See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/19946072

Four channels mediate the mechanical aspects of touch

Article in The Journal of the Acoustical Society of America · December 1988

DOI: 10.1121/1.397184 · Source: PubMed

CITATIONS READS
991 3,238

4 authors, including:

George A Gescheider
Syracuse University
129 PUBLICATIONS 8,008 CITATIONS

SEE PROFILE

All content following this page was uploaded by George A Gescheider on 06 June 2014.

The user has requested enhancement of the downloaded file.

Four channels mediate the mechanical aspects of touch
S.J. Bolanowski,
Jr.a)
Centerfor Brain Research,Unioersity
ofRochester
MedicalSchooland Centerfor YisualScience,University
ofRochester,Rochester,New York14642andInstitutefor Sensory
Research,SyracuseUnioersity,Syracuse,
New York 13244

G.A. Gescheider
Department
ofPsychology,
HamiltonCollege,
Clinton,New York13323
R.T. ¾errillo
Institutefor Sensory
Research,
SyracuseUnioersity,
Syracuse,
New York13244

C.M. Checkosky
Centerfor Brain Research,Universityof Rochester
MedicalSchool,Rochester,
New York14642

(Received16December1987;acceptedfor publication9 June1988)

Althoughpreviousphysiological andanatomicalexperiments haveidentifiedfourafferentfiber
types(PC, RA, SAII, and SA I) in glabrous(nonhairy) skinof the humansomatoscnsory
periphery,only threehavebeenshownto mediatetactile (mechanoreceptive) sensation.
Psychophysical evidencethat four channels(P, NP I, NP II, and NP III) do, indeed,
participatein the perceptualprocessis presented.In a seriesof experiments involvingselective
maskingof the variouschannels,modification of the skin-surfacetemperature, andtesting
cutaneous sensitivitydownto very low-vibratoryfrequencies, the fourthpsychophysical
channel(NP III) is defined.Basedon theseexperiments and previouswork from our
laboratory,it is concludedthat the four channelswork in conjunctionat thresholdto createan
operatingrangefor the perceptionof vibrationthat extendsfrom at least0.4 to greaterthan
500 Hz. Eachof the four channelsappearsto mediatespecificportionsof the overallthreshold-
frequencycharacteristic. Selectionof appropriateneural-response criteriafrom previously
publishedphysiological data and correlationof their derivedfrequencycharacteristics with the
fourpsychophysical
channels
indicates
thateachchannel
hasitsownphysiological
substrate:
P
channel and PC fibers,NP I channel and RA fibers,NP II channel and SAII fibers,and
NP III channeland SA I fibers.Thesechannelspartiallyoverlapin their absolutesensitivities,
makingit likely that suprathreshold
stimulimay activatetwo or moreof the channelsat the
sametime.Thustheperceptualqualitiesof touchmaybedeterminedby the combinedinputs
from four channels.

PACS numbers:43.66.Wv, 43.63.Vi [WAY]

INTRODUCTION sonet aL, 1982a).This segregation of responsecharacteris-

ticsis apparentlypresentbeyondthe primaryafferentlevel,
Microneurographic recordingsfromnervesinnervating
the glabrousskin of the human hand have isolatedfour extendingeven to cortical levels (for review, see Dykes,
1983).
groupsof low-threshold,mechanoreceptive fibers:(a) rap-
idly adapting(RA) fibers;(b) type I, slowlyadapting(SA In spiteof the fact that thereappearto be four separate
I ) fibers;(c) typeII, slowlyadapting(SAII ) fibers;and (d) classes of fiberspresumably involvedin mediatingthe non-
Paciniancorpuscle(PC) fibers (Johanssonand Vallbo, painful aspects of taction, psyehophysicalexperimentson
1979;for review, seeVallbo and Johansson,1984). In addi- the glabrous skin of human observershave assayedonly
tionto possessing differentratesof adaptationin theirre- three such information channels (Capraro et al., 1979;Bo-
lanowskiandVerrillo, 1982;GescheideretaL, 1985;Verrillo
sponses to pulselikemechanicalstimuli,the fibertypeshave
and Bolanowski, 1986). The P channel, as measured in
other functionalcharacteristicsthat help permit classifica-
tion into the four categories.For example,PCs and SAIIs thresholdexperiments,typically operatesover the vibra-
have large receptivefieldsthat lack distinctborders,while tory-frequencyrange of 40-800 Hz and possesses a U-
the receptivefieldsof RAs and SA Is are smaller but better shapedthreshold-frequencycharacteristic
with a maximum
defined (Johansson, 1976, 1978). Furthermore, the differ- sensitivitynear 300 Hz (for review,seeVerrillo, 1975}. This
ent fiber typesseemto havedistinctcapacitiesto respondto channelis highly sensitiveto changesin skin-surfacetem-
perature (Bolanowskiand Verrillo, 1982) and to changesin
specificfrequencyrangesof vibratorystimuli,as displayed
stimulus size and duration (Verrillo, 1962, 1963, 1966b).
by the"tuning"propertiesofisoresponsecontours(Johans-
Since increases in both the size and duration of a stimulus

'• Requests
for reprintsshouldbesentto Dr. S.J. Bolanowski,
Jr.,Institute preferentially
activatingtheP channelproduces decreasesin
for SensoryResearch,SyracuseUniversity,Syracuse,NY 13244. threshold(Verrillo, 1965, 1968;Craig, 1968), it is conclud-

1680 J. Acoust.Soc. Am. 84 (5), November1988 0001-4966/88/111680-15500.80 ¸ 1988 AcousticalSocietyof America 1680
edthattheP systemiscapableof spatialandtemporalsum- trical activation of individual SAII fibers in human observ-
mation. The sensoryattribute routinely ascribedto the P ersdid not elicita consciousperception.They pointedout,
channelis that of"vibration" (Talbot et al., 1968). It is well however,thatperhaps severalSAlI fibersmustbeactivated
documented thatPaciniancorpuscle afferents
aretheinputs or recruitedbeforea sensationcanbeelicited,a formof spa-
to the P channel(Verrillo, 1966a;Mountcastleeta!., 1972; tial summation.Alternatively,they suggestthat the SAII
Bolanowskiand Verrillo, 1982). fibersmay be responsible, not for perception,but for pro-
The non-PaeinianI (NP I) channel(Capraroet al., prioceptivefeedbackin motorcontrol(seealso,Knibest/31,
1979)islesssensitive
to changes
in stimulusfrequency(Ver- 1975). Contraryto thislatterideaarethe resultsof Harring-
rillo, 1962, 1968; Gescheider, 1976; Verrillo and Bolan- ton and Merzenich (1970) who recordedfrom peripheral-
owski, 1986) than the P channel and has, at threshold, a nervefibersin the hairyskinof themacaque.They conclud-
relativelyflat responseacrossvibratoryfrequencies. The NP edthat SAII fibersareinvolvedin mediatingtheintensityof
I channelpossesses a nominaloperatingrangeof vibration tactile sensations.Unfortunately,the glabrousskin of the
frequencies between10 and 100Hz (Labset al., 1978;Ges- macaque containsonlythreetypesof response profiles(PC,
cheideret al., 1985;Verrillo and Bolanowski,1986). It ap- RA, and SA; see,'for example,Freemanand Johnson,
pearsnot to be affectedby variationsin temperature(Ver- 1982a,b,and Johnson,1987), supportingthe notionthat
rillo and Bolanowski,1986) nor does it display either perhapsonly threechannelstransmittactileinformation
temporalor spatialsummation(Verrillo, 1965;Gescheider, from the volar surfaceof the hand to the somatosensory cor-
1976). The sensoryattributeascribedto thischannelis that tex.

of "flutter" (Talbot et al., 1968). The physiological corre- Interpretationof thephysiological resultsreliesheavily
late of NP I is presumablythe RA fibers(Lindblom and onassumptions regardingtheneuralcodeusedfor signaling
Lund, 1966; Talbot etal., 1968; Mountcastleet al., 1972), peripheralevents,but thecodefor somatosensation hasnot
which are believedto innervate Meissnercopuscles(Lind- beenadequately deciphered. It ispossible that,eventhough
blom, 1965). afferentfibersmay be rigorouslyexcited,unlessit is done
A secondnon-Pacinianchannel,NP II, the existenceof with the precisecode,the artificiallyinducedactivitymay
whichwasoriginallyproposed byCapraroetal. (1979), has not enter into consciousness. It could then be erroneously
beenshownby Gescheideret al. (1985) to operatein the concludedthat the activatedunitsplayno rolein mediating
vibratory-frequency rangesimilarto theP channel(i.e., 15- the senseof touch.
400 Hz) but at a muchlowersensitivity.In their study,the Rather than rely on physiologicaltechniquesthat re-
channelwasdefinedby desensitizing the P channelthrough quireassumptions of criterionresponse (Bolanowskiand
theuseof a smallstimulusarea (i.e., minimizingspatialsum- Verrillo, 1979), we havechosento usepsychophysical mea-
mation,seeabove)andby preventingNP I fromcontribut- surements to reassesswhether a fourth, low-threshold me-
ing to the overall threshold-frequency characteristic chanoreceptive channelcan or doescontributeto the sense
througha maskingparadigm(Gescheider et al., 1979;Ges- of touch.Sincethe fourthchannel,if present,mightbe ob-
cheideret al., 1982;Gescheideret al., 1983). Verrillo and scuredby other, more sensitivechannelsor may operateat
Bolanowski(1986) similarlyhave shownthe existenceof vibratoryfrequencies not previouslyinvestigated, we have
two separateNP channelsin the glabrousskinby usinga measured psychophysical thresholds from 0.4-500 Hz and
smallstimulusprobeto eliminatecontributions to threshold have used a forward-masking paradigm, as well as variations
by the P channelin conjunctionwith variationsof skin-sur- in stimulusdimensions andtemperature, to differentiate the
facetemperature.They founda differentialeffectof tem- various channels. Since electrical stimulation of SA I fibers
peratureon thethreshold-frequency response of humanob- apparently produces thesensation of"pressure" (Torebj/3rk
servers, indicating two separate channels. At low and Ochoa, 1980; Vallbo, 1981; Ochoa and Torebj/Srk,
frequencies (12-100 Hz), temperaturechangeshad no ef- 1983), a staticor slowlychangingexperience,it wasexpect-
fecton theaveragesensitivity asexpected for theNP I chan- ed that a fourth channelwould mostlikely operateat very
nel.For stimulusfrequencies above100Hz, however,tem- low stimulusfrequencies. This notionwasalsosuggested by
peraturehad a large effect on the thresholdresponses. thefrequency-response characteristics of SA fibersrecorded
Because of thestimulusconfiguration usedin thisstudy,the in macaqueperipheralnerve (Freeman and Johnson,
high-frequency systemmusthavebeenNP II. In additionto 1982b)andby the resultsofJohansson etal. (1982a,b) who
beingtemperature sensitive,NP II isalsocapableof tempo- recordedactivityof all four fibertypesin man.Bothgroups
ral summation (Gescheider et al., 1985). It is not known if of investigators foundthat SA fiberscouldbedrivenby low-
NP II displaysspatialsummation.Neither the sensoryat- frequencyvibratorystimulation.
tributenor the physiological correlateof NP II is known The results indicate that a fourth channel does indeed
presently, althoughit is highlylikelythat SA fibersare in- existandoperatesin the vibratoryfrequencyrangebetween
volved. 0.4 and 100 Hz. The channel, which we call NP III; has a
Priorto thepresentstudy,the psychophysical evidence sensitivitycomparableto that of the NP I channeland is
that only threechannelsapparentlycontributeto vibrotac- affectedby skin-surfacetemperature.NP III doesnot ap-
tile sensationsuggestedthat at leastoneof the four physio- pear to have the capabilitiesof spatialsummation.These
logicallyidentifiedafferentgroupsdoesnot contributeto facts, togetherwith a reinterpretationof previouslypub-
somatosensory perception.This ideawassupported by the lishedpsychophysical andphysiological data,formthebasis
work of OchoaandTorebj/Srk(1983) who foundthat elec- of a four-channel model for the sense of touch.

