Perception & Psychophysics1975, Vol. 17 (4). 337-345

Understanding prism adaptation:An individual differences approach

DAVID H. WARREN and BRUCE B. PLATTUniversity of California, Riverside, California 92502

Adaptation of college students to wedge prisms was studied in terminal exposure and continuousexposure conditions. The adaptation experience was preceded by the administration of a battery ofpretests designed to evaluate several visual and motor abilities thought to be involved in adaptationperformance. Ability indices were formed for each subject on the basis of his performance on the pretests.These indices were then used in multiple regression analyses with several measures of adaptation asdependent variables. The ability indices accounted for substantial portions of the dependent measurevariance. In the terminal exposure condition, amount of adaptation was positively correlated with indicesof eye ability and negatively correlated with indices of hand ability. Intermanual transfer of adaptationwas positively correlated with hand abilities and negatively correlated with eye abilities. In thecontinuous exposure condition, amount of adaptation was positively correlated with eye abilities andnegatively correlated with hand abilities. It was suggested that theoretical and experimental approachesmay be oversimplified if they do not take into account the possibility that at least some of the differencesin the adaptation shown by different subjects is attributable to differences in abilities that are basic to theprocess of adaptation.

Studies of adaptation to the lateral visualdisplacement produced by wedge prisms havetypically found substantial between-subject varia­bility. Under identical experimental conditions, somesubjects adapt very little, while others adapt almostcompletely. There has been very little seriousattention paid to this variability-it has routinely beentreated as error variance, either explicitly orimplicitly. At the same time. there has been a greatdeal of attention paid to the theoretical controversyover whether adaptation consists of a shift inperceived visual position or in felt position of the handor arm. Harris (1%5) has served as a majorspokesman for the hand shift hypothesis, whileMcl.aughlin, Rifkin, and Webster (1966) provide anexample of the visual shift hypothesis.

These two topics. response variability and visual vs.hand shift hypothesis. may be more related than isapparent at first glance. If adaptation consists of twoor more components (e.g., both visuomotor andproprioceptive hand shifts). then it may be possible toaccount for some of the variability in adaptationresults by taking between-subjects differences into

lhc rcscurvh was supported by a University of CaliforniaIntramural Research Grant to the first author. Part of themaleri,d was presented at the Conference on the RecombinationProcedure as a Tool for the Study of Visual Perception.November 147J. at the University of Kansas. We thankProtcvsor Roy D. Goldman for his patient statistical consultationand we thank Steve Morrell. Jill Osborn. David Rapaport.lcrrv Schmitt, and David Huttner for their long hours ofrese,;rch assistance. Send requests tor reprints to David H.Warren. Department of Psychoiogy. University of California.Riverside. Calilornia 42502.

accou nt. Specifically, it may be useful to considerabilities that might lead subjects to experiencedifferent amounts of visuomotor and proprioceptivehand shifts.

There is evidence that adaptation to the lateraldisplacement produced by prisms consists of at leasttwo components. In a long-term study, Hay and Pick(i %6) found that a proprioceptive shift occurred inthe early part of the exposure period, and that anoculomotor shift gradually appeared, replacing theproprioceptive shift. McLaughlin and Bower (1965)provided evidence for the concurrent development ofoculomotor and proprioceptive hand-shift compo­nents. The oculomotor component produced a shift inthe responses of both exposed and nonexposed hands,while the proprioceptive hand shift occurred only forthe exposed hand. McLaughlin et al. (1966) provideda similar demonstration that both proprioceptivehand-shift and oculomotor components of adaptationoccur at the same time. Further. the oculomotorcomponent was separable into two types. onedependent on perceived lateral shift and the otherdependent on the apparent rotation of a planeperpendicular to the line of sight. Welch and Rhoades(1%9) measured the magnitude of negative aftereffect(NA) and of proprioceptive shift (PS) under variousinformation conditions in an attempt to determinewhat sources of information were used in producingeach type of adaptation. Their results supported thenotion that at least two types of adaptation occursimultaneously in the same subjects: "NA and PSrepresent qualitatively different kinds of adaptation.since they are neither highly correlated with each

337

338 WARREN AND PLATT

other nor affected in the same manner by themanipulation of informational variables" (p.425).Thus it is evident that adaptation to the lateraldisplacement produced by prisms consists of two ormore components, and that the different componentsdepend on different sources of information.

The experiments reported here were designed toinvