Acupuncture, Pain, and Signal Detection Theory

Clark and Yang (1) have underscored an

important aspect of the analgesic effectsof acupuncture-that it tends to generatepositive attitudinal shifts which lead theexperimental subject to report pain lessreadily. However, their study miay haveprovided an inadequate evalua tion of theeffects of acupuncture on sensory sensitiv-ity to painful radiant heat stiniulation, andthree aspects of their report warrant con-

sideration.First, the available Chinese literature in-

dicates that little or no analgesic eflectshould be evident in either of the experi-mental conditions einployed by Clark andYang. A test on acupuncture anesthesiastates that 15 to 20 minutes of electricalacupuncture stiiulation are required forthe development ol' analgesia, and an in-

duction period of this length precedes sur-

gery (2). Furthermore, Peking investiga-tors using potassium iontophoresis do-lorimetry reported that cutaneous pain

thresholds followed a negatively accelera-ted growth function as electrical acupunc-

ture stimulation progressed, becomingstable only after about 50 minutes (3).Thresholds began to drop rapidly after

needles were removed. Although thresholdstudies can be misleading, as Clark andYang noted, the Chinese report does sug-

gest that Clark and Yang did not allowsufficient time for an analgesic efTfct to de-velop. Their subjects were tested during an

acupuncture stimulation period of 15 to 20minutes and again after needles were re-

moved.Second, Clark and Yang picked a difli-

cult areal in which to demonstrate acupunc-

ture ancalgesia. One of us (C.R.C.) was in-

formed, while visiting Chinese operating

theaters, that acupuncture analgesia is

most often successful for surgery of thehead and trunk, while efTects are less reli-able on the extremities. Kaada et al. (4) ob-tained the sanie information.

Third, sensory sensitivity (d') values in

each treatment conditionl were calculatedon the basis of only 12 stimuli at each ofthe two test intensities. The use ol a sm.Illnumber of trials reduces the precision of

signal detection indices as well as their reli-ability. While the mean values reportedwere remarkably consistent across ana-

tomical test sites, the possibility exists thata difference might have been missed be-cause so few trials were used.We examined the aLnalgesic efIects ol bi-

lateral acupuncture at a single locus on thedorsal aspect of the hand in 42 volunteers,comparing its effects to those of nitrousoxide (5). High, medium, low, and zero in-

tensity electrical stimuli were delivered to

4 JULY 197i

the central incisors in randoin order, with75 trials at each level, and subjects gaverating scale responses to indicate the paLinexperienced with each stimulus in bothbaseline and test sessions. Electrical stiniu-lation of the needles for acupuncture sub-jects began 20 minutes before testing andcontinued lor the rest of the session. Ana-lyzing the data by signal detection meth-ods, we observed that controls improved instimulus detection and discrimination over

two sessions, while acupuncture subjectsshowed significant decreases in d' betweenbaseline and test sessions, as did subjectsgiven 33 percent nitrous oxide. In addition,significant changes in response bias, sim-

ilar to those reported by Clark and Yang(1), were observed.

Acupuncturc generates sensory analge-sic effects for experimental dental pain, butsensory changes are of small magnitudeand many trials are required to demon-

strate them. Whether acupunciure afTectscutaneous sensitivity to radiant heat stimu-lation has not been conclusively answeredby the Clark and Yang report.

C. RICHARD CHAPMANDepartments oJ Anesthesiology,Psychiatry and Behavioral Sciences, andPsyc hology, University of WashingtonSchool ofMedicine, Seattle 98195

JOHN D. GEHRIGDepartment of Oral Surgery, University\of Washington School ofMedicine

MICHAEL. E. WIL SONDepartment of A nesthesiology, Universityof Washington School of Medicine

References

1.W. C. Clirk aind J. C. Yang, Science 184, 1096(1974).

2. Board of Acupuncture Anesthesias Acupunt lure.Anesthesia (in Chinese) (People's. Publishing Soci-ety, Shanghai, 1972).

3. Research Group of Acupuncture Anesthesia, Pe-king College, Chin. Med. J. 3, 5I (I1973).

4. B. KaadaL, I. Hoel, K I eseth, B. Nyg.iard-Ostby,.1. Setekleiv,.1. Stovuer, TidiKr. \orINr.L ge/oren94, 417 (1974).

5. C. R. C'hapimiaiz, .1. D. Gehrig, M. E. Wilson,Anesthe.viologic in press.

8 August 1974

Clark and Yang (1) report that acupunc-

ture treatment did not alter the physicaldiscriminability (d') of the two radiantheat stimuli presented, but did result in a

reduecd willingness of subjects to reportpain (that is, a change in response bias to a

stricter criterion) and hence "did not pro-

duce acupunctural analgesia." We feel thattheir conclusions regarding the effects of'

acupuncture are unjustified because of fail-

ure to meet the basic assumptions of thetheory of signal detection (TS D).

Specifically, Clark and Yang f;ail to in-

elude an evaluation of subjects' responding

in the presence of a clearly nonnoxiousstimulus. As Swets (2) points out, "Weimagine signal detection to be a choice be-tween two Gaussian variables. One, ha'vinga mean equal to zero, is associated withnoise alone; the other, having a mean equalto d', is associated with signal plus noise."Traditional applications oltTSD allow ob-jective measurement of both the natureand physical presence of signal (S) andnoise (N) independent of subjects' reports.With pain, however, objective verificationis not possible. Any stimulus that elicits re-ports of pain must be presumed paintul,and its use would not, theretore, allow anestimation of the N distribution necessaryto evaluate changes in pain sensitivity re-flected by d'. To avoid the logical con-tradiction inherent in labeling as N a stim-ulus that elicits a pain response, one mustdevise a means of measuring responding inthe objectively veritiable absence ofnoxious stimulation (for example, no stim-ulation at all). While Clark (3) makes thequestionable contention that stimuli pro-ducing the perception of warmth may beemployed for estimating the N distribu-tion, data included in that same papershow a 370 mcal sec-I cm 2 stimulus (thesame stimulus treated as N in the acupunc-ture study) to be reported as painful 87percent of the time. Thus, Clark and Yang,by comparing two radiant heat stimuli ofdifferent intensities without an appropriateN comparison, have violated the proce-dural requisites of TSD.

Although Clark and Yang interprettheir failure to find a change in d' as an ab-sence of analgesia, it seems just as likely toexpect analgesia to result in a decrease inpain resporises to all noxious stimuli, in-cluding those in their experirnent asso-ciated with the lower stimnulus intensity of370 mcal sec cm 2. At a theoretical level,analgesia could result in the distributionsof both noxious stimuli shilting aboutequally to the left along the decision axis.This hypothesis would predict that anal-gesia need not always be attended bychanges in d' between stimuli. In the ab-sence ot an N evaluation, such shifts wouldbe misconstrued as a change in responsebias to a stricter criterion. Empirical sup-port for this latter interpretation is foundin a report by Chapman et al. (4), whodemonstrated that 33 percent nitrous oxidedid reduce d' for radiant heat stimuli whencompared to a zero intensity condition butdid not alter d' for adjacent nonzero stimu-lus pairs quite similar to those of Clarkand Yang. Consequently, it could be ar-gued that Clark and Yang's results are notinconsistent with the induction of "true"analgesia, and their omission of an ap-propriate N estimate led them to interpret

65

on February 1, 2019

http://science.sciencem

ag.org/D

ownloaded from

distribution shifts as changes in responsebias.

In conclusion, we maintain that pub-lished applications of TSD to analgesia re-search have produced, at best, equivocalresults. Clark and Yang (1) and Chapmanet al. (4), in the presence of similar data,have drawn opposite conclusions. Evcn cx-plicit operational recognition of the as-sumptions of the model still poses seriousquestions about its applicability too de-tailed to be considered here. Without suchoperations, however, we feel results cannotbe interpreted within the context of TSD.

RONALD L. HAYESGARY J. BENNETTDAVID J. MAYER

Department of PhysiologyvMedical College of Virginia.Virginia Commonwealth UniversityvRichmond 23298

References and NotesI. W. C. Clark and J. C. Yang, Science 184, 1096

(1974).2. J. A. Swets, in Signal Detection and Recognition

by Huhman Observeri. Contemporary Readings, J.A. Swets, Ed. (Wiley, New York, 1964), p. 51.

3. W. C. Clark, Anesthesiologjr 40, 272 (1974).4. C. R. Chapnman, T. M. Murphy, S. H. Butler, Sci-

ence 179, 1246 (1973).5. Supported by PHS grant OA 00576-01.

28 June 1974

Clark and Yang (I) present an ingeniousstudy of acupuncture analgesia but intrying to avoid certain problems in the def-inition of pain they have measured thewrong type of sensitivity. As they pointout,. the measurement of pain is com-plicated by the interplay of sensory and at-titudinal variables. Signal detection theoryallows one to separate these variables ex-perimentally, provided one can present twostimuli that, by definition, differ with re-spect to some variable, such as the pres-ence or absence of a light. The stimulusthat contains the light is called the signaland the one that does not is called noise.This example wouid allow the measure-ment of the absolute sensitivity to the light.Instead of absolute sensitivity, one mightbe interested in the differential sensitivity,the ability to discriminate two intensitiesof light from one another. In that case, themore intense light would be the signal andthe less intense, the noise.As Clark and Yang realize, pain is prob-

lematic because it is not possible to presenta stimulus that is either painful or is notperceived at all. If heat is used as the painstimulus, an intensity too weak to evokepain will produce the sensation of heat. Soone cannot study the absolute sensitivityto pain, per se, using signal detection the-ory, because the independent variable isheat, not pain.

Clark and Yang try to get around the

66

problem in the following way: two stimuliare presented that vary in intensity and thesubject is instructed to use different labelsto describe them. They analyze the differ-ences in the way labels are attributed to thepair of stimuli. This is a legitimate proce-dure in signal detection theory but it yieldsa measure of differential sensitivity to vari-ous stimuli that are warm or painful (orboth), not the absolute sensitivity to pain.

If the measurement of differential sensi-tivity to thermal stimuli is the best that canbe done, why am I objecting? Two reasons.First, absolute sensitivity and differentialsensitivity are not two aspects of a singlesensory capability. It is not uncommon fordifferential sensitivity to increase and ab-solute sensitivity to decrease as a result ofthe same manipulation. Two well-knownexamples are sensory adaptation (2) andphysical damage to a receptor (3). Adapta-tion is often thought of as simply a de-crease in sensitivity of a system to somestimulus as a result of steady presentationof the stimulus. However, while absolutesensitivity has decreased, differential sensi-tivity has increased. Adaptation has beenlikened to a device that adjusts the range ofmaximum sensitivity of a sensory systemto prevailing conditions. In sensory-neuraldeafness the threshold is elevated but theear is more sensitive to variations in loud-ness, as revealed in the slope of the psycho-metric function at threshold and the rate ofgrowth of loudness just above threshold.The fact that acupuncture produced nochange in differential sensitivity to thermalstimuli does not imply that there might nothave been a decrease in absolute sensitivityto pain.

Second, and more fundamental, becausethey must settle for measuring sensitivityto warmth or pain (or both) rather thanpain per se, they confuse the ability tosense a stimulus with the quality of the sen-sation.

Since the requirements of signal detec-tion theory have not been met, it is fruitlessto discuss whether changes occurred insensitivity (d') or criterion (Lx). Unfortu-nately, it does not seem possible to meetthose requirements in principle.

DONALD H. MCBURNEYDepartnment ofPsychology,University of Pittsburgh,Pittsburgh, Pennsylvania 15260

References

1. W. C. Clark and .1. C. Yang, Science 184, 1096(1974).

2. W. D. Keidel, U. 0. Keidel, M. E. Wigand, in Sen-sorY Commiunication, W. A. Rosenblith, Ed. (M ITPress, Cambridge, Mass., 1961), pp. 319 338.

3. G. von Bekesy, A cta Oto-Larvngol. 35, 411 (1947);L. D. Hedgecock, A.M.A. Arch. Otolarvngol. 66,515 (1955); J. Jerger, J. L. Shedd, E. Harford, ibid.69, 200(1959).