1681 J. Acoust.Sec. Am., VoL84, No. 5, November1988 Bolanowski.Jr. etal.: Mechanicalaspectsof touch 1681
I. METHODS followedby a periodof 25 msafterwhichthe teststimulus
The methodsand proceduresusedin the presentstudy waspresented. No stimulationwaspresentduringthe25-ms
weresimilarto thoseusedpreviouslyin thislaboratory(Ver- temporalgap. As shownby severalinvestigators(Hamer,
rillo, 1962, 1963, 1965, 1966a-c;Verrillo and Capraro, 1974; 1979;Gescheideret al., 1982;Hamer et al., 1983) masking
Verrillo and Gescheider, 1977; Bolanowski and Verrillo, occursonly whenthemaskerand the teststimulusexcitethe
1982; Gescheider et aL, 1985; Verrillo and Bolanowski, samechannel.It is thuspossible,
dependinguponthe mask-
1986). Sinusoidal,vibratorydisplacements of the skinwere ing- and test-stimulus
frequencies,
to differentiatebetween
producedwith a Goodman's390A vibratorundercomputer two channelsby maskingoneof them without affectingthe
controland appliedto the thenateminence.The displace- other. For example,if the frequencyand intensityof the
mentswereproducedarounda staticindentationof 0.5 mm maskingstimulusissuchthat it stimulatesthesamechannel
and monitoredwith an electromagnetic, linear-voltagedis- that the test stimulusactivates,detectionof the test stimulus
placementtransducerthat sensesthe displacementof the will be impaired.The amountof maskingdependsuponthe
movingelementof the vibrator. Stimuli were appliedwith intensityof the maskingstimulus(Gescheideret al., 1982).
either of two circular contact surface areas, 0.008 and 2.9 If, on the other hand, the maskingand the test stimuli acti-
cm2 and weresurrounded
by a rigidsurface(33 cm2) that vate differentchannels,then no maskingwill occur. Mask-
confines the deformations to the area of stimulation. The ingisdefinedasa lossof sensitivity in detectinga teststimu-
contactorsurfaceand surroundwere separatedby a l-mm lus in the presenceof a masker,and is expressedas a
gap.The temperature of theskinsurfaceunderlyingboththe thresholdshiftrelativeto the unmaskedthreshold.By vary-
contactorand the surroundwas controlledduring a given ingtheintensityandfrequencyof themaskingstimulusand
experimentalseries.The temperaturewasmaintainedbycir- the frequencyof the test stimulus,it is possibleto differen-
culatingwaterat the appropriatetemperaturethroughhol- tiate any underlyingand independentchannels.In our ex-
low chambers in both the aluminum contactor and sur- periments,the intensitiesof the maskingstimuli are ex-
round. A Lauda/Brinkman K-2/R heating and pressedrelativeto the thresholdof themasker,i.e., sensation
refrigerationunit and pumpcontrolledboththe watertem- level (SL), and are givenin dB referencedto the masking-
perature and the circulation flow rate. The contactor and stimulusthresholdlevel (dB SL ). In eachexperimentalses-
surroundwereplacedin seriessuchthat the water from the sion, the detectionthresholdof the maskingstimuluswas
circulatingunit firstflowedthroughthe contactorand then determinedand the intensityof the maskerwas set in dB
throughthe surround,eventuallyreturningto the circula- relative to that value. Thresholds were then measured for
tionunit.Thesizeofthe0.008-era
2contactor
didnotpermit detectingthe teststimulusalone.Finally, thresholdsfor de-
water to be circulatedthrough it, but, for both contactor tectingthe teststimulusin the presenceof the maskingstim-
sizes,the skin-surfacetemperatureat both the contactor/ ulus were measured.The maskingexperimentswere con-
skin interfaceand the surround/skininterfacewere always ductedonly at 30 øC.
within 0.5 øC. This was monitored by small thermistors Thresholdsweremeasuredby a two-alternative,forced-
placedeitherat (2.9-cm
2contactor)
or nextto (0.008cme) choicetrackingprocedure(Zwislockiet al., 1958) in which
the contactor/skin interface. The active surface of the ther- subjectsjudged which of two sequentialintervalsof time
mistor touched the skin. To monitor the surround/skin tem- containedthe teststimulus.Their judgmentswere signaled
perature,an additionalthermistorwasalsoplacedat that bydepressing
oneof two switches,
theresponse
recordedby
junction (seeBolanowskiandVerrillo, 1982). Prior to each computer.
Theprobability
of theteststimulus
occurring
in
testingsession, thesubject's handwasallowedto adaptto the one interval versus the other was 0.5. The intervals were
contactor/surround temperaturefor a periodof no lessthan separatedin time by 1750 ms. The stimulusintensitywas
10 min. In all, the effectsof threeskin-surfacetemperatures increasedby I dB for everyresponse
erroranddecreased by
( 15, 30, and 40 øC) wereinvestigated. 1 dB for everythreecorrectresponses.In this manner,the
In one groupof experiments,sinusoidalburstswith du- criterion used to determine threshold was 75% correct de-
rationscalculatedat the half-powerpointsof 700 or 2400 ms tectionovera periodof 3 min. The actualperiodof stimula-
were used as test stimuli. The rise-fall time at the onset and tion was considerablylongerthan this sincethe stimulus
offset of the stimulus burst was 500 ms when the burst dura- intensitywas alwaysa suprathreshold level prior to each
tion was 700 ms, and 1250 ms for the 2400-ms stimulus. test.Nominally,a periodof 3-6 min wasrequiredbeforethe
These durations were chosen to ensure that most of the ener- subjectreachedthe approximatethresholdlevel at which
gy in the stimulus,regardlessof the testfrequencyused,oc- timethe3-minperiodof 75% detectionwasstarted.Thresh-
curredat the fundamentalfrequency.This wasparticularly oldswereexpressed in dB referenced
to l-/.tmpeakdisplace-
importantsincethe vibratory-frequency rangeinvestigated ment. In the maskingexperiments,the maskingstimulus
extendedfrom 0.4-500 Hz requiring fairly long-duration waspresentedin eachof the two observationintervals,and
stimuli (see the Appendix). Thresholdvaluesare given in the subjectwas requiredto selectthe interval that alsocon-
decibels(dB) referencedto 1-/zmpeakdisplacement.For a tained the test stimulus.Every subjectwas presentedwith
secondgroup of experiments,only test stimuli of 700-ms each test condition three times over several sessions. A total
duration were used,but a conditioningstimulusthat pro- of five subjects,two females (ages 20 and 27) and three
ducedforwardmasking(GesheideretaL, 1979,1982,1983) males(ages26, 36, and49) wereused.All werewell experi-
of the teststimuluswaspresentedfirst. The maskingstimu- encedwith thistypeof tactileexperiment.Individual results
lus was presentedfor 4 s (rise-fall times, 50 ms) and was closelyfollow the group averages.Becauseof this closecor-

1602 J. Acoust.Soc. Am., Vol. 84, No. 5, November1988 Bolanowski,Jr. et al.: Mechanicalaspectsof touch 1682
respondence, asgroupaverages. frequency
the resultsare presented portions
ofthethreshold-frequency
characteristic
The standard error of the mean for individual observers'
suggests the presence of two separatechannels,onefor the
thresholdjudgmentsis, on average,1.5 dB. The standard low-frequency portionanda second,probablyNP I, for the
error of the meanfor groupestimates
are almostidentical middle-frequency part. Sinceno previousstudieshavemea-
being,on average,1.3dB. suredtheresponse ofNP I forfrequencies lowerthanabout2
Hz, however,it is possible that thelow-frequencyregionis
II. RESULTS alsomediatedby NP I. Maskingexperiments wereconduct-
ed to determine if this was the case.
A. The vibrotactile-frequencyresponse
Fiigure1 showsaveragepsychophysical
thresholdsob-
tainedfor frequencies from0.4-500Hz. Skin-sur- B. Effects of masking stimuli
ranging
facetemperaturewasheldconstantat 30 *C. Stimulusdura- In the maskingexperiments,a frequencyof 0.7 Hz was
tion was700 ms, and the contactorarea was2.9 cm:. The used as the test stimulus since it falls in the middle of the low-
maskingstimuluswasnot used.The line throughthe data
frequencyportionof the threshold-frequency characteristic
of Fig. 1.The frequencies
pointshasbeenfittedbyeyewith theerrorbarsdepictingthe of themaskingstimuliwerechosen
standarderrorof themeans.Threesegments of thecurvecan
to selectivelyaffectthe other portionsof the threshold-fre-
bediscerned.First, thereis a low-frequency quencycharacteristic.For example,in the first caseto be
portionextend-
ingbe•:ween presented,the maskingstimuluswas40 Hz.
0.4 and3.0 Hz whichappearsfrequencyinsensi-
tive.Second, thereappears tobea middle-frequency After measuringthethresholdof the0.7-Hz testand40-
portion,
Hz maskingstimuli alone,the maskingstimuluswas pre-
whichis frequencydependent,havinga slopein displace-
ment-frequencyunits of about -- 5.0 dB/oct. Third, a U-
sentedat preselectedSL levels,and the subjecttrackedthe
shapedportioncan be seenin the highestfrequencyrange,
thresholdfor the test stimulusin the maskingparadigm.
betweenthe frequencies of 40 and 500 Hz. The slopeof the
Five subjectsweretestedin eachconditionduring two ses-
low-frequencyportionof the U is -- 12 dB/oct. sionsin whicheachtestconditionwaspresented twice.Skin-
The high-frequency portionof thecurveis producedby
surfacetemperaturewas maintainedat 30 øC. Figure 2
the P channeland has the expectedshape,sensitivity,and
showsthe averagethresholdshift as a functionof masker
locationfoundin manyotherstudies assessing intensityfor the40-Hz maskingstimulus.In thisandsimilar
theproperties
of thischannel(seethe Introduction).The middle-frequen-
figuresto follow,a least-squaresanalysiswasusedto fit the
cy portionis probablyproducedby the NP I channel.A1-
datapoints(seefigurelegendsfor thecorrelationcoefficient
thoug]h r and slopem) affectedby the maskingstimulus.This was
in previousworkfromthislaboratoryNP I appeared
insensitiveto frequency,.this conclusionmay have resulted
done to producebreakpointsbetweenthe two segmentsof
fromthefactthat thelowestfrequency the maskingfunctions,sincethe exactpositionof the transi-
.investigatedwasonly
12 Hz. Other investigators(B6k6sy,1939;Mountcastleet
tion from no maskingto the point of maskingwasnot mea-
al., 1972) have usedlower frequencies and havefoundan
sured.In additionto usingthosepointsshowingan effectby
the masker,in the least-squares
effectof stimulusfrequencyon sensitivityin thisregion.The analysis,the highestpointat
fact[hata breakpoint
exists
between
thelow-andmiddle- whichmaskingdid not occurwasalsoincluded.A horizon-

8O
I I I I I III I I I I I I I I1• I I I I I I II I I I I I I I II

6,0

• 40
E FIG. 1. Threshold-frequency char-
acteristicrelatingstimulusintensity
2,0 to stimulus frequency.The results
are the averagesof five observers.
The errorbarsin thisand thefigures
I--
z to follow signifythe standarderror
I.u 0 of the means, their absenceindicat-
IJJ ingthat theerrorwastoosmallto be
depicted.Skin-surfacetemperature
.,_1 was maintained at 30øC. Stimulus
•. -2'0
contactor size was 2.9 cm •.

-40

-60
0. 1.o lO lOO 1000
FREQUENCY (Hz]

1683 J. Acoust.Soc. Am., Vol. 84, No. 5, November 1988 Bolanowski,Jr. ot aL: Mechanicalaspects of touch 1683
 15
40 Hz MASKER
133 ß 10 Hz
• o20Hz

C3 5
o

f,/)
uJ (3 ß
rr

-• i I I I I I I I
0 5 10 15 20 25 30 35 40 45
0 5 10 15 20 25 30 35 40 45
MASKER LEVEL (dB SL)
MASKER LEVEL (dB SL)
FIG. 3.Maskingfunctions
obtained
withvariousmasker
stimuli( 10Hz, O,
FIG. 2. Theeffectsofa 40-Hz masking stimuluson thethreshold ofa 0.7- r = 0.99, m = 0.41; 20 Hz, O, r = 0.99, m = 0.43; and ]00 Hz, &, r = 0.99,
Hz teststimulus.
Thelinesthroughthedatapointsfrom 14-to 38-dB mask- m = 0.46) and with a test stimulusof 0.? Hz. The breakpointsof these
er levelhavebeenfittedby linearregression
(r = 1.0,m = 0.46). curvesare plottedas X's in Fig. 4.

tal line wasfittedto all of the datapointsthat wereunaffect- breakpointsof the maskingfunctionsshownin Figs.2 and 3
edbythemaskingstimulus,thepointof intersection between were used to determine the thresholds of the channel mediat-
the two straight-linesegments thusobtainedbeingreferred ingdetectionof the0.7-Hz stimulus.Sincethe maskingex-
to asthebreakpoint.As shownin Fig. 2, maskinglevelsof 0, perimentswereconducted only at 30 øC,we havereplotted
6, 10,and 14dB SL producedno maskingof the teststimu- the threshold-frequency characteristic obtainedat thistem-
lus.The horizontalline determinedby the datapointssigni- peratureand described in Fig. 1 (i.e., unmaskedthreshold-
fiesthis fact. This meansthat the maskingstimuluswasnot frequencycharacteristic)into Fig. 4. In sodoing,we have
affectingthechannelmediatingthe0.7-Hz teststimulus.For eliminatedthe data pointsfor clarity. In addition,we have
maskinglevelsgreaterthan 14dB SL, however,themasking plottedthe breakpointsof the maskingfunctionsshownin
stimulusproduced thresholdelevations abovetheunmasked Figs.2 and3 intoFig. 4 andasshownby the X 's.The dashed
thresholdof the test stimulus.The presenceof masking line (seeFig. 4) connecting the thresholdvaluesobtainedat
above14dB SL indicatesthat thechannelmediatingthe test 0.7 Hz for the unmasked condition (and for the flat low-
stimuluswas being affectedby the maskingstimulus.A frequencyregionin general)with the maskedthresholds
least-squares analysisof the data from 14 dB SL and above givenby the X's defines,to a firstapproximation, the func-
indicatesthat the effectof the maskingstimulusoccurswith tionalcharacteristic of thechannelmediatingthe0.7-Hz test
an efficiency(i.e., slope) of 0.46. The breakpointat 10-dB stimulus. Since the obtained curve does not follow the over-
SL maskinglevel definesthe approximatelocationof the all threshold-frequency characteristic,a differentchannel
thresholdat 40 Hz of the channelmediatingthe 0.7-Hz test must be operatingin the low-frequencypart of the curve.
stimulus.Specifically,the channelmediating0.7 Hz has a Neither the P or the NP II channelscan be mediating the
thresholdat 40 Hz, which is 10 dB above the 30 øC curve low-frequencyportion of the overall threshold-frequency
shownin Fig. 1. characteristic sinceprevious experiments (Gescheider etal.,
Additional maskingexperimentswere performedat 1983;Gescheideret al., 1985) haveshownthat their sensiti-
maskingfrequencies of 10,20, and 100Hz in orderto deter- vities below 2.5 Hz are at least 27 dB above the unmasked
minethe breakpoints for variousmaskingfunctions.These threshold measured here.
breakpoints, asdemonstrated in Fig. 2, definethelocationof To substantiatethe independence of this low-frequency
the channel that in this case mediates the 0.7-Hz test stimu- channel(Fig. 4, ---), a maskingexperimentwasperformed
lus, at the frequencyof the maskingstimulus.Averagere- in which a 40-Hz maskingstimuluswascoupledwith a 25-
sultsfrom theseexperimentsare shownin Fig. 3. Masking of Hz test stimulus. Based on previous results (see the Intro-
the0.7-Hz teststimulusbeginsto occurat variousintensities duction), it was reasonable to assume that the 25-Hz test
of themaskingstimulus,depending on themaskerfrequen- stimuluswould be detectedby the NP I channel.The results
cy.For example,maskingstimuliat 100Hz producemask- are plotted in Fig. 5. Also shown is the maskingfunction
ing of the teststimulusonly at maskerlevelsgreaterthan obtainedwith the samemaskingfrequencybut with the 0.7-
about24 dB SL (asdefinedby thebreakpoint).On theother Hz teststimulus,the resultsreplottedfrom Fig. 2. Note that
hand,maskingoccursat verylow SL levelsof the 10-and 20- maskingof the 25-Hz teststimulusoccursat maskinginten-
Hz maskingstimuli. In eachinstance,the efficiencyof the sitiesaslow as I dB SL. In thisparticularcase,thefunctionis
maskerasit masksthe 0.7-Hz teststimulusis approximately bestfit by two segments,onefor low-maskinglevels(least-
the same ( 10 Hz, 0.41; 20 Hz, 0.43; and 100 Hz, 0.46) indi- squaresfit from 0-8 dB) and a secondfor the higherlevels
cating commonality in the channel being masked. The (least-squaresfit from 8-30 dB). Neither segmentoperates

1684 J. Acoust.Soc. Am., Vol. 84, No. 5, November1988 Bolanowski,Jr. et al.: Mechanicalaspectsof touch 1684
 60

FIG. 4. Threshold-frequency char-

40 acteristicsrelating stimulusampli-
tudeto stimulusfrequency.The sol-
id curvehasbeenreplottedfromFig.
1. The X's are the breakpointsob-
tainedin relationto Figs. 2 (40 Hz)
and 3 ( 10, 20, and 100 Hz). Dashed
line has been fitted to the X's and to
0
the flat, low-frequency
regionof the
L• solidcurvelyingbetween0.4 and 1.0
Hz. The + signifies the lowest
-20 breakpoint on the 40-Hz masking
stimulus/25-Hz test stimulus mask-
ing functionof Fig. 5.

-6( , , , , , ,,,I
o.1 1.o lO lOO 000
FREQUENCY (Hz)

with •,n efficiencyof 0.45, the lower portionpossessing an peratureincreased boththe averagesensitivityaswellasthe
etiiciencyof 0.8, while the upperhasan efficiencyof 0.31. frequencyat which maximum sensitivityoccurs,the so-
The maskingfunctionis quiteunlikethat obtainedwith the called"bestfrequency."The high-frequency portionof the
0.7-Hz teststimulus.The lowestpoint ( 1 dB SL) of the 25- three curvesis producedby the P channel.The significant
Hz teststimulus/40-Hzmaskingfunctionhasbeenplotted effectsof temperatureon its responseprofilehave already
in Fig. 4 ( + ), sinceit signifies
the pointat whichthe chan- been described(Bolanowski and Verrillo, 1982). The mid-
nel mediatingthe 25-Hz teststimulusoperates.The break- dle-andlow-frequency portionsof thecurvesshownin Fig. 1
point,asexpected,liesalmostdirectlyon the overallthresh- are producedby the NP I andNP III channels.Exclusiveof
old-frequencycharacteristic,indicating that both the the P region (40 to 500 Hz), changesin skin-surfacetem-
maskingandthe teststimuliarebeingmediatedby the same peraturehavethe greatesteffectin the regionbetween1.5
channel (NP I). Sincethis doesnot occur when a 0.7-Hz test and4 Hz, that regionon all threecurveswherethe sensitivi-
is usedin conjunctionwith a 40-Hz maskingstimulus,a sep- tiesofNP I and NP III arebelievedto mergeandcrossover.
aratechannelmustbe responsible for detectingthe 0.7-Hz To determineif temperaturehad a significanteffecton
test.In otherwords,two independentchannelsmustbe de- thresholdsbelow 40 Hz, a two-way analysisof variance
terminingthe psychophysical thresholdbetween0.4 and 30
Hz: oneresponsible for signalingdetectionat 0.7 Hz (and
the fiat portionof Fig. 1) andthe othermediatingthreshold
at 25 Hz (and the -- 5.0-dB/oct portionsof Fig. 1). The 15
previouslyknown NP I systemis responsible for the 25-Hz 40 z MASKER

channel.Followingpreviousnomenclature,we call the new-

ly revealedchannelNP III. 10
25
HZ
TEST
C. Effects of skin-surface temperature
It would be possibleto definemore preciselythe fre-
quencyat which detectionchangesfrom onechannelto the
other if the two low-frequencychannelswere differentially
affectedby changesin skin-surfacetemperature.Therefore,
threshtold measurements were made at two additional tem-
peratures, 15 and 40 øC. Other stimulus conditions were -5 I I I I I I I I
0 5 10 15 20 25 30 35 40 45
identicalto thoseusedpreviously
(Fig. 1,2.9-cm2contactor
area,700-msduration,and 500-msrise-falltimes).The re-
MASKER LEVEL (dB SL)
sults t¾omthe experimentsin which the skin-surfacetem-
peraturewasvariedareplottedin Fig. 6, alongwith the FIG. 5. The effectsofa 40-Hz maskingstimulusonthethresholdofa 25-Hz
teststimulus.The linesthroughthedatapointsfrom0- to 8-dBmaskerlevel
resultsobtainedat 30 øC.The linesthroughthe datapoints havebeenfittedby linear regression(r = 0.97, m = 0.8) asis the line fitted
havebeenfittedby eye.Temperatureclearlyhasan effecton to the maskerlevelsof 8-30 dB ( r = 0.98, rn = 0.31 ). The maskingfunction
the U.-shapedportion of all three curves.Increasesin tem- shownwithout data pointshasbeentakenfrom Fig. 2.

