obtainedby Houtgast (1977),
Patterson (1976), Weber (1977), Fidell et al. (1983), Patterson et al. (1982), and Shailer and Moore (1983). The
ERB valuesbasedon thosedata are similar in size (slightly

smaller) and slope, over the fitted part of the frequency
range,to critical bandwidthsas expressedby Greenwood's
( 1961b, 1974) critical-band function.
The accuracyof the slightlyvariant gerbil functions

shownin Fig. 6, if more frequency-position
data were ob-
Moore and Glasberg (1983) assumed,in effect, that

their ERBs followeda second-order
polynomialfunctionof
frequency.They fittedthe functionto data overthe frequency rangefrom 125 Hz-6.5 kHz, integrated,and obtaineda
frequency-position(ERB-rate) function, when ERB units
were assumedto correspondto a constantdistancein mm,
specificallyabout0.9 mm. Greenwood's1961frequency-po-
tained, is, of course,uncertain,and they seemnot to agreeas
sition function had indicated to them that, in the above fre-
2600
Donald D. Greenwood:Cochlearfrequency-position
function
of about 0.35 rather than 0.85 is needed. However, the data
of Ryan and Bone (1978) suggestthat 500 Hz may be associatedwith a point2 mm (or slightlymore) from the apex,
which is more consistent with k = 0.85 or 1.
2600
quencyregion, the estimatedERB valuescorresponded

closelyto 0.9 mm (Moore, personalcommunciation,1983).
The ERB-ratefunctionobtainedby integratingwasquantitativelysimiliarin a part of the fitted frequencyregionto
Greenwood's1961 frequency-position
function in Fig. 1.
Given that Function (1) reasonablyapproximates
the Bksy-Skarsteincochlearmap, it may be usefulto consider
brieflytherelationof thebandwidthestimates
aboveto equal
distancesaccordingto Function (1) and to comparethe
ERB-ratefunctionto the cochlearmapspresented
in earlier
lOO
i
lOOO
J i i i i ,I
1oo
1ooo
! ! i , ,[
, ! , iii
lOOOO
i
i i
[ ,
10000
ooo
I 1000
._
lOO
figures.
The bandwidth estimatesof Houtgastand Patterson

hadcloselyfollowedtheslopeof the 1961critical-bandcurve
and the two-toneconsonance
data originallycomparedto
the latter.Hence,theyalsocorrespond
closelyto equalbasilar distances
calculatedby Function(1), as seenin Fig. 7,
wherethe data fall closeto superimposed
curvesrepresentingfrequencyintervalscorresponding
to particularconstant
distances.
In addition,the da(aof ShailerandMoore (1983)
andFidellet al. (1983), citedabove,agreerathercloselyto a
constantdistanceof 0.9 mm, in the maingraphof Fig. 8. It
wasunclearhow Moore and Glasberg(1983, their Fig. 1)
hadplottedWeber'sdata,whichwereobtainedat fivespectrum levelsfrom 10 to 50 dB SPL, and variednoticeablyat
the two highestlevels.Therefore,all Weber'sdataare plottedin theinsetgraphof Fig. 8, and,at the threelowestlevels,
thefrequency
intervalscorrespond
quitecloselyto a distance
of 0.53 mm.
loo
,
, t i I iJ
lOOO
i
, 'Nil]
10000
I
i t till
.... !,,o,o
..... .Lo,oo
.... !.o.0,oo
.
N
I 1000
._
lO
,
1O0
, , ,
i 0[00
FrecLuency
in
10000
Hz
FIG. 8. Main graph:The circlesare the data of Shailerand Moore (1983).

the squaresare the data ofFidell et al. (1983). Both setsof bandwidthestimatesareobtainedfrom temporalgapdetectiondata.The curverepresents
frequencyintervalscorresponding
to 0.9 mm. Insetgraph:data of Weber
(1977), analogous
to Patterson'sin Fig. 7. Obtainedat fivespectrumlevels
from 10to 50dBSPL.Thecurverepresents
frequency
intervalscorresponding to a distanceof 0.53 mm. The uppermostcirclesrepresentresultsat the
50-dBspectrumlevel,andthe0.9-mmcurve(not shown)passes
just above
thosecirclesat 1000and4000Hz, by nearlythesameamount.The squares
indicatedataobtainedat the 40-dBspectrumlevel.The remainingdataat
the 30-, 20-, and 10-dBlevelsbunchup, with thebrokencrossinglinesindicatinglittle or no systematiceffectof level.As in Fig. 7, the curveis not
fitted to the data, but rather is proportionalto the derivativeof Function
( 1), as in Greenwood (1974).
Note that the calculatedcurvesin Figs. 7 and 8 are used

as "measuringsticks" to assessthe bandwidths'conformanceor nonconformance
to constantdistancesasgivenby
Function( 1) andthe cochlearmapin Fig. 1. The curvesare
neither fitted to the bandwidth data nor presentedto draw
supportfrom them, sincethey have an independentstatus
providedby the cochleardata. However,beyondindicating
the bandwidths' relation to cochlear distance, the calculated
curvesserveasmore than adequatedescriptivefunctionsfor
these data.
lOO
lo
, , 1,1
'
1000
Frequency
' ' ''1
10000
in
Hz
FIG. 7. Estimatesof auditory-filterbandwidths.Main graph: Houtgast