I1 June 1974

The above comments are welcome, espe-cially since they serve to focus attention onthe value of signal detection theory, which,unlike the traditional psychophysicalmethods, successfully distinguishes be-tween the sensory and response bias com-ponents of the classical perceptual thresh-old (1). Pain is an extremely complex per-ceptual experience in which sensory andpsychological factors are closely inter-laced. Signal detection theory specificallyaddresses itself to such problems. The pro-cedure may not be perfect, but it is cer-tainly the best model presently availablefor the investigation of experimentallycontrived pain. If nothing else, signal de-tection theory has made us aware that adecrease in pain report does not necessar-ily indicate an amelioration of the pain ex-perience.We cannot agree with the statement of

Chapman et al. (2) that "the available Chi-nese literature indicates that little or noanalgesic effect should be evident" underthe conditions used in our study. With re-spect to duration of stimulation, ourChinese (Shanghai) manual AcupunctureAnesthesia (3) states that, for surgery ofthe forearm, "the acupuncture anesthesiainduction time should be around 20 min-utes." In our study (4), no analgesia wasevident in any of our subjects even thoughthe duration of acupunctural stimulationfor all subjects except one ranged from 20to 28 minutes (mean, 24.7) (one was stimu-lated for 46 minutes). The work of the Pe-king group (5) cited by Chapman gives thetime course of pain threshold changes invarious parts of the body following acu-puncture at the Ho-Ku or lsu-San-Li sites(or both). The results of this excellentstudy do not contradict the instructiohs ofour manual. The Peking group used one ortwo acupuncture sites and studied thresh-old changes in eight sites in the body, notincluding the forearm. In contrast, we usedsix acupuncture sites specifically pre-scribed for forearm surgery. Furthermore,their data on the leg (the only limb studied)indicates that two-thirds of the maximum(50-minute) analgesic effect was obtainedin 25 minutes, the same duration as ourstimulation period. Instead, we obtainedabsolutely no analgesic (lower d') effect.

In a recent symposium (6), durations of8 to 30 minutes (7) were reported as beingsufficient to produce fewer pain reports(so-called "analgesia"). Man and Baragar(8) reported an effect in I minute, andChang (9) found that acupuncture tookonly a few seconds to stop the character-istic unit discharges found when the nu-cleus centralis lateralis of the thalamus isresponding to noxious stimulation, in thisinstance a surgical wound.

SCIENCE, VOL. 189

on February 1, 2019

http://science.sciencem

ag.org/D

ownloaded from

With respect to the persistence of theanalgesia following removal of the needles,our manual (3) states: "When the stimu-lation of the surgery is not too strong youcan stop manipulating the needle or stopthe electric current.... Whenever the stim-ulation of surgery becomes too strong, youmust twist or stimulate the needle to keepthe effect." Since the thermal experimentalpain, from which the subject could with-draw if it became too painful, is innocuousrelative to surgical pain, we believe that theanalgesia could be expected to persist forsome time. The Peking group (5) reporteda decrease in analgesia almost immedi-ately following the removal of the needles;however, even 10 minutes later about two-thirds of the analgesic effect remained inthe limb. However, in the period after acu-puncture we obtained neither a thresholdeffect (fewer pain reports) nor a decrease ind' or L,. Again, there is very little agree-ment about duration of analgesia (10).Chapman et al. (2) suggest that it is eas-

ier to demonstrate acupunctural analgesiain the head and trunk than in the ex-tremities. Perhaps, but the manual (3)claimed that analgesia sufficient to permitsurgery could be induced in the forearm.Furthermore, in figure 2 of the reference(5) cited by Chapman et al. (2), the degreeof analgesia in the head is shown to beequal to that found in the ipsilateral leg,and even less than that found in the con-tralateral leg.

In sum, there is no compelling evidencethat our failure to find analgesia was due todeparture from accepted practice. Al-though it might prove effective to uselonger stimulation times, as Chapman sug-gests, our conclusion is that the intensity ofacupunctural stimulation should be thefirst candidate for future investigation. Inour experiment (4) we increased the in-tensity of electrical stimulation until thesubject refused to bear any more pain. In-deed, physical signs such as grimacing andclenching were occasionally present. How-ever, success in China could be due to anethnoculturally induced willingness to ac-cept considerably higher intensities of acu-punctural stimulation. Also, the use of pre-operative analgesics and sedatives appearsto be common in China (11), and wouldpermit the administration of still higher in-tensities of acupunctural stimulation. Suchintensities might induce a true analgesia(lower d') in addition to the raised pain re-port criterion (Lx) which we found. Per-haps the positive results reported byChapman for dental pain were due tohigher intensities of stimulation.