1685 J. Acoust. Sec. Am., Vol. 84, No. 5, November 1988 Bolanowski, Jr. otal.: Mechanical aspects of touch 1685
 80 [ Ilia[ [ [ [

SKIN-SURFACE TEMPERATURE

o 15øC
6C
ß 3oøc
ß 4oøc

FIG. 6. Threshold-frequencychar-
acteristicsrelating stimulus ampli-
• 20 tudeto stimulusfrequency.Skinsur-
face temperatureis the parameter:
15 øC,¸; 30 *C, O; and 40 øC,l Er-
z o ror barsarenotplottedfor clarityal-
though the standarderror of the
means are identical to those found
for Fig. 1 (seeSec.I).

-41

-60 t I I I I •'•l • • , i , ,•l , , ,

0.1 1.0 10 100 .... ;0oo
FREQUENCY (Hz)

(ANOVA) wasperformed,usingthedataobtainedforstim- D. Effects of stimulus size and duration

ulusfrequencies from 07, •.0 Hz. The resultsof the analysis
showedthat theeffectof temperature washighlysignificant In orderto determineif the NP III channelis affectedby
[F(2,204)--7.39, p<0.001]. Sincethe maskingexperi- stimulus
size,thresholds
wereobtained
usinga 0.008-cm
2
mentsshowthat two separatechannelsappearto be operat- contactorareaanda skin-surface
temperatureof 30 øC.Two
ingoverthisregion,two ANOVAs wereperformed,onefor stimulusfrequencies
wereinvestigated
(0.5 and 1.0Hz), and
the frequencyrangebetween2.5 and40 Hz andthe second thethresholds
compared
to thoseobtained
withthe2.9-cm
2
for the rangebetween0.4 and 2.0 Hz. In bothinstances, the contactor.Spatial summationin NP III would be demon-
effectsof varyingskin-surface temperatureprovedto besig- strated if the smaller stimulusarea producedthresholds
nificant(2.5-40 Hz, [F(2,120) = 4.08,p<0.01]; 0.4-2.0 higherthan the largerstimulusarea. When the thresholds
Hz, [F(2,84) = 5.2,p < 0.01] }. If therange
between
2.5and obtainedwiih the smaller contactor (0.5 Hz, 26.7 dB re: l-
40 Hz ismediatedsolelyby the NP I channel,thenit is likely /•m peak;1.0Hz, 27.6dB re: 1-ttmpeak) werecomparedto
that temperatureaffectsthe sensitivityof this channel.Al- thoseobtainedwith the large contactor (0.5 Hz, 27.7 dB re:
thoughin previousworkfrom thislaboratory(Verrillo and 1-ttmpeak;1.0Hz, 26.4dB re: l-ttm peak),nodifferences
in
Bolanowski, 1986)NP I appeared to tempera- threshold
insensitive wereevidentor statistically
significant
{0.5 Hz,
ture,thisconclusion
mayhaveresultedfromthefactthat the [t(4)=0.51, p>0.5]; 1.0 Hz, [t(4)=0.63, p>0.5]}.
lowestfrequencyinvestigated
in thatstudywasonly 12Hz. Whenthe200-Hz testfrequency wasused,however,thresh-
Similarly,sincetemperature apparentlyaffectssensitivity in oldsobtainedfor a contactsurfaceof 0.008cm2 ( 11.2dB re:
the frequencyrangebetween0.4 and 2.0 Hz and, if this re- 1-ttm peak) were considerably higher [t(4) =8.69,
gionis controlledsoleyby NP III, then this channel,too, p<0.001] than thoseobtainedwith the 2.9-cm 2 area
must be temperaturesensitive.The ANOVAs performed ( -- 15.4dB re: 1-/•mpeak), a resultpredictedby the pres-
alsoshowedthat the middle-frequencyregionwas strongly enceof spatialsummationin the P channel.The resultssug-
affected by vibration frequency [F(9,120) = 4.08, gestthatspatialsummation is notfoundin theNP III chan-
p < 0.001] asthe -- 5.0-dB/octslopefoundat a temperature nel.
of 30 øCwould predict.The low-frequency protionsof all The effectsof stimulusduration on the sensitivityof NP
three curves are not affected by stimulus frequency III was also investigated.Increasesin sensitivityfor in-
[F(6,84) = 1.6, p>0.1]. An ANOVA performedon the creasesin stimulusduration might suggestsomeform of
data from 0.4 to 40 Hz also showedthat frequencyhas a temporalintegrationof the stimulus.This was not found.
significanteffect on threshold throughout this range For example,thresholdsobtainedat 0.4 Hz (27.6 dB re: 1-
IF(32,204) ----72.98,p < 0.001]. However,sincethe interac- /_tmpeak) and 1.0Hz (26.4 dB re: 1-gmpeak) usingstimu-
tionbetweenstimulusfrequencyandtemperatureon thresh- lus duration of 2500 ms with rise-fall times of 500 ms
oldsin all cases wasnotfoundto besignificant (0. •. •.0Hz, showedno statisticallysignificantdifference(0.4 Hz,
[F(32,204)=1.36, p>0.10]; 0.4-20 Hz, [F(12,84) [t(4) = 0.86,p>0.4]; 1.0 Hz, [t(4) = 0.15,p>0.5]} in
= 1.15, p>0.25] and 2.5-40 Hz, [F(18,120) = 1.45, sensitivitywhen comparedto the thresholdsobtainedwith
p > 0.10] }, it isnotpossible
to determine
theexactmannerof the 700-msstimulus(0.4 Hz, 25.6 dB re: 1-/•mpeak;1.0 Hz,
the effectsof temperatureon eitherNP I or NP III. 28.1-/zmpeak).This findingalsoconfirmsthat the thresh-

1686 J. Acoust.Soc.Am.,Vol.84, No.5, November1988 Bolanowski,

Jr. ot a/.: Mechanical
aspectsof touch 1686
olds obtained at the shorter burst durations were not con- has a slopeof 0.31, it is unlikely that the P or the NP II
foundedby low-frequencyenergycomponents
in the stimu- channelsthemselvesare detectingthe test stimulusbut that
lus (seethe Appendix). thisisbeingdoneby NP III. It is alsopossible,however,that
the otherchannelsaresequentially mediatingthe teststimu-
lus for increasesin masker-stimulusintensity. In other
III. DIISCUSSlON
words, the 0.31 efficiencycalculatedmay be the result of
The resultspresented indicatethat vibratorystimulican theseother channelscontributingto this segmentof the
be detectedat frequencies between0.4 and > 500 Hz, the maskingfunctionin a complicatedway.
threshold-frequency characteristic beingU shapedfor high The experimentsin whichskin-surfacetemperaturewas
frequencies andpartiallyslopingfor the lowerones(Figs. 1, varied(Fig. 6) indicatethat temperaturecanhavea signifi-
4, and 6). The generalshapeof the characteristicwas first canteffecton thresholdsensitivityovermostof the frequen-
reportedby B6k•sy (1939) and later by Verrillo (1963, cy rangeinvestigated. Exclusiveof the P-channeleffectal-
1966a-c), usingmethodssimilar to thoseusedhere,except ready describedand discussedby Bolanowskiand Verrillo
that theydid not controlskin-surfacetemperature.The ma- (1982), temperatureseemsto affectthresholdsmediatedby
jor differencebetweenthisandtheearlierstudies,whencon- both NP I and NP III. The fact that the NP I channel is
sideredin conjunctionwith previousexperiments from our affectedby changesin skin-surfacetemperatureis inconsis-
laboratory(e.g., Capraroet al., 1979;Bolanowskiand Ver- tent with the report of Verrillo and Bolanowski (1986).
rillo, 1982;Gescheideret al., 1985;Verrillo and Bolanowski, However,sincethey only investigatedfrequencies above12
1986),isthatweconclude thatthereareprobablyfoursepa- Hz, the effectof temperaturein theirstudycouldhavegone
rate channelsmediating the senseof touch. The present unnoticed.
studyalsoshowsthat the previouslyunknownfourthchan- Sincethereappearto befourseparatechannelsinvolved
neloperatesat low-stimulus frequencies.B•k•sy (1939) and in signalingthesenseof touch,previousphysiological results
Verrillo (1963), on the other hand, postulateda two-chan- suggesting that onlythreefibertypesareinvolved(e.g.,Kni-
nel (duplex) modelfor vibrotactionbasedon their earlypsy- best61,1975;Ochoaand Torebj6rk, 1983) mustbe reevalu-
chophysicalresults. ated.The factthat repetitive,electricalstimulationof SA II
The fourthchannel,calledNP III, appearsto operatein fibersapparentlydoesnot evokea sensation(Ochoa and
thefrequencyrangefrom 0.4 to > 100Hz andhasan average Torebj6rk, 1983) suggests that the fibersin their studywere
sensitivitysimilarto that of the NP I channel(seeFig. 4). As not stimulatedappropriately.Failure to evokea sensation
with iNPI, NP III doesnot possess spatialsummation.The couldeitherbea resultof an improperimpulsepatternor the
presenceof NP III was determinedby maskingstudies fact that only a singleSAII fiber is activated,perhapsan
(Figs..2 and 3) in whichmaskingstimuliwerepairedwith a insufficientnumber (Ochoa and Torebj6rk, 1983). When
0.7-I-t[zteststimulus.The procedureelevatedthe thresholds reevaluating previousphysiological results,it wouldbequite
of P, NP I, and NP II for the purposeof definingNP III. useful,for the purposes of modeling,to correlateeachpsy-
Whenthe maskingstimuliwereintenseenoughto maskNP chophysically definedchannelwith its underlyingphysio-
III also,a powerrelationshipbetweenmaskingintensityand logicalsubstrate.Of primaryimportance in bothinstancesis
threslholdshift of the teststimuluswasfound. The efficiency the needto selectthe appropriateneuralcodeusedby each
(i.e.,:slope;seeGesheider etal., 1983)at whichthisoccurred fiber groupto signaldetectionof a stimulus(Bolanowski
wasapproximately0.44 asdeterminedby least-squares fit to and ¾errillo, 1982). The fact that artificial activation of indi-
the data. It is importantto notethat all four maskingfunc- vidualfibersmayproducea perceptualresponse (Torebj6rk
tionsobtained
withthe0.7-Hzteststimulus(Figs.2 and3) andOchoa,1980;Vallbo, 1981;OchoaandTorebj6rk, 1983)
havethesamemaskingefficiencies,providingadditionalevi- may be misleadingsinceit doesnot provethat the imposed
dencethat the samechannel(NP III) wasmediatingdetec- stimulusis the codeusedby the nervoussystem.In fact, the
tion of the 0.7-Hz test stimulus in each case. In contrast to sensations evokedby suchstimulationhavebeenreportedto
this, Fig. 5 showsthat pairinga maskingstimulusof 40 Hz feel somewhatunnatural (Ochoa and Torebj6rk, 1983).
witha teststimulusof 25 Hz produces a verydifferentmask- In general,previousreports on the responseof peri-
ingfunction.This isexpectedsinceboththe 25-Hz teststim- pheralnervefibersto vibratorystimulationusuallypresent
ulus and the 40-Hz masking stimulus, even at very low frequency-response curvesin whichtheintensityof thestim-
maskinglevels,are mediatedby the samechannel.NP I is ulusneededto achievea predetermined responsecriterionis
knownto havea maskingefficiencyof approximately0.8 plottedasa functionof vibrationfrequency.Only two crite-
(Gescheideret al., 1983). The lowerportionof the function ria are usually used:one neural impulseper stimuluscycle
obtained with the 40-Hz masker, combinedwith a 25-Hz test (entrainment) or one neural impulse per stimulus burst.
stimulus(Fig. 5) has a slopeequalto 0.81, indicatingthat They do not describewhat happensto the shapeof the fre-
this portion of the maskingfunction is indeedproducedby quencyresponseasa functionof changesin responsecriteria
NP I. The upperportionof the 40-Hz masking-,25-Hz test- (e.g., Lindbiomand Lund, 1966;Talbot et al., 1968;Mer-
stimulus condition could be due to the 25-Hz test stimulus zenichandHarrington,1969;Tapperetal., 1972;Mountcas-
being:detectedby either the P, NP II, or NP III channels. tle et al., 1972; Pertovaara and Iqiim•iliiinen, 1981; Freeman
The P and NP II channelshave maskingefficienciesof 0.62 and Johnson, 1982a). In fact, the reasonthat an entrainment
(Gescheider et al., 1983) and 0.9 (Gescheider et al., 1985), criterionwasinitially usedin the literatureon the topicwas
respectively.
Sincetheupperportionof themaskingfunction that it producedfrequency-response characteristicsof peri-