( 1977); upper set of estimatesare of Gaussianbandwidthsderived from
ripple-resolution
dataobtainedin simultaneous
maskingexperiments.
The
uppercalculatedcurverepresents
the frequencyintervalcorresponding
to
1.1 mm, accordingto Function ( 1). Lower curveand points:estimatesof
filterbandwidths
derivedfrom ripple-resolution
dataobtainedin pulsation
thresholdmeasurements.
The calculatedcurverepresents
thefrequencyinterval corresponding
to 0.65 mm asabove.Inset graph:Patterson( 1976);
uppersetof pointsare estimatesof Gaussianfilter bandwidthsderivedfrom
notched-noise
simultaneous
maskingexperiments.
The Pattersonpointsat
500 and 2000 Hz nearly coincidewith Houtgast'supper set of pointsat
thosefrequencies.
The uppercalculatedcurverepresents
the frequencyintervalcorresponding
to 1.16mm, asabove.Lower curveand points:Estimatesof equivalentrectangularfilter bandwidthsderivedfrom samedata.
The calculatedcurverepresentsthe frequencyinterval corresponding
to
As for the derivedERB-rate (frequency-position)function, Moore (1986, his Fig. 5) has comparedit directly to
someof B6k6sy'sand to Skarstein'sdata (Kringlebotnet al.,
1979). Moore statesthat thebestcorrespondence
is obtained
if eachERB bandwidthcorrespondsto about0.89 mm and
readjustshis constantsaccordinglyfrom those originally
published.However, his Fig. 5 restrictsitself to the range
from 400-6500 Hz by omittingfour of B6k6sy'seightpoints
from the Kringlebotnfigure,all of thosebelow400 Hz, two
of which are at and above100 Hz in the samerangeof frequenciesoverwhichthe ERB curvewasfittedto the empirical bandwidth
estimates.
0.89 mm, as above.
However, the comparisonof the ERB-rate function to

basilar coordinatesis as close as his figure indicatesonly
above 400 Hz, in the selectedregion and in the largely
straightsectionof the function,just prior to its increasing
convexity.In Fig. 9 of this paper,Moore's functionis comparedoverthe wholehumanfrequencyrangeto the omitted
Bk6sypointsand to Function ( 1) of Fig. 1. The ERB-rate
functionexhibitsa lesseragreementwith B6k6sy'sdata in
2601
DonaldD. Greenwood:Cochlearfrequency-position
function
2601
loo
lOOO
10000
35
c30
.--
--
x25 --
20-
L15-
/,.?
o10
C
m-, 5Itl
'
'
'
' ' ''1
'
'
'
lOO
'
' ' ''1
'
'
lOOO
Freckoency
in
' ' ''!
lOOOO
Hz
FIG. 9. The solidcurveis Function ( 1) from Fig. 1.The dashedcurveis the

ERB-rate functionof Moore and Glasberg(1983) and Moore (1986), as
modifiedby Moore for direct comparisonwith B6k6sy's(1960) and Skarstein's(Kringlebotnet al., 1979) data, which relatepositionof maximum
amplitudeto frequency.
the lower frequencies.A lesseragreementwould also be

shown by Zwicker's critical-band-rate function (Zwicker
andTerhardt, 1980) if it werecomparedto thesephysiological data. The comparisonsalsoshowthat the curvaturebeyond the high-frequencyend (6.5 kHz) of the rangeof the
originalpsychoacoustic
dataleads,not surprisingly,to higher than usualestimatesof the upperfrequencylimit of hearing and to an unrealisticform, if we can generalizefrom
other species.
In severalspecies,
the evidencesincethe 1940sindicates
In all of the comparisons

of morerecentdata,it hasbeen
possibleto scalethe slopeconstanta exactlyfrom the original slopeconstantof 0.06 found to be suitablefor man, by
multiplyingit by the ratio of basilarlengths(man'sdivided
by that of the other species).That is, the productof a times
basilarlength is approximatelyconstantamongthesespecies,specifically2.1; thus,if distancealongthe membraneis
expressed
asproportionallength,from 0 to 1,a canequal2.1
in all thesecases(if physicaldistanceis required,divide2.1
by basilarlength). Obviously,it isnot shownnor certainthat
the slopeparameteris completelyconstant.However, it is
clearenoughthat it isverysimilaramongthesespecies
and is
not likely to vary much amongthem if more definitivedata
are obtained.Apparentlythesecochleasare sufficientlygeneralizedfor this relationto hold, althoughthereseemsto be
little reasonto expectsucha convenientrelationand function to apply unmodifiedto more specializedcochleas,such
as thoseof certain bats with an enlargedpatch of cochlear
partitiondevotedto a particuIarfrequencyor thoseof a burrowing rodent like the "mountain beaver" (Aplodontia
Rufa) that is reportedto be specializedfor low frequencies
(Merzenich et al., 1973).
Hence,the functionshownin 1961to fit quiteaccurately the data then availablefrom eightspecieshasbeenshown
hereto fit closelyconsiderable
additionaldata from someof
the samespecies--human,cat, andguineapig--as well from
the chinchillaand a speciesof macaque(at leastin respectto
slopeand upper frequencylimit), on which data had not
previouslybeenavailable.Most of theseadditionaldata are
fromlivingspecimens.
The basicfunctioniscertainlya plausiblecandidateto fit both gerbil data and data from other
macaqueandsquirrelmonkeys,if moredatafrom thesespecies become available.
that, over most of the basilar membrane, and most of the
frequencyrange, log frequencyversusbasilar positionis