In the more readily specified realm ofpsychophysics, Chapman et al. (2) cor-rectly note that when there are too few4 JULY 1975

Table 1. Various meanings of absolute and differential sensitivity; N, noise; S, intensity; the sub-scripts indicate the amount of activity in the sensory system; versus indicates the physical statesbeing discriminated. N, is used rather than No to indicate that spontaneous neural activity is presenteven in the absence of physical stimulation.

Discrimination

Background Absolute sensitivity Differential sensitivityBlank = 0 Blank > 0

Absolute sensitivity A BSurround = 0 N versus S, + N, S, + N, versus 5 + N,

Differential sensitivity C DSurround > 0 N2versusS, + N2 S, + N2versusS2 + N2

stimulus presentations the resulting esti-mates of d' and Lx may be highly variable,with the result that statistical tests maylack power to detect differences. We wouldprefer to use more than 12 stimuli per in-tensity; however, one runs into the problemeither of tiring the subject or of omittingessential controls such as the simultaneoustesting of a nonacupunctured area. In ourstudy (4) (which included four other in-tensities in addition to the two mentionedin the report), each subject judged at least432 stimuli in a period of about 2 hours.This approaches the edge of endurance ofmost observers. lp any event, 12 presenta-tions per intensity appear to be sufficient.First, as Chapman recognized, the meanvalues of d' were remarkably uniform atboth sites and under all conditions. Thisuniformity makes it unlikely that differ-ences existed which were not detected be-cause tests lacked power. Second, it seemsunlikely that the tests used could be pow-erful enough to detect differences in Lx butnot powerful enough to detect differencesin d'. Third, in earlier studies, Clark andco-workers have used as many as 25 stim-uli (12) and as few as 12 (13) without en-countering any marked change in variabil-ity. Finally, using fewer than 12 stimuli perintensity we found a dose-related decreasein d' following a median nerve block withdilute Carbocaine. If acupunctural anal-gesia had even begun to approach the ef-fectiveness of a drug-induced peripheralnerve block, it would have been apparentin our data. We agree that the effect of thenumber of presentations on the standarderror of d' and Lx deserves investigation.To this end we have presented the standarderrors of the mean; we hope that other in-vestigators will do likewise.

Hayes el al. (14) have drawn the atten-tion of pain researchers to the importanceof using nonnoxious stimuli including zerointensity. As they point out, if the d's arenot ultimately anchored to zero intensity,it becomes possible for the distributions toshift to the left along the decision axiswithout a change in d' becoming apparent,and the effect of an analgesic might bemissed. In fact, our present study (4), like

our earlier ones (12, 13), did include stimu-lus intensities of 0, 120, 240, and 305 mcalsec-' cm-' as well as the two higher in-tensities of 370 and 420 mcal sec cm 2;data at the lower intensities were omittedfrom the Science report because of spacerestrictions. We found no significant ef-fects for treatment (acupunctured versuscontrol arm) or for period (before, during,or after acupuncture) for any of the otherintensity pairs. Furthermore, in the criticaltest for the presence of analgesia, the treat-ment by period by intensity interaction, d'failed to approach significance (F = 0.97;d.f. = 8, 80; P < .25). Contrary to Chap-man et al. (15), who found that nitrous ox-ide reduced d' at low heat intensities, we

found no effect with acupuncture at any in-tensity.

Although the applicability of signal de-tection theory to pain has not been conclu-sively established, no other viable alterna-tive exists. Studies of analgesia by tradi-tional threshold methods yield far more

equivocal results than do studies using sig-nal detection theory. Threshold measures

hopelessly confound sensory and attitudi-nal components, and lead to fruitless dis-putes about whether a procedure affectssensitivity to stimulation or the criterionfor reporting pain. The need is for more re-

search, not a return to outmoded psycho-physical procedures.McBurney (16) contends that the re-

quirements for signal detection theoryhave not been met, and cannot be met inprinciple. Fortunately for researchers inpain, this conclusion is false. Since it is notclear which assumption has not been met,let us match those introduced by Greenand Swets (1) to the present experiment. (i)Two or more quantifiable states of theworld exist, in this instance, radiant in-tensities of 0, 120, 240, 305, 370, and 425mcal sec-' cm-2. The independent variableis neither the sensory experience of heatnor that of pain, but the intensity of ther-mal radiation. (ii) The various physicalstates give rise to different distributions ofobservations. The observation may be re-