1687 J. Acoust. Soc. Am., Vol. 84, No. 5, November 1988 Bolanowski,Jr. eta/.: Mechanical aspects of touch 1687
pheralfibersthatqualitatively mimicked threshold-frequen-channel,a portion(12-150 Hz) of whichwastakenfrom
cy responses obtainedfromhumanobservers (e.g.,Mount- Verrillo and Bolanowski(1986) with the lower frequency
castleet al., 1972). Unfortunately,no reportsshowintensity aspect(2.5-12 Hz) basedon the resultspresented abovein
characteristicsrelatingafferentneuralresponsesto stimulus this study (seeFigs. 1 and 4). The shapeof the two curves
intensitiesat specifiedfrequencieswith sufficientsampling are similar.The averagesensitivityof the RA units,how-
pointsto ascertainthe effectsof response criterionon the ever,is 18 dB lesssensitivethan the psychophysical results.
shapeof the frequency characteristics
(but seeBolanowski This disparitymay be explainedby differences in stimulus
and Zwislocki, 1984). conditionsand the fact that averagedata were usedin con-
Someof themostcarefullyconducted research hasbeen structingthe physiological curve.As Mountcastleet al.
on the macaquemonkeyusedasa modelfor man.However, (1972) and Bolanowskiand Zwislocki (1984) haveshown,
as Johnsonpointsout (Freeman and Johnson,1982a,b; the range of sensitivitiesfor mechanoreceptors can be as
Johnson,1987), the macaqueapparentlyhas only three large as 20 dB. Thus the most sensitiveRA fiber, which
typesof response profilesandthe SA-unitprofilecannotbe would be considerablymore sensitivethan the average,
subdivided. Thus it may not be advantageous to usethese would be mediatingthreshold.Aside from this considera-
data sincetheymay not givea completedescriptionof the tion, Verrillo (1979) hasshownthat the presenceof an edge
mechanismsused by the human somatosensorysystem. eitherat or near the placeof stimulationcan significantly
Aside from the fact that there may be a speciesdifference ( 10-15 dB) increasethe sensitivityof the NP I channel.The
betweenmacaqueandman,recentdevelopments in micro- presentexperiments,as well as most of the previousones
neurography makeit possible to usephysiological results from our laboratory,wereperformedwith the useof a rigid
obtainedfrom humansas correlatesto the psychophysical surfacethat surroundedthe stimulusprobe.This stimulus
resultspresented here.At present,themostcomplete study configuration introducestwo largecircularedges,oneat the
assessing single-unitresponses to vibratorystimuliin hu- boundaryof the surroundand the other at the contactor,
mansisthatofJohansson etal. (1982a). They reportedtheir producingincreased sensitivityof NP I. Johansson et al.
resultsasequiresponse contours in whichstimulusintensity (1982a) did notusea surroundin theirexperiments and,asa
washeldconstantat variouslevels,stimulusfrequencyvar- consequence, the sensitivities
that theymeasurefor the RA
ied,andthenumberof impulses percyclerecorded. Theydid populationwould be expectedto be much lessthan that
thisovera stimulus-frequency rangeof 0.5-400 Hz andfor foundpsychophysically. Indeed,Johansson et al. (1982b)
all fourfibertypes.We replottedtheirdataasintensitychar- haveshownthat the presenceof an edgein a stimuluspro-
acteristics,relatingfiring rate to stimulusintensityand by ducedgreaterresponses in RA fibersthan the flat, smooth
interpolating
andextrapolating
theirdatawhennecessary, surfaceusedin their companionstudy (Johansson et al.,
we were then able to constructisoresponse
contoursacross 1982a) on which our analysesare based.
stimulus
frequency.
Choosing
appropriate
response
criteria Unlike NP I, the P channel[seeFig. 7(b) ] is capableof
as discussed below, we then fitted the neurophysiological temporalsummation.Sinceincreasesin stimulusduration
resultsto a four-channel,psychophysical model.The results canlowerdetectionthresholds,the neuralcodeusedby theP
of theanalysisto bedescribedaregivenin Fig. 7; thespecific channelmustincorporatea methodby whichstimulusdura-
neurophysiological-response characteristiccorrelatedwith tion is signaled.A singleimpulse-per-stimulus eventcannot
eachpsychophysically determinedchannel.Eachpanelof dothis.Therefore,at leasttwo impulsesperstimulusmustbe
thefigurerefersto onlyonechannel[ (a), NP I; (b), P; (c), requiredona singlePacinianfiberfor a thresholdeventto be
NP III; and (d), NP II]. perceived.Furthermore,asVan Doren (1985) haspointed
The RA units,andby inferencethe NP I channel[see out, a criterion basedon entrainmentsimilarly cannot be
the Introduction;Lindblom, 1965;Talbot et al., 1968 and usedby the P channel,againbecauseof the constraintof
Fig. 7(a) ], areapparently
ableto signala threshold
event temporalsummation. That is,lengthening thedurationof a
with a singleneuralimpulseon a singlefiber (Henseland stimulusthat isalreadyproducinga thresholdeventsignaled
Bowman,1960;Vallbo andJohansson,1976). This hasbeen by entrainmentwill not lowerthe intensityrequiredfor that
supported byexperiments
usingelectrical
stimulation ofRA entrainment. Lastly, Verrillo (1971) has shown that
fibersby Torebj6rkandOchoa(1980), Vallbo ( 1981), and changesin stimulusrisetimedo not changethresholdlevels.
Ochoaand Torebj6rk (1983). Thus the neuralcodefor a SincePacinianfibersare known to dischargeat higherin-
thresholdeventin the NP I channelmay, in fact, be a single stantaneous ratesfor increases in the onsetvelocityof ramp-
impulse
perstimulus
burst.Thisis consistent
withthefact like stimuli (Iggo and Ogawa, 1977), it is unlikely that a
that NP I doesnot displaytemporalsummation,a phenome- firing-rate code is used.
nonrequiringat leasttwo impulses to demarcatethe dura- Verrillo (1965) hasshownthat the P channelcan signal
tion of a stimulus.The uppercurvein Fig. 7(a) showsthe detection of a 100-Hz stimulus burst of 20-ms duration. The
average (n = 8) frequency response ofRA fibersasreported duration and the rise-fall times that he used allowed for 2
by Johansson et al. (1982a), but usinga criterionof one cyclesof vibrationat 100 Hz with the amountof energyin
impulse/stimulus.The data points on the physiological the stimulus concentrated at 100 Hz. For intense mechanical
curvein thisandthe subjacentpanelsof Fig. 7 areinterpola- stimulation, Bolanowskiand Zwislocki (1984) have shown
tionsand/or extrapolations of the data of Johanssonet al. that the near-maximalfiring rates that can be achievedin
(1982a). The curved lines have been fitted by eye. Also isolatedPaciniancorpusclesin responseto 100-Hz vibra-
shownin Fig. 7 (a) isthefrequency
characteristic
of theNP I tionscorrespondto 2 spikes/stimulus
cycle.Thusa thresh-

1688 J. Acoust.Sec.Am.,Vol.84, No.5, November

1988 Bolanowski,
Jr. et a/•:Mechanical
aspectsoftouch 1688
 8C

6•

4O

2O

80

60
(b)
PC

4C

FIG. 7. Relationshipbetweenphysiologically
2O
measuredfrequencycharacteristics for differ-
ent fibertypes[ (a) RA; (b) PC; (c) SA I; and
0
(d) SAII] and psychophysically obtained
thresholds-frequency characteristics
[ (a) NP I;
(b) P; (c) NP Ill; and (d) NP 11]. Neurophy-
siological datapointsareinterpolations andex-
Pacinian
cor13uscle trapolationsof the averageresultspresented
$ohansson et al. (1982a) for selected
by
response
-40 criteria: (a), 1 impulse/stimulus;(b), 4 im-
8O pulses/stimulus; (c), 0.8 impulses/s,and (d), 5
spikes/s. For the P channel (b), an additional
60
(c) physiologicalcurve (Pacinian corpuscle)de-
rived from unpublished resultsof Bolanowski
( 1981) hasbeenplotted.This curveis theaver-
age (n = 6) responseof excisedPaciniancor-
40
pusclesmaintainedat 33 øCand for a response
criterionof four impulsesoccurringduring the
central 200 ms of a 300-ms stimulus burst.
2O

-20
80'

6C

40

20

N• 11
0

-20
0. 1.0 10 100 1000

FREQUENCY (Hz)

old eventcan be signaledby as little as 4 spikes/stimulus. pulses/stimulus.