nearly a straightline. Thus the almost-exponential
form of
the 1961 function, and the constancyof normalized slope
amongseveralspecies,indicatesthat it more plausiblyapproximatesthe form of the functionthat shouldbe theoretically derived. Zwislocki (1965) has long sincederived an
almost-exponential
function;someothermodelsandempirical curveshave usedsimple exponentials.For the species
consideredhere,the 1961almost-exponential
frequency-positionfunction seemsat this moment as satisfactoryin form
as a descriptionof physiologicaldata as it did 29 yearsago.
SUMMARY
AND DISCUSSION
A. Present status of the cochlear frequency-position

function
Sincepossible
cochlearfrequency-position
functionsare
chieflydependentultimatelyon the accuracyof the available
physiologicalfrequency-position
data, it hasbeenunfortunate that those data were initially, with the exceptionof
those from the cat, only from dead specimens.Partly
counter-weighting
that fact were the observations
that the
slopeconstantcouldbe scaledor normalizedacrossspecies,
importantlyincludingthe live cat, and that, in man, the upper frequencylimit was consistentwith behavioralestimates.
2602
B. Desirability of further physiological support for

function (1) for homo sapiens in particular
Sincefor humanbeingsour cochleardata mustremain

indirect,onereasonfor surveyingthe interspecies
comparisonsin thispaperhasbeenfor theirbearingonthedegreeof
confidencethat may, or may not, bejustifiedin the possibility that the frequency-versus-position
data from humancadaversapplyalsoto livingears.Sincethe samesimpleform
of functionhasbeenshownto fit reasonablywell the data of
up to tenspecies,
utilizinga singleconstanta for normalized
slope,the useof thisfrequency-position
functionfor the livinghumanearseemsto bereinforced.The caseof thecat,for
which data are mostcomplete,seemssihglyto be the most
supportiveat this time (Liberman,1982), with the guinea
pig a closesecondsincedataof basicallyfour kindssupport
the sameslope.Whateveruncertainities
remain,at leastthe
comparisons
havebeenbroughtup to date.
But it would be valuable,for the mostnearly direct relevanceto homosapiens,to obtaindetailedfrequency-place
correlationsfrom primates.The squirrelmonkeyandM. Nemestrina are reasonable candidates. In the first instance, we
already have the mechanicaldata of Rhode (1971, 1978).

Study of the squirrelmonkey could permit any necessary
revisionof cochlearlength,as well as of estimatesmadeof

the cochlearloci of Rhode'srecordingsitesand the longituDonaldD. Greenwood:Cochlearfrequency-position
function
2602
dinal extentof the apicalsegmentand peakregionof the

displacementenvelope(Greenwood,1974). In the second
instance,wehavethedataof $tebbinsandMoody (1979) in
Fig. 5, which would affordthe opportunityto clarify the
relationof basalheatinglossdata to, still nonexistent,data
from the monkeythat would relate primary fiber CF to
cochlearlocationusingthe horseradish
peroxidase(HRP)
techniques
employedby Liberman(1982).
sapiens,and/or to confirm its form. However, hesitation