garded as unidimensional even when thesensory system has many dimensions, be-

67

on February 1, 2019

http://science.sciencem

ag.org/D

ownloaded from

cause it can be represented as a point in m-dimensional space; and there exists a likeli-hood ratio, or a unique set if the number ofstimulus intensities is greater than two, foreach such point. (iii) The observer re-sponds with a decision or hypothesis aboutthe physical state of the world. This deci-sion arises from some strategy, that is, adecision rule, which partitions the set ofpossible observations into a number of re-sponse classes; in the present instance,"nothing" through "warm" and "hot" to"painful." (iv) In accepting or rejectingvarious hypotheses about the state of theworld, the observer makes use of a poste-riori probabilities. (v) The likelihood ratiois a single real number that expresses thestrength of evidence associated with eachobservation. The dimensionality of the ob-servation or event in no way influences thecalculation of the likelihood ratio by theobserver. It is assumed that observers cancompare various sensory events with an or-dinal scale that is monotonic with the like-lihood ratio.Green and Swets (1, p. 123) state: "It is

important to note that the objectivity ofdetection theory methods does not requirethat the experimenter be able to score thesubject as right or wrong; he need onlyknow which value of the signal hepresented on each trial. The experimentercannot score the subject as right or wrongwhen he is measuring a transition from hotto cold, from not painful to pain, frombeats to roughness, or from achromaticityto chromaticity. He can, however, deter-mine the reliability with which an observercan discriminate between any two (ormore) signals on one of these continua bydetermining an ROC curve ..." Further-more, it is not necessary to obtain pain re-ports; d's can be based on intensity reports(high or low) or on confidence ratings (17).McBumey's other objection is that ab-

solute sensitivity and differential sensitivityare not two aspects of a single sensory ca-pability, since, for example, manipulationof sensory adaptation may cause the twotypes of sensitivity to move in opposite di-rections. This is certainly true. However, inthe study of visual contrast thresholds as afunction of stimulus duration, one wouldnever attempt to plot a point from the ab-solute sensitivity function (scotopic thresh-old) in the midst of a relative sensitivity(photopic) curve. Confusion arises here be-cause McBurney, and indeed many of thehandbooks, use but two terms to describethe four situations portrayed in Table 1.Thus, it is not always clear whether abso-lute and differential sensitivity refer to theintensity of the stimulus background or tothe intensity of the blank. (In signal detec-tion theory "blank" refers to the zero orlower intensity stimulus.) In his second

68

paragraph, McBumey's definition of abso-lute sensitivity is that of type A, and hisdifferential sensitivity refers to type B andperhaps type D measurements. It is per-fectly legitimate to compare type A mea-surements with type B, since both are abso-lute sensitivity with respect to background;similarly, types C and D measurementsmay be compared. It is improper, asMcBurney points out, to compare type A,absolute surround sensitivity, with eithertype C or D. This we did not do. Since thethermal stimuli were applied to warm skinof a constant intensity (single adaptationlevel), we were only concerned with type Cand D measurements, that is, with differ-ential sensitivity with respect to back-ground, and with both absolute (0 versus120 mcal sec-' cm-2) and differential sensi-tivity (120 versus 240 to 425 mcal sec-'cm-2) with respect to discrimination. Clark(18) obtained 17 d' values at intervals of 25mcal sec-' cm-2 from 0 to 425 mcal sec-'cm-2. The "absolute sensitivity point" at 0to 25 mcal sec' cmM2 was not unique andfitted the d' versus thermal intensity func-tion. Defined in terms of discrimination,but not in terms of surround, absolute anddifferential sensitivity are, indeed, relatedaspects of a single sensory capability.

Let us recapitulate our argument. Val-ues of d', independent of the quality of thesubjective report ("warm," "pain,""high," or "yes"), measure the capacity ofthe nervous system to transmit informa-tion about the amounts of thermal radi-ation reaching the skin. Conceivably,transmission in fibers mediating pain couldbe blocked without affecting d'. This seemsunlikely in principle, since d' depends upontotal information transmitted, nor has itbeen found in practice. Recognized anes-thetics (15) and analgesics decrease d', ac-upuncture did not; thus, we conclude thatacupuncture is not an effective analgesic.For what it is worth, and forgetting signal

detection theory entirely, both during andfollowing acupuncture, our subjects feltthat a skin incision would inflict the usualamount of pain.