The top curvein Fig. 7(b) is a replotof the
Basedon this fact and that at leasttwo impulses/stimulus resultsof Johansson et al. (1982a) usinga four impulse/
are required,we suggestthat the codeusedby the P channel stimulus for average (n = 4) PC responses.The actual
for stimulus detection may be between two and four ira- shapeof the curve is not significantlyaffectedby usinga

1689 d. Acoust. Sec. Am., Vol. 84, No. 5, November 1988 Bolanowski, Jr. et,31.:Mechanical aspects of touch 1689
criterionbasedon eithera lower (2 and 3) or higher( 5 and exhibitstemporalsummation(Verrillo, 1965), a centrally
6) numberof impulses/stimulus, althoughhighernumbers mediatedphenomenon(Zwislocki, 1960). As measuredby
( > 7) do changethe shapeof the curveconsiderably. Also Verrillo ( 1965) andZwislocki(1960), temporalsummation
plottedisthepsychophysical resultof BolanowskiandVer- will imparta 3-dB/octincrease in sensitivity
for increasesin
rillo (1982) showingthe frequency responseof theP chan- stimulusfrequency. Sincethepsychophysical resultsinclude
nelmeasured with a skin-surfacetemperature of 30 øCanda thiscentrallymediatedeffect,thiscurveshouldhavea low-
2.9-cm2 contactorsurface.Basedon previousreportsindi- frequencyslopethat is 3 dB/oct steeperthan that found
catingthat theP channelis mediated by PC afferents (see physiologically for peripheralfibers.The slopeof the psy-
the Introduction),it is not surprisingthat the shapeof the ohophysical curve,ascalculatedbetween20 and 100Hz is
physiologicalandpsychophysical curvesaresimilar.How- -- 12dB/oct.Thelow-frequency
slopeof thePC'curve,
ever, the bestfrequency,overall sensitivity,and low-fre- measured between 20 and 100 Hz, is - 9 dB/oct. On the
quency(10-100 Hz) slopesof the two curvesare not the other hand, the Paciniancorpuscledata showa - 12-dB/
same.The third curve in Fig. 7(b) representsresultsob- octslope.It ispossiblethat themannerin whichtheselatter
tainedfromisolatedPaciniancorpuscles, itssignificancede- data werecollected(i.e., in the steadystate,seeabove)arti-
scribed below. factuallyintroducedan additional - 3-dB/oct slope.We
The differencein thelocationof thebestfrequencies
and note that the differencein the low-frequencyslopesof
possibly
thesensitivity
ofthePCandP curves
ofFig.7(b) physiological and psychophysical data that resultsfrom
maybedueto thefactthatJohansson etal. (1982a)did not temporal
summation
requires
thatthepsychophysical
chan-
controlthe skin-surface
temperaturethat canbe as low as nel exhibittemporalsummation.Temporalsummationis
20 øC for relaxed,sedentaryhuman subjects(Bolanowski not a propertyof theNP I channelandFig. 7 (a) showsthat
and Verrillo, 1982;seealsoHardy and Bard, 1974). Lower thepsychophysical andphysiological
curvesarefairly paral-
temperatures decreaseboththesensitivityandbestfrequen- lel. The NP II channel (see below) doesexhibit temporal
cy of the P channel(Bolanowskiand Verrillo, 1982). The summation, and the physiologicaland psychophysical
curveobtainedfor Paciniancorpuscles and shownin Fig. curves[seeFig. 7(d) ] differby the expectedamount.
7 (b) providesan exampleof this.The curvewasgenerated The NP III channel[seeFig. 7(c) ] is functionaldown
by averagingthe stimulusintensities,
at eachstimulusfre- to verylow frequencies. Sinceseveralinvestigators (Vallbo,
quency,requiredto producea fixed-criterion
responsein six 1981;OchoaandTorebjrrk, 1983) haveshownthat electri-
Paciniancorpuscles isolatedfromcat mesentery andmain- cal activationof SA I fibersproduce"pressure"or a very
tainedat 33 øC (unpublishedresultsof Bolanowski,1981, low-frequency sensation, it is likely that they are the sub-
but seeBolanowskiand Verrillo, 1982). This temperature stratefor NP III. BecauseNP III doesnot displayspatial
waschosen since theactualtemperature ofPacinian corpus- summation,it is likely that the codingusedto signalthresh-
cles,in situ,probablyliesbetween coretemperature (37 øC) old in thischanneloccursona singlefiber.OchoaandToreb-
andthetemperature of theskinsurface, in thiscase,30øC. jrrk (1983), in fact,haveindicatedthat a sensation canbe
Unlike the PC fiber data, the Paciniancorpuscleexperi- produced by activating a singleSA I fiber.Theyalsofound
mentsmonitoredactivityonlyduringthecentral200msof a that severalimpulseswereneededfor thesensation to occur.
300-msburst,that is, in the steadystate.The specificcrite- We replotted theaverage(n = 5) results of Johansson etal.
rionthatweusedto generate theaveragePaciniancorpuscle (1982a) using severaldifferentcriterion responses: im-
curvewasfourimpulses perthe 200-msassessment period. pulses/cycle, impulses/stimulus, andimpulses/s,andcom-
There is fairly goodcorrespondence betweenthe bestfre- paredthemto theshapeof theNP III channelfoundin this
quency ofthepsychophysical andPacinian corpuscle curves study[shownasthedashedlinein Fig. 7(c) ]. The shapeof
andthesensitivity of thePaciniancorpuscle curveisgreater NP III isbasedon thebreakpoints of themaskingfunctions
than that found for the PC fibers. obtainedwith the 0.7-Hz teststimulus(seeFig. 4) and the
Severalother factorscan explainthe differencesin the fiat portionof the overallthreshold/frequency characteris-
sensitivityof the psychophysical andPC fibercurves,aside tic shownin Fig. 1. The bestfit tOthe psychophysical func-
from thefactthat corpuscles canvaryin theiroverallsensi- tion occurswhenan impulse/scodeis used,in thiscase0.8
tivityby20dB ( Bolanowski andZwislocki,1984).Onesuch impulses/s, althoughusinghigherfiringratessimplylowers
factoristhepresence of spatialsummation in theP channel. the overall sensitivityof the physiologicalcurve without
For example,thepsychophysical resultswereobtainedwith changingits shape.The othercriteriaproducefrequency
a contactorsurfaceof 2.9 cm2, which couldsubstantially characteristics that are U shapedwith bestfrequencies near
increasethe sensitivityof this channelover that found in 0.5 Hz, quiteunlikethe shapefoundfor the sensitivityof the
singleunit recordings asa resultof the centralnervoussys- NP III channel.Thus the response criterionof this system
tem integratingresponses from adjacentPaciniancorpus- may be an impulse rate at or higher than 0.8 impulses/s.
cles.A secondpossibility is that Johansson et al. (1982b) Since the P, NP I, and NP III channelsare probably
maynothavestimulatedthemostsensitive regionof anindi- mediated, respectively, by the PC, RA, andSA I fibertypes,
vidual unit's large receptivefield. by elimination NP II [see Fig. 7(d)] must be mediatedby
The otherapparentdiscrepancy betweentheP channel SAII fibers.Since SAII fibers possess spontaneous activity
and the PC fibercurvesshownin Fig. 7(b) is that thereis a (Knibest61,1975;Johansson et al., 1982a), it isunlikelythat
differencein the low-frequency( 10-100 Hz) slopesof the a low firing rate or a smallnumberof impulses/stimulus
two curves. It is well known, however, that the P channel burst could act as the neural code for threshold in this chan-

1690 J. Acoust.Sec.Am.,Vol.84, No.5, November1988 Bolanowski,

Jr. st aL:Mechanicalaspectsof touch 1690
nel.SinceOchoaandTorebjbrk(1983) haveindicatedthat The completepsychophysical modelis givenin Fig. 8
SAII fiberscannotproducea sensation evenwhendrivento andisa composite of theworkpresented here,aswellasthat
ratherhigh repetitionrates,it is possible that the codere- reportedin previouspublications from thislaboratory(see
quirementis severalfibersbeingactivatedat onetime or a discussionin relationto Fig. 7). Figure8 showsfour chan-
statisticalchangein thefiringratedistribution.In reference nelswith partiallyoverlapping sensitivities
that combineto
to the..
latteridea,therepetitivepatternof activationusedby mediatetactileperception. It is importantto realizethatthe
OchoaandTorebjbrk( 1983) maynothavehadsignificance absolutesensitivityat a particularfrequencyas measured
for thischannel.SinceNP II displaystemporalsummation psychophysically will normallybedeterminedby the chan-
(Gescheider etal., 1985),a criterionof oneimpulse/stimu- nel havingthe lowestthreshold,thisbeinga functionof the
lus burstcannotbe the appropriatecode.At the present stimulus conditions:stimulus size, duration, skin-surface
time, it is not possibleto determinewhetherseveralSAII temperature,as well as other factorssuchas body siteand
fibers must be activated at the same time for a threshold staticindentation.This is demonstrated
by replottingthe
eventto take place,sinceit is not known whetherNP II is datapointsof Fig. 1 into Fig. 8.
capableof spatialsummation.Similarly,the resultsof Jo- For the stimulus conditions used here, the P channel
hansson et al. (1982a) do not givethe temporalpatternof determinesthresholdbetweenabout 35 and 500 Hz, which is
activity producedby the SAII fibersin responseto their consistentwith the findingsand predictionsof Verrillo
stimulation.Thus it is impossibleto analyze their results (1963, 1966a, 1968). The sensitivityof this channel is U
usinga stochastic process(e.g., variance)asa criterionre- shapedand its low-frequency slope( - 12 dB/oct) ensures
sponse. It isclearthat, sinceSAII fiberspossess spontaneous that, evenif vibrationsbelow10 Hz are sufficientlyintense,
activity,a substantialincreasein firing abovethis baseline the P channelwill be activated.A secondchannel,NP I, has
mayberequiredfor a thresholdevent.As a firstapproxima- its maximumsensitivity,regardlessof stimulusdurationor
tionin tryingto fit thepsychophysical resultsandin orderto size (see the Introduction), between about 3 and 100 Hz.
overcomethe effectsof spontaneous activity, we selecteda The low-frequencyportion of the NP I characteristicex-
firing rate criterion of 5 spikes/sand plot the average tendsto low frequencies with a slopeof about - 5.0 dB/oct.
( n = 5) resultsin Fig. 7 (d). The othercurve( --- ) showsthe Sincethe NP I channelis much lesssensitiveat vibratory
NP I][ channelasmeasuredpsychophysically by Gescheider frequenciesgreaterthan 35 Hz than is the P channel,it is
et al. (1985; 15-150 Hz) and Verrillo and Bolanowski ineffective in mediatingthresholdin thisrange.NP II, given
( 198(5;100-500 Hz). The correspondence betweenthe two thestimulusconditions of thisstudy,isnotreadilyapparent.
curvesis excellent.Plotsof the SAII response profilesfor Its operating-frequency range,asascertained by Gescheider
response criteriaof impulses/stimulus
or evenimpulses/cy- et al. (1985), is similarto the P channel,althoughits sensi-
cle producefrequency-response characteristics
that do not tivity is considerably less.Of course,if a smallcontactorhad
matclhanyof the psychophysically determinedchannels,in- beenusedexclusivelyin thisstudy,the P channelwouldhave
cludingNP III. beendesensitized to sucha degreethat the overallthreshold-