about plotting data on physiologicallysupportedcochlear
frequencyscaleswhilewaitingfor furthersupportseemsunnecessarily
cautious,for man and a numberof otherspecies
considered.The easeof plotting computer-storeddata on
anytransformedscaleaswell asontheoriginalindependentvariablescalearguesfor the useofcochlearcoordinatescales
wheneverit may proveinteresting.Moreover,wherethereis
willingness
to plot on purely psychoacoustically
generated
C. Relation of the cochlear frequency-position function
scales,
there
should
be
little
reluctance
to
use
physiologically
to psychoacoustic data
supportedscales.
Although the frequency-position
functionconsidered
Given the increasinginterestin speechrecognitionalherewasobtainedoriginallybyintegrationof an exponential
gorithmsand hypotheticalneuralnetworksto processaudifunction fitted to some of the critical-band estimates availtory-nerveinput, it may be especiallyto the pointto plot the
ablein 1961,it wasthencomparedto cochleardatafor the
spectralanalyses
of speechsoundson a physiologically
supkindof confirmation
thatsucha functionrequiresto serveas portedfrequency-position
scalerather than on scalesbased
a cochlearfrequency-position
function, rather than as a
only on psychoacoustic
data, sincethe latter will havebeen
purelypsychoacoustical
construct,howeverusefulthe latter
influenced
by whatever
factorsotherthanspatialmayhave
might be. Psychoacoustic
data cannotbe knownin advance
contributedto the originalmeasurements.
actuallyto reflect,unconfounded
by any other factor,only
To usescalesknown not to conformaccuratelyto existtheseparation
of cochleardisplacement
maxima,or a "spaing physiological
data is, on the onehand,eitheran implicit
tial" factor,howevernamed.That hypothesisin respectto
argumentthat the physiologicaldata are defective(or more
any given body of data requiresassessment
and may not
likely to be defectivethan the psychoacoustic
data and asapply, without any fault in thosedata. For example,more
sumptions
usedto arriveat the scale)or, on the otherhand,
than oneoperativefactormay co-varywith changein a sinan argumentthat the psychoacoustic
scalereflectssomeesgle independentvariable,producingpotentiallydifferential
eontlal "norportrayal"
nr intogrativenqvt-hnat-nllqtitequiveffectsin differentperformancetasks.Althoughthe validity
alence(by presumablylumping the effectsof all operative
of any testof correspondence
of frequency-resolution
estifactors) that the user explicitly wants to incorporateas a
matesto constantdistanceson an independentphysiological
desiredpreprocessing
expectedto further his psychophysscalewill dependon theaccuracyof the physiological
data,if
ical ends.The first argumentthat physiologicaldata are the
the latter aresoundandif somepsychoacoustic
measuresdo
weakestlink is becomingincreasinglylessplausibleastechcorrespondto equaldistancesand othersdo not, the latter
niquesimproveand as similaritiesamongsomespeciesbemeasuresmay ipsofactobe especiallyinterestingand useful
come clearer. Is it likely that a psychoacoustically
based
to studyto determinewhat otherfactorsactuallyoperatein
cochlearscalefor the catwouldnowbeproposedif the requithe variousperformancetasksin question.Thispoint,made
site data for the cat existed but led to visible conflict with
in 1961,still seemsnoninvidiousand unexceptionable.
Liberman'sHRP data and the descriptionby Function ( 1)
Thus, if the future yields accurate cochlear data that
of those data?
continueto reinforcethe frequency-position
function, the
An argumentof the secondtype for usinga frequencysupportprovidedby the conformance
of a givensetof psyposition
scalethat is known not to agreewith existingphychoacoustic bandwidths to a constant cochlear distance will
siologicalspecifications
of cochlearfrequencycoordinates
turn out to be for the equaldistancehypothesis,
asappliedto
shouldbe madeexplicit.However,whateverthe scale'sapthosedata,rather than for the frequency-position
function.
propriatenessfor its own ends,in modelingauditory proAt thistime,the supportmaybe regardedasto someextent
cesses,
it wouldnot seema genuinebenefitthat lumpedfacfor both only if, and to the extentthat (a) the physiological
tors influencingreal post-cochlearauditoryprocessing
will
data may be regardedas doubtfulon their own, and/or (b)
be already embeddedin the assumedspectraltransformathe equal distancehypothesismay be independentlysuptionsbasedon psychoacoustic
scalesdifferingfrom physioportable,deductivelyor evidentially.Moreover,othersetsof
logicaldata. It seemsarguablymore conservativeand flexipsychophysical
estimatesor measuresof frequencyresolubleto usea cochlearfrequencyscalethat is,sofar aspossible,
tion may not conformto equaldistances,withoutany necesphysiologicallyfoundedand to incorporateexplicitly into
saryadversebearingon the accuracyof the frequency-posilater analysisof the spatiallytransformedspectrumany subtion functionor on their own repeatabilityand validity as
sequentreal or hypotheticalphysiologicalor other processdata.Giventhe performancetaskin questionandthe nature
ing desired--basedon inferencesfrom eitherphysiologyor
of the system,the equaldistancehypothesis
may simplynot
psychoacoustics.
Moreover, differingcochlearscales,based
be correctin a givencase.
on differingsetsof psychoacoustic
discriminativeend-prodD. Uses and advantages of physiologically accurate
uct data, cannot all reflect actual cochlear coordinates with
basilar-frequencytransformationsof data and complex
equalindependence
from otherfactorsnor the optimalspecspectra
tral transform.
Furtherin oiooprimatedatamightpermitusto inferany

advisablemodificationsof Function ( 1) constantsfor homo
2603
J. Acoust.Soc.Am.,Vol.87, No.6, June1990
In somecontrast,the physiologicaldata availablefor

homo sapiens,althoughin vivodata are lackingand unobDonaldD. Greenwood:
Cochlearfrequency-position
function
2603
tainable,havebeenextendedand remain in closeagreement

with Function( 1), whichappearsto havebeenstrengthened
in respectto its slopeconstantby thelatter'sapplicability,in
scaledor normalizedform, to severalother speciesfrom
Anderson,D. J., Rose,J. E., andBrugge,J. (1971). "Temporalpositionof

discharges
in singleauditorynervefiberswithin the cycleof a sine-wave
stimulus:Frequencyandintensityeffects,"J. Acoust.Soc.Am. 49, 1131-
which in vivo data have been obtained. The increased inde-
(Saimiri sciureus)," J. Acoust. Soc. Am. 55, 196-198.

Beecher,M.D. (1974b). "Hearing in the owl monkey(Aotus trivirgatus):
pendenceaffordedFunction (1) by this physiological

support suggests
that a potentiallyimportantuseliesin describing systematicallythe relations of various sets of
psychoacoustically
significant
bandwidths
to cochlearlocation and distance,as done, in part, in Figs. 7 and 8. Constraintson theapplicabilityof theequaldistancehypothesis,
longspareda testagainsttheavailablecochleardata,maybe
expectedin someperformancetasks.Caseby caseanalysis,
givententativeor "working"acceptance
of thecochlearmap
in Fig. 1, might clarify responsible
factorsand other func-
1139.
Beecher,M.D.
(1974a). "Pure-tonethresholdsof the squirrelmonkey
I. Auditory sensitivity,"J. Comp. Physiol.Psychol.86, 898-901.