W. CRAWFORD CLARKDepartment ofResearch Psychology,New York State Psychiatric Institute,New York 10032

J. C. YANGDepartment ofAnesthesiology,Columbia University College ofPhysicians and Surgeons,New York 10032

WAYNE HALLDepartment ofPsychology,University ofNew South Wales,Kensington, A ustralia, 2033

References and Notes

1. D. M. Green and J. A. Swets, Signal DetectionTheory and Psychophysics (Wiley, New York,1966); W. C. Clark, Anesthesiology 40, 272 (1974).

2. C. R. Chapman, J. D. Gehrig, M. E. Wilson, Sci-ence 189, 64 (1975).

3. Board of Acupuncture Anesthesia, AcupunctureAnesthesia (in Chinese) (People's Publishing So-ciety, Shanghai, ed. 1, 1972), pp. 214-215.

4. W. C. Clark and J. C. Yang, Science 184, 1096(1974).

5. Research Group of Acupuncture Anesthesia, Pe-king College, Chin. Med. J. 3, 151 (1973).

6. P. H. Jenerick, Ed., Proceedings NIH Acupunc-ture Research Conference (Publication No. NIH-74-165, Department of Health, Education, andWelfare, Washington, D.C., 1973).

7. Durations of acupuncture treatment were reportedas follows: 8 minutes, P. L. Man (6, p. 69); 15 min-utes, G. Smith (6, p. 69); 20 minutes, R. W. Bren-nan (6, p. 46) and P. C. Tang (6, p. 65); and 30 min-utes, D. E. Bresler (6, p. 68).

8. S. C. Man and F. D. Baragar, Can. Med. Assoc. J.109, 609 (1973).

9. H. T. Chang, Sci. Sin. 16, 25 (1973).10. For example, analgesia was reported to persist 40

minutes by P. C. Tang (6, p. 65), 12 hours by D. E.Bresler (6, p. 68), but only 30 seconds by M. H. M.Lee (6, p. 77).

11. A. Taub, Science 178,9 (1972).12. W. C. Clark,J. Abnorm. Psychol. 74, 363 (1969).13. and J. S. Goodman, ibid. 83,364 (1974).14. R. L. Hayes, G. J. Bennett, D. J. Mayer, Science

189,64(1975).15. C. R. Chapman, T. M. Murphy, S. H. Butler, ibid.

179, 1246 (1973).16. D. H. McBurney, ibid. 189,65 (1975).17. W. C. Clark and L. Mehl,J. Exp. Psychol. 97, 148

(1973); W. C. Clark and D. J. Dillon, Percept. Psy-chophys. 13, 491 (1973).

18. W. C. Clark, paper presented to the annual meet-ing of the Psychonomic Society, St. Louis, 1971.

10 March 1975

Hemispheric Asymmetry and Musical Performance

Bever and Chiarello (1) described an un-

expected right ear superiority (putatively a

left hemisphere dominance) for musiciansin a melody recognition task. The usualobservation (2) that recognition of melodicstimuli produces left ear superiority (righthemisphere dominance) was seen in a sec-

ond group of listeners who were musically'naive." The counterindication was inter-preted to be a result of musical training inwhich specialized left hemisphere analysishad been developed. The trained ("sophis-ticated") musicians apparently had learnedto analyze the tonal sequence according to

what Bever and Chiarello called "internal

relationships of its components," an aspectof the serial or sequential (analytic) pro-cess of the left hemisphere (3). In contrast,musically naive listeners lacking special-ized training had processed melodic pas-sages according to the more holistic or

unit mode of cognition in the right hemi-sphere (4).To support the effect of musical train-

ing, Bever and Chiarello cited my report(5) of college musicians who also failed toshow the usual left ear dominance for mel-odies. But my results did not demonstrateright ear superiority, whereas they didshow left ear dominance for musical

SCIENCE, VOL. 189

on February 1, 2019

http://science.sciencem

ag.org/D

ownloaded from

Letter: Acupuncture, pain, and signal detection theory

DOI: 10.1126/science.1135629 (4196), 65-68.189Science 

ARTICLE TOOLS http://science.sciencemag.org/content/189/4196/65.citation

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceGovernment Works. The title reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S.Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights

(print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for theScience

on February 1, 2019

http://science.sciencem

ag.org/D

ownloaded from