80' I I I I I I II I I I I I I Ili{ I I I I I I I II I I I I I I

(5O

• 40
E

FIG. 8. The four-channel model of

m ;_•0 vibrotactionshowingthe threshold-
frequencycharacteristics
of the var-
iouschannels:- - -, PC; - - -, NP I; -
ua 0 --, NP II; andZ-, NP III. Seetext
for discussionregarding origin of
data. Data pointssuperimposed on
Gescheider el. al., 1985 the modelare replottedfrom Fig. 1.
_a
o.=
............. Verrillo and Bolanowski,198'6
Bolanowski and Verrillo, 1982
-40

-60
O. 10, 10 100 •oo

FREQUENCY {Hz)

1691 J. Acoust. Soc. Am., Vol. 84, No. 5, November 1988 Bolanowski,Jr. et aL: Mechanical aspects of touch 1691
frequency
characteristic
wouldhaveextended to higherfre- necessary
to establish,acrossall receptortypes,the criterion
quencies
alongtheNP I/NP II curvesasshownby Verrillo response(psychophysically and physiologically) that sig-
andBolanowski(1986). The low-frequency
slopeof the NP nals a sensoryevent in both singleand multiple channels.
II channel(about -- 6 dB/oct) makesit important to note The developmentof multichannelstimulationtechniques
that, like the other channels,NP II canbeactivatedby vibra- may alsobe necessaryto appropriatelyengagethe requisite
tory stimulioccurringat suprathreshold levels.Lastly,NP channels,especiallyfor the purposesof devisingprosthetic
III operates fromverylowfrequencies tofrequenciesgreater devicesassurrogateinputsfor visionandaudition.
than 100 Hz. We havenot, as yet, determinedthe full capa-
bilitiesofNP III, althoughit isclearthat it canbeexcitedby
vibratorystimuliover a considerable rangeof stimulusin-
tensitiesand frequencies.
The significance of the psychophysical modeland its
ACKNOWLEDGMENTS
physiological substrate is that a singlechannelalonemay
signalthethreshold event,depending uponthefrequency of This work was supportedby grantsfrom NINCDS,
stimulation.However,andmoreimportantly,at certainfre- NS09940 and NSF, BNS-8411768 to RTV and NINCDS,
quencies (e.g.,2.0-4.0 and35 Hz), morethanonechannel NS23933 to SJB.The authorswishto thank Dr. R. W. Doty
may be stimulated,even at threshold-levelintensities.In and Dr. C. L. Van Doren for scientific and editorial com-
fact, probabilitysummationbetweenthe channelsmay be ments and G. Lanni and D. Arpajian for technicalassis-
occurringat thosefrequencies wheretwo or morechannels tance.
havethe samethresholdvalue.There are alsoregionsin the
intensity-frequency spaceof Fig. 8 wheremultiplechannels
canbe activatedwith stimulithat are only slightlyabovethe
psychophysical threshold.A typicalexampleof thiswould
bea 35-Hz sinusoidal stimulus,or a morecomplexstimulus
withanenergycomponent at 35 Hz, whichhasanamplitude APPENDIX
greater
than40dBre:1.0-$tm
peak.Because
ofthesubstan-
tial overlapping
ofsensitivities
of thefourchannels,themod- Spectralanalysisshowedthat the energyspectrumof
elhasconsiderableimportance fortheunderstanding ofboth the test-stimuluswaveformis approximatedby
physiologicalandpsychophysical resultsobtained
withnon-
F(n/T)={cos[O.407•(n--1)/T]
sinusoidalstimuli having broad-frequencyspectra (e.g.,
pulse,ramp,noise).Thesecomplexstimuliwouldgenerally --cos[l.407•(n--1)/T]}/n•--2n+l
activateseveralchannelsat the sametime, makingit impor-
tant to take into accounteach channel'sresponsewhen at- +{cos[0.407•(n+ I)/T]
temptingto explainperceptual phenomena in termsof phys-
iologicalmechanisms. Clearly,mechanical stimuliusedin -- cos[1.407v(n+ 1)/T]}/n • + 2n + 1,
physiological experiments wherepsychophysical responses
arenotobtainedmaybeactivatinganensemble of units,thus where T is the periodof the fundamentalfrequencyand
preventing theexperimenter fromtrulyascertaining theex- -- m < n < o•. The first, second,third, and fourth harmonic
act codeusedat the somatosensory periphery.Discussions frequencycomponents
haveamplitudes,respectively,
that
of experiments dealingwith suprathreshold stimuluscondi- are: 22.7, 34, 41.8, and 54 dB smaller than that of the funda-
tionsmustconsiderthe possibility that perceptualqualities mental, indicatingthat the resultsare not affectedby the
with anintensive component(e.g.,itch,tickle,texture)may high-frequencycomponents of thestimulus. For thesubhar-
be a resultof all four channelsbeingstimulatedat the same monics,however,a substantial0-Hz componentcan be in-
time andnot merelythe resultof activityin a singlechannel troducedas the carrier frequency(i.e., test-stimulus fre-
supposedly encodingintensity(Werner and Mountcastle, quency) approachesvery low frequencies.The spectral
1965;Mountcastle,1967;HarringtonandMerzenich,1970; analysisindicatedthat,for thefrequencies usedin thestudy,
Krugerand Kenton,1973;but see,Knibest61 and Vallbo, theharmoniccomponents belowthefundamental frequency
1980;Meiet al., 1983;Johnson,1983;Pouloset al., 1984). did nothavean amplitudecomponent greaterthanthefun-
Lastly,the modelsuggests that, at suprathreshold lev- damental.For example,at 0.7 Hz, the largestsuchcompo-
els, the code for perceptualquality may be considerably nent occurredat 0 Hz and was4 dB lower in amplitudethan
more sophisticated than had beenpreviouslyrealized,re- the fundamental.Thus detectionby the observerwas to the
quiringthatseveralchannels contribute information forthis fundamentalfrequencyand not confoundedby subhar-
purpose.In otherwords,fundamental qualitieslike "pres- monicfrequencycomponents spuriously introducedby the
sure," "flutter," and "vibration" may combineto form the stimulusparameters.Additional assurancethat stimulus
many sensoryattributesascribedto the somatosensory sys- artifact wasnot responsible for the resultswasprovidedby
tem.One implicationof the four-channel modelis that, be- the use of the 2500-ms test stimulus that substantiallyde-
fore a true understanding of the mannerin whichsensory creasesthe 0-Hz componentof the spectrum.The results
experiences
suchas"roughness," or "intensity" obtained were the same as those found
"softness," for the 700-ms test
(amongmyriadotherattributes)canbeachieved,it maybe stimulus (see Sec. II).

1692 J. Acoust.Soc. Am., Vol. 84, No. 5, November1988 Bolanowski,Jr. otaL: Mechanicalaspectsof touch 1692
B•k•sy,
G.,yon(1939).
"Ober
dieVibrationsempfindung,"Akust.
Zeits.
4, discrimination," Fed. Proc. 42, 2542-2547.
316-334. Johnson,K. O. (1987). Personal communication.
Bolanowski, S. J., Jr. (1981). "Intensityandfrequency characteristics
of Knibestbl,M. (1975). "Stimulus-responsefunctionsof slowly adapting
Paciniancorpuscles," Ph.D. dissertation
andSpecialRep.,ISR-S-20,In- mechanoreceptors in the human glabrousskin area," J. Physiol. (Lon-
stitutefor Sensory Research, SyracuseUniversity,Syracuse,NY. don) 245, 63-80.
Bolanowski, S.J.,Jr.,andVerrillo,R. T. (1979). "Effectsofresponse
crite- Knibest61,M., and Vallbo, A. B. (1980). "Intensityof sensationrelatedto
rion on the frequency characteristics
of a mechanoreceptor (Pacinian activityof slowlyadaptingmeehanoreceptive
unitsin thehumanhand,"
corlmscle),"J. Acoust.Soc.Am. Suppl.I 65, SI 1. J. Physiol.(London) 300, 251-267.
Bolanowski, S.J.,Jr.,andVerfillo,R. T. (1982). "Temperature andcrite- Kruger, L., and Kenton,B. (1973). "Quantitativeneuraland psycho-
rioneffectsina somatosensory subsystem:A neurophysiological
andpsy- physicaldatafor cutaneous
mechanoreceptorfunction,"Brain.Res.49,
ohophysical study,"J. Neurophysiol.48, 836-855. 1-24.