B6k6sy,G. von (1960). Experimentsin Hearing (McGraw-Hill, New
York).
Bohne,B. A., and Carr, C. D. (1979). "Location of structurallysimilar

areasin chinchillacochleasof differentlengths,"J. Acoust.Soc.Am. 66,
411-414.
Bredberg,G. (1968). "Cochlearstructureandhearingin man," in Hearing

Mechanismsin Vertebrates,
editedby A. V. S. de Reuck and J. Knight
(Churchill, London,England), pp. 126-138.
Ehret, G. (1975). "Masked thresholds,critical ratios, and scalesof the basi-
lar membraneof the housemouse(Mus musculus),"J. Comp. Physiol.

103, 329-341.
tional relations of interest.
Eldredge,D. H., Miller, J. D., and Bohne,B. A. (1981). "A frequencypositionmap for the chinchillacochlea,"J. Acoust.Soc.Am. 69, 1091-
ACKNOWLEDGMENTS
Fidell, S., Horonjeff, R., Teffeteller,S., and Green, D. (1983). "Effective

maskingbandwidthsat low frequencies,"J. Acoust.Soc.Am. 73, 628-
1095.
I would like to thank Dr. J. E. Hind and the faculty of

the Department of Neurophysiologyat the University of
Wisconsin,where this paper was begun in the summer of
1985, and where its ancestor was written in 1960, for their
provisionof a congenialplaceof work andtheirhospitality.I

would
like to thank
those who have commented
on the
manuscriptand especiallyto thank JohnNicol for essential

computerassistance.
Work supportedby NSERC, Canada.
Althoughtheagreement
of theB6k6sy
andSkarstein
datapointsisclose,
satisfactionmight seemto be temperedby recallingthat in Bredberg's
(1968) studyof a largecorpusof humantemporalbones,thelengthof the
organof Corti exhibiteda total variationof about28% of the mean(standarddeviationnotreported).Thisfigurewasquitesimilarto Hardy's 33%
(1938), who alsoreporteda standarddeviationof 6.8%, and very similar
to the total variationof lengthwithin a numberof the macaquespecies,
wherethe total rangeand standarddeviationwerealmost29% and 6% to
7% of the mean,respectively(figuresobtainedfrom datapersonallycommunicatedby Stebbins,1986). In chinchilla,Bohneand Carr (1979) report the rangeand standarddeviationof cochlearlengthsas 26% and almost 5% of the mean, respectively.However, they also report that the
dimensionsof the chinchillacochlea,suchas its width, are very similar at
corresponding
normalizedpoints,whichwould tendto keepconstantintraspeciesparameterssuchas the normalizedslopea in Function ( 1). In
cat, Liberman (1982) findsthat normalizingbasilarlengthreducesacrosscatvariabilityofunit-CF versusplace,i.e.,indicatesalsoa commonparameterA and upperfrequencylimit.
2However,
thereappears
tobea complication
in theWilsonandJohnstone
data, in that the frequenciesthey measuredover the basal4 mm of the
partitionare cutofffrequencies(as definedin their paper) and not peak
frequencies.
They reportthe peakvalueswouldbe about10% loweror up
to 20% lower, dependingupon whetheror not an additionalcorrection
wasalsonecessary
for drainageof the scalatympani.However,it may also
be notedherethat a countervailingcorrectionin the otherdirectionmay
benecessary
ifa progressive
basalpeakshiftwith timeanddeteriorationof
the preparationoccurred,of the type reportedby other experimenters
(Kohll/Sffel, 1972b;Rhode, 1973; Khanna and Leonard, 1982;Sellick et
al., 1982;Robleset al., 1986). Hence,if, owingto experimentaltrauma,
themeasured
cutofffrequencies
aretoolow andshouldbe raised--before
then makinga downwardcorrectionto peakvaluesand for the effectsof
drainagesthenthe opposedcorrectionswouldto someunknownextent
counteracteachother.Althoughestimationof the net resultmay well be
too uncertainto be useful,in the eventitself we observethat the upper
frequencylimit of about 43 to 45 kHz that is consistentwith their cutoff
frequenciesis alsoconsistentwith the other in vivodata.
2604
638.
Fletcher, H. (1940). "Auditory Patterns," Rev. Mod. Phys. 12, 47-65.
Fletcher,H. (1953). Speechand Hearing in Communication(Van Nostrand, New York).
Greenwood,D. D. (1961a). "Auditory maskingand the criticalband," J.

Acoust. Soc. Am. 33, 484-502.
Greenwood,D. D. (1961b). "Critical bandwidthandthe frequencycoordinates of the basilar membrane," 33, 1344-1356.
Greenwood, D. D. (1974). "Critical bandwidth in man and some other
species,"in Sensationand Measurement:Papersin Honorof S. S. Stevens,

editedby A. R. Moskowitz, B. Scharf, and J. C. Stevens(Reidel, Dordrecht,Holland), pp. 231-239.
Greenwood,D. D. (1977). "Empirical travel time functionson the basilar
membrane,"in Psychophysics
and Physiology
of Hearing,editedby E. F.
Evansand J.P. Wilson (Academic,New York), pp. 43-53.
Guild, S. R. (1921). "A graphicreconstructionmethodfor the studyof the
organofCorti," Anat. Rec. 22, 141-157.
Hardy, M. (1938). "The lengthof the organof Corti in man," Am. J. Anat.
62, 291-311.
Houtgast,T. (1977) "Auditory filter characteristicsderivedfrom directmaskingdata and pulsation-threshold

data with a rippled-noisemasker,"
J. Acoust. Soc. Am. 62, 409-4 15.
Igarashi,M., Mahon, R. G., and Konishi, S. (1968). "Comparativemeasurementsof cochlearapparatus,"J. Speech.Hear. Res. 11, 229-235.
Johnstone,B. M., and Taylor, K. (1970). "Mechanicalaspectsof cochlear
function,"in Psychophysics
and Physiology
of Hearing, editedby E. F.
Evansand J.P. Wilson (Academic,New York), pp. 81-93.
Kelly, J. B., and Masterson,B. (1977). "Auditory sensitivityof the albino
rat," J. Comp. Physiol.Psychol.91, 930-936.
Khanna, S. M., and Leonard,D. G. B. (1982). "Basilarmembranetuning
in the cat cochlea," Science 215, 305-306.
Kohll/Sffel,L. U. E. (1971). "Studiesof the distributionof cochlearpotentialsalongthe basilarmembrane,"Acta Otolaryngol.Suppl.288, 1-66.

Kohll/Sffel,L. U. E. (1972a). "A study ofbasilar membranevibrationsI.
Fuzziness-detection:
A new methodfor the analysisof microvibrations
with laserlight," Acustica27, 49-65.
Kohll/Sffel,L. U. E. (1972b). "A study ofbasilar membranevibrationsII.
The vibratoryamplitudeand phasepatternalongthe basilarmembrane
(post-mortem)," Acustica27, 66-81.
Kohll/Sffel,L. U. E. (1972c). "A studyofbasilar membranevibrationsIII.
Thebasilarmembranefrequencyresponse
curvein thelivingguineapig,"
Acustica 27, 82-89.
Kohll/Sffel,L. U. E. (1974). "Recordingsfrom spiralganglionneurons,"in

Factsand Models in Hearing, edited by E. Zwicker and E. Terhardt,
(Springer,New York), pp. 193-203.
Kohll/Sffel,L. U. E. (1975). "A studyof neuroneactivityin the spiralganglionof thecat'sbasalturn," Arch. Oto. Rhino.Laryngol.209, 179-202.
Kringlebotn,M., Gundersen,T., Krokstad, A., and Skarstein,O. (1979).
"Noiseinducedhearinglosses,"Acta Otolaryngol.Suppl.360, 98-101.
LePage,E. L., andJohnstone,B. M. (1980). "Nonlinear mechanicalbehavior of the basilarmembranein the basalturn of the guineapig cochlea,"
2604
Hear. Res. 2, 183-189.
Russell,I. J., and Sellick, P.M. (1978). "Intracellular studiesof hair cellsin
LePage,E. L. (1981). "The roleof nonlinearmechanical

processes
in mammalianhearing,"Ph.D. thesis,The Universityof WesternAustralia,
Nedlands, Australia.
Liberman,C. E. (1982). "The cochlearfrequencymapfor thecat:Labeling

auditory-nervefibersof knowncharacteristic
frequency,"J. Acoust.Soc.
Am. 72, 1441-1449.
the guineapig cochlea,"J. Physiol.284, 261-290.

Russell, I. J. (1988). Personal communication.
Ryan, A., and Dallos, P. (1975). "Effect of absenceof cochlearouter haircellson behavioralauditorythreshold,"Nature 253, 44-46.
Ryan, A., and Bone, R. C. (1978). "Noise induced threshold shift and
cochlearpathologyin the mongoliangerbil," J. Acoust.Soc.Am. 63,
Liberman,C. E., and Kiang, N. Y.-S. (1978). "Acoustictrauma in cats,"

Acta Otol. Suppl.358, 1-63.
Mayer, A.M. (1894). "Researchesin acoustics--IX" Philos. Mag. 37,
259-288; or (1894)." Art. I.--Researches in acoustics,"Am. J. Sci. 47,
1145-1151.
Moore, B.C. J. (1983). Personal communication.
Schmiedt,R. A., andZwislocki,J. J. (1977). "Comparisonof soundtransmissionand cochlear-microphonic

characteristics
in Mongoliangerbil
and guineapig," J. Acoust.Soc.Am. 61, 133-149.
Schuknecht,H. (1953). "Techniquesfor studyof cochlearfunctionand
pathologyin experimentalanimals,"Acta Oto-Laryngol.58, 377.
Schuknecht,
H., and Neff, W. D. (1952). "Hearinglosses
after apicallesionsin the cochlea,"Acta Otolaryngol.42, Fasc.3, 263-274.
Schuknecht,
H., andSutton,S. (1953). "Hearinglosses
afterexperimental
lesionsin basalcoil of cochlea,"A.M. A. Arch. Otolaryngol.57, 129-
Moore, B.C. J., and Glasberg,B. R. (1983). "Suggested