Bolanowski,
S.J.,Jr.,andZwislocki,J. J. (1984). "Intensityandfrequency Labs,S. M., Gescheider,G. A., Fay, R. R., and Lyons,C. H. (1978). "Psy-
characteristics
of Paciniancorpuscles. I. Actionpotentials,"J. Neuro- chophysicaltuningcurvesin vihrotaction,"Sen.Process.2, 231-247.
physiol.51, 793-811. Lindbiota,U. (1965). "Propertiesof touchreceptors in distalglabrousskin
Capraro, A. J.,Verrillo,R.T., andZwislocki,J.J. (1979)."Psychophysical of the monkey,"J. Neurophysiol.28, 966-985.
evidencefor a triplex systemof mechanoreception," Sens.Process.3, Lindblom, U, and Lurid, L. (1966). "The dischargefrom vibrationsensi-
334--352. tive receptorsin the monkeyfoot," Exp. Neurol. 15, 401-417.
Craig, J. C. (1968). "Vibrotactile spatial summation," Percept. Psy- Mei, J., Tuckerr, R. P., Poulos, D. A., Horch, K. W., Wei, J. Y., and Bur-
chophys.4, 351-354. gess,P. R. (1983). "The neural signalfor skin indentationdepth. If.
Dykes,R. W. (1983). "Parallelprocessing of somatosensory information: Steadyindentations,"J. Neurosci.3, 2652-2659.
A theory," Brain Res. Rev. 6, 47-115. Merzenich,M. M, andHarrington,T. (1969). "The senseof flutter-vibra-
Freeman,A. W., andJohnson, K. O. (1982a). "Cutaneous mechanorecep- tion evokedby stimulationof the hairy skinof primates:Comparisonsof
totsin macaque monkey:Temporaldischarge patternsevokedby vibra- human sensorycapacitywith the responses of mechanoreceptive affer-
tion, anda receptormodel,"J. Physiol.323, 21-41. entsinnervatingthe hairy skinof monkeys,"Exp. Brain Res.9, 236-260.
Freeman,A. W., and Johnson,K. O. (1982b). "A modelaccountingfor Mountcastle,V. B. (1967). "The problemof sensing andthe neuralcoding
effectsof vibratoryamplitudeon responses of cutaneous mechanorecep- of sensoryevents,"in The Neuosciences, editedby G. C. Quarton,T.
tots in macaquemonkey,"J. Physiol.323, 43-64. Melnechuck,and F. O. Schmidt(Rockefeller,New York), pp. 393-408.
Gescheider,G. A. (1976). "Evidencein supportof the duplextheoryof Mountcastle, V. B., LaMotte, R. H., and Carli, G. (1972). "Detection
mechanoreception," Sens.Process.1, 68-76. thresholdsfor stimuliin humansand monkeys:Comparisonwith thresh-
Gescheider, G. A., Frisina, R. D., and Verrillo, R. T. (1979). "Selective oldeventsin mechanoreceptive afferentnervefibersinnervatingthemon-
adaptationof vibrotactilethresholds,"Sen.Process. 3, 37-48. key hand," J. Neurophysiol.35, 122-136.
Gescheider,G. A., O'Malley, M. J., and Verrillo, R. T. (1983). "Vibrotac- Ochoa, J. L., and Torebj6rk, H. E. (1983). "Sensationsevokedby intra-
tile forwardmasking:Evidencefor channelindependence," J. Acoust. neuralmicrostimulation of singlemechanoreceptor unitsinnervatingthe
Soc. Am. 74, 474-485. humanhand,"J. Physiol.342, 633-653.
Gescheider, G. A., Sklar, B. F., Van Doren, C. L., and Verrillo, R. T. Pertovaara, A., and H•im•il•iinen, H. (1981). "Vibrotactile thresholds in
(1985). "Vibrotactileforwardmasking:Psychophysical
evidencefor a non-Pacinianafferents:the importanceof temporalparameters,"Acta
tripltextheoryof cutaneous
mechanoreception,"
J. Acoust.Soc.Am. 74, PhysioLScand.113, 519-522.
534--543. Poulos, D. A., Mei, J., Horch, K. W., Tuckett, R. P., Wei, J. Y., Cornwall,
Gescheider,G. A., Verrillo, R. T., and Van Doren, C. L. (1982). "Predic- M. C., and Burgess,P. R. (1984). "The neuralsignalfor the intensityof a
tionsof vibrotactilemaskingfunctions,"J. Acoust.Soc.Am. 72, 1421- tactile stimulus," J. Neurosci. 4, 2016-2024.
1426. Talbot, W. H., Darian-Smith, I., Kornhuber, H. H., and Mountcastle, V. B.
Hamer,R. D. (1979). "Vibrotactilemasking:Evidencefor a peripheralen- (1968). "The senseof flutter-vibration:
Comparisonof thehumancapac-
ergy'threshold,"Ph.D. dissertationand SpecialRep., ISR-S-18,Institute ity with response
patternsof mechanoreceptive afferentsfrom the mon-
for .%nsoryResearch,SyracuseUniversity,Syracuse,NY. key hand,"J. Neurophysiol.31, 301-334.
Hamer, R. D., Verrillo, R. T., and Zwislocki, J. J. (1983). "Vibrotactile Tapper,D. N., Galera-Garcia,C., and Brown,P. (1972). "Sinusoidalme-
maskingof Pacinianand non-Pacinianchannels,"J. Acoust.Soc.Am. chanicalstimulationof the tactilepad receptor:tuningcurves,"Brain
73, 1293-1303. Res. 36, 223-227.
Hardy, J. D., and Bard, P. (1974). "Body temperatureregulation,"in Me- Torebj6rk, H. E., and Ochoa,J. L. (1980). "Specificsensations
evokedby
dicalPhysiology, editedby V. B. Mountcastle(Mosby,St. Louis,MO), activityin singleidentifiedsensoryunitsin man," Acta. Physiol.Searid.
pp. 1305-1342. 110, 445-447.
Harrington,T., andMetzenith,M. M. (1970). "Neuralcodingin thesense Vallbo, A. B. (1981). "Sensationsevokedfrom the glabrousskin of the
of touch:Human sensations of skin indentationcomparedwith the re- humanhand by electricalstimulationof unitary mechanosensitive
affer-
spoasesof slowly adaptingmechanoreceptive afferentsinnervatingthe ents," Brain Res. 215, 359-363.
hairy skinof monkeys,"Exp. Brain Res.10, 251-264. Vallbo,A. B., andJohansson, R. S. (1976). "Skinmechanoreceptors in the
Hensell,H., and Bowman,K. A. (1960). "Afferentimpulsesin cutaneous humanhand: Neural and psychophysical thresholds,"in SensoryFunc-
nervesin humansubjects," J. Neurophysiol.23, 564-578. tionsof theSkin of Primates,withSpecialReference toMan, editedby Y.
lggo, A., and Ogawa,H. (1977). "Correlativephysiologicaland morpho- Zotterman (Pergamon,New York), pp. 185-198.
logicalstudiesof rapidlyadaptingmechanoreceptors in cat's glabrous Vallbo, A. B., and Johansson,R. S. (1984). "Propertiesof cutaneousme-
skin," J. Physiol.266, 275-296. chanoreceptors in the humanhand relatedto touchsensation,"Human
Johansson, R. S. (1976). "Receptivesensitivityprofileof m½chanosensitive Neurobiol. 3, 3-14.
unitsinnervatingtheglabrousskinof the humanhand,"BrainRes.104, Van Dom, C. L. (1985). "Temporal summationby Paciniancorpuscles
330--334. precludes
entrainmentat the detectionthreshold,"J. Acoust.Soc.Am.
Johansson,R. S. (1978). "Tactilesensibilityof the humanhand:receptive 77, 2188-2189.
fieldcharacteristics
of mechanoreceptive unitsin the glabrousskin," J. Verrillo, R. T. (1962). "Investigationof someparametersof the cutaneous
Physiol. 281, 101-123. threshold for vibration," J. Acoust Soc. Am. 34, 1768-1773.
Johansson,
R. S.,Landstr6m,U., andLundstr6m,R. (1982a). "Responses Verrillo, R. T. (1963). "Effect of contactor area on the vibrotactile thresh-
of mechanoreceptive
afferentunits in the glabrousskin of the human old," J. Acoust. Soc. Am. 35, 1962-1966.
handto sinusoidal
skindisplacements,"
BrainRes.244, 17-25. Verrillo, R. T. (196S). "Temporalsummationin vibrotactilesensitivity,"J.
Johansson,
R. S., Landst6m,U., and Lundstr6m,R. (1982b). "Sensitivity Acoust. Soc. Am. 37, 843-846.
to edges
of mechanoreeeptive
afferentunitsinnervating
theglabrous
skin Verrillo, R. T. (1966a). "Vibrotactilesensitivityand the frequencyre-
of the human hand," Brain Res. 244, 27-32. sponseof the Paciniancorpuscle,"Psychon.Sci. 4, 135-136.
Johansson,R. S., and Vallbo, A. B. (1979). "Tactile sensibilityin the hu- Verrillo, R. T. (1966b). "Effectsof spatialparameterson the vibrotactile
manhand:Relativeandabsolutedensities of fourtypesof mechanorecep- threshold,"J. Exp. Psychol.71, 570-574.
riveunitsin glabrousskin,"J. Physiol.(London) 286, 283-300. Verrillo, R. T. (1966e). "Vibrotactile thresholdsfor hairy skin," J. Exp.
Johnson,K. O. (1983). "Neural mechanismsof tactual form and texture Psychol.72, 47-50.

1693 J. Acoust. Sec. Am., Vol. 84, No. 5, November 1988 Bolanowski,J r. otal.: Mechanical aspects of touch 1693
Verrillo, R. T. (1968). "A duplexmechanism of mechanoreception,"in Verrillo,R. T., andCapraro,A. J. (1974). "Effectsof simultaneous
audi-
TheSkin Senses, editedby D. R. Kenshalo(Thomas,Springfield,IL), tory stimulationon vibrotactilethresholds,"Percept.Psychophys.16,
pp. 139-159. 597-600.
Verrillo, R. T. (1971). "The effectof risetime on vibrotactilethresholds," Verrillo, R. T., andGescheider,
G. A. (1977). "Effectsof priorstimulation
Psychon.Sci. 25, 199-200. on vibrotactile thresholds," Sen. Process.1, 292-300.
Verrillo, R. T. (1975). "Cutaneoussensation,"in ExperimentalSensory Werner, G., and Mountcastle,V. B. (1965). "Neural activityin mechano-
Psychology,
editedby B. Scharf(Scott,Foresman,Glenview,IL), pp. receptivecutaneous afferents:
Stimulusresponserelations,Weberfunc-
159-184. tions,and informationtransmission,"J. Neurophysiol.28, 359-397.
Verrillo, R. T. (1979). "The effectsof surfacegradientson vibrotactile Zwislocki,J. J. (1960). "Theory of temporalauditory summation,"J.
thresholds," Sen. Process.3, 27-36. Acoust. Soc. Am. 32, 1046-1060.
Verrillo, R. T., and Bolanowski,S. J., Jr. (1986). "The effectsof skin tem- Zwislocki,
J.J.,Maire,
F.,Feldman,
A.S.,andRubin,
H.J.(1958)."Onthe
peratureon the psychophysical
responses
to vibrationon glabrous
and effectof practiceand motivationon the thresholdof audibility,"J.
hairy skin,"J. Acoust.Soc.Am. 80, 528-532. Acoust. Soc. Am. 30, 254-262.

1694 J. Acoust.Soc. Am., Vol. 84, No. 5, November1988 Bolanowski,Jr. eta/.: Mechanicalaspectsof touch 1694

View publication stats