formulaefor calculating auditory-filterbandwidthsand excitationpatterns,"J. Acoust.
Sellick, P.M., Patuzzi, R., and Johnstone,B. M. (1982). "Measurement of
1-28.
Merzenich, M. M., Kitzes, L., and Aitkin, L. (1973). "Anatomical and
physiological
evidencefor auditoryspecialization
in the mountainbeaver
[AplodontiaRufa]," Brain Res.58, 331-344.
Moody, D. B. (1988). Personalcommunications.
Soc. Am. 74, 750-753.
Moore, B.C. J. (1986). "Parallelsbetweenfrequencyselectivitymeasured

psychophysically
andin cochlearmechanics,"
Scand.Audiol. Suppl.25,
139-152.
Neubert, K., and Wiistenfeld, E. (1985). "Nachweis der zellularen Ans-
prechgebiete
im Innenohr,"Naturwissenshaften
42, 350-351.
Patterson,R. D. (1976). "Auditory afilter shapesderivedin simultaneous
and forwardmasking,"J. Acoust.Soc.Am. 70, 1003-1014.
Patterson,R. D., Nimmo-Smith, I., Weber, D. L., and Milroy, R. (1982).
"The deteriorationof hearingwith age:Frequencyselectivity,the critical
ratio, the audiogram,and speechthreshold,"J. Acoust. Soc. Am. 72,
1788-1803.
Pestalozza,G., and David, H. (1956). "Electricresponses

of the guineapig
Plomp, R., and LeveR,W. J. M. (1965). "Tonal consonance
and critical
bandwidth," J. Acoust. Soc. Am. 37, 548-560.
Plomp,R., andSteeneken,H. J. M. (1968). "Interferencebetweentwo simple tones,"J. Acoust. Soc.Am. 43, 883-884.
Rhode, W. S. (1971). "Observations of the vibration of the basilar mem-
brane in squirrelmonkeysusingthe M6ssbauertechnique."J. Acoust.

Soc. Am. 49, 1218-1231.
Rhode, W. S. (1973). "An investigationof post-mortemcochlearmechanismsusingthe M6ssbauereffect,"in BasicMechanismsin Hearing,edited by A. Mller (Academic,New York), pp. 49-68.
Rhode, W. S. (1978). "Some observations on cochlear mechanics," J.
Acoust. Soc. Am. 64, 158-176.
Robertson,D., and ManIcy, G. (1974). "Manipulationof frequencyanalysisin thecochlearganglionof the guineapig,"J. Comp.Physiol.91, 393375.
Robertson,D., Cody, A. R., Bredberg,G., and Johnstone,B. M. (1980).

"Response
propertiesof spiralganglionneuronsin cochleasdamagedby
direct mechanical trauma," J. Acoust. Soc. Am. 67, 1295-1303.
Robles,L., Ruggero,M. A., and Rich, N. C. (1985). "M6ssbauermeasurementsof the mechanicalresponseto single-toneand two-tonestimuli at
the baseof the chinchillacochlea,"in PeripheralAuditoryMechanisms,
editedby J. B. Allen, J. L. Hall, A. Hubbard, S. T. Neely, and A. Tubis
(Springer-Verlag,Berlin), pp. 121-128.
Robles,L., Ruggero,M. A., and Rich, N. C. (1986). "Basilar membrane
mechanicsat the baseof the chinchillacochlea.I. Input-output functions,tuningcurves,andresponse
phases,"J. Acoust.Soc.Am. 80, 13641374.
142.
basilarmembranemotionin the guineapig usingthe M/Sssbauer

technique," J. Acoust. Soc.Am. 72, 131-141.
Shailer,M. J., andMoore,B.C. J. (1983). "Gap detectionasa functionof
frequency,bandwidth,and level," J. Acoust. Soc.Am. 74, 467-473.
Smith,K. R., andWever,E.G. (1949). "The problemof stimulationdeafness.III. The functionalandhistological
effectsof a highfrequency
stimulus." J. Exp. Psychol.39, 238-241.
Sokolich, W. G., Hamernik, R. P., Zwislocki, J. J., and Schmiedt, R. A.
(1976). "Inferred responsepolaritiesof cochlearhair cells,"J. Acoust.
Soc. Am. 59, 963-979.
Stebbins,W. C. (1970). "Studiesof hearingand hearinglossin the monkey," in Animal Psychophysics:

TheDesignand Conductof SensoryExperiments,editedby W. C. Stebbins(Appleton-Century-Crofts,New
York) '-'- '
Stebbins,W. C. (1986). Personal communication.
Stebbins,W. C. (1988). Personal communication.
Stebbins,W. C., Clark, W. W., Pearson,R. D., and Weiland, N.G. (1973).
"Noise-anddrug-inducedhearinglossin monkeys,"in Advancesin Otorhinolaryngology:
Otophysiology,
editedby J. E. Hawkins,Jr. andM. Lawrence(Karger, Basel), pp. 42-63.
Stebbins,W. C. and Moody, D. B. (1979). "Comparativebehavioraltoxicology,"NeurobehavioralToxicology,I, Suppl. 1, 33-44.

Weber, D. L. (1977). "Growth of maskingand the auditory filter," J.
Acoust. Soc. Am. 62, 424-429.
Wilson,J.P., and Evans,E. F. (1977). "Cochlearfrequencymap for the

cat," in Psychophysics
and Physiology
of Hearing,editedby E. F. Evans
and J.P. Wilson (Academic,New York), p. 69.
Wilson,J.P., andJohnstone,
J. R. (1972). "Capacitativeprobemeasures
of
basilarmembranevibration,"in Hearing Theory(IPO, Eindhoven,The
Netherlands),pp. 172-181.
Wilson, J.P., and Johnstone,J. R. (1975). "Basilar membrane and middle-
earvibrationin guineapigmeasured
by capacitiveprobe,"J. Acoust.Soc.
Am. 57, 705-723.
Zwicker, E., Flottorp, G., and Stevens,S.S. (1957). "Critical bandwidthin

loudnesssummation," J. Acoust. Soc. Am. 29, 548-557.
Zwicker,E., andTerhardt,E. (1980). "Analyticalexpressions

for criticalbandrate and criticalbandwidthasa functionof frequency,"J. Acoust.
Soc. Am. 68, 1523-1525.
Zwislocki, J. J. (1965). "Analysisof someauditory characteristics,"in

HandbookofMathematicalPsychology,
Vol.III, editedby R. D. Luce,R.
R. Bush,and E. Galanter (Wiley, New York), p. 69.
Ruggero,M. A., and Robles,L. (1984). Personalcommunication.
2605
2605

1990 - Greenwood - Cochlear Frequency-Position Function For Several Species - 29 Years Later

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1990 - Greenwood - Cochlear Frequency-Position Function For Several Species - 29 Years Later

Uploaded by

Copyright:

Available Formats

A cochlear frequency-position function for several

( Received30 May 1989;accepted

to the derivative with a constant of 1 rather than with some

1961a,b). The developmentof the function assumedthat

other constant,as would have beenimplied by correspondenceto a differentdistancein millimeters.The comparison

CB = 10(ax+ b), of distance

J. Acoust.Soc.Am.87 (6), June1990

also be related to the questionof what is the "effective"

especiallyits slopeconstant,to the augmenteddataof recent

wheresuitableconstants(for man) are:A = 165.4 (to yield

ment envelopesalong the basilar membrane [see also

exerted some influence on both setsof data, that effect in the

andotherwise0.85 throughoutthispaperastwo valuesthat

J. Acoust.Soc. Am., Vol. 87, No. 6, June 1990

35t ......., ........ , ........ ' /

FIG. 1. Top panel:Takenfrom Kringlebotnet al. (Fig. 1, 1979). Distance

whenthe vibratingpoint wasobservedasit passedthrough

patternposition,at leastof relativepatternposition[and

J. Acoust. Soc. Am., Vol. 87, No. 6, June 1990

k(sy'stemporalbonepreparationsmay havebeencomparably freshand that the compositedata may possess

A. Primary fiber CF and cochlear site of innervation

matesof cochlearlength [from thoseof Schuknechtand

ofbasilarlengthor, if averagemillimetermeasureis desired,

B. Other cat data: Spiral ganglion cell CF and

but three,when groupedfor analysis,fell in 2-dimensional

20 and 35 kHz. The coordinatesof the cellswere in quite

whose value of A was too small, demonstrated that, for an

the basal end, with both Liberman's data and Function ( 1)

appropriatelyselectedA, generatingan empiricallycorrect

in Fig. 2 and with the original 22-mm functionsincethe

well, or better than, it had fit the old.

Hence, as Greenwoodhad done in 1961 for the eight

FIG. 2. AdaptedfromLiberman(1982). CurveisFunction( 1) fittedto the

Two bodies of mechanical data had been obtained over

the basal 9 mm of the cat cochlea,by Wilson and Evans

goodagreementin slopewith Schuknecht's(1953) basal

Wilson and Evans slopeis in still closeragreementwith

C. Why might mechanical peak frequencies be

cochlearlength;k = 0.8 (rather than 1).

Althoughthe slopesarethe same,thepeakfrequencies

Donald D. Greenwood:Cochlear frequency-positionfunction

than 2.1/22 as in 1961) ifx is to be scaledin mm and referencedto mean

J. Acoust. Soc. Am., Vol. 87, No. 6, June 1990

relating unit CF to position.Liberman listedtwo possible

On the hypothesisof cochleardisruption,a shift of the

The same basal shift of the displacementmaximum,

affected,however,for tone frequencieslower than the former CF, frequencies

innervatingthe damagedpointsshouldbe higher,making

guineapig ( Greenwood,1974), it isprobablynomorethana

largerthan the lengthof the apicalsegmentof the envelope,

sameinterval noted above).

A map of frequencyversuspositionfor the chinchilla

Donald D. Greenwood: Cochlear frequency-positionfunction

D. Why might frequency-place correlations of hearing

lationsof the CFs of singleauditoryunitsshowingthreshold

J. Acoust. Soc. Am., Vol. 87, No. 6, June 1990

ever, the chinchilla data, like the cat data of Schuknecht

(1953) and Liberman and Kiang (1978), are basedon a

Also shownin Fig. 3 (solid circle) is the determination

theconstantA in Eq. ( 1) issetto about163.5,a valuenearly

direct determinationslike those of Robles et al. (1985).

log-linearplot. The coefiScient

Thus an almost-exponentialfunction, Function (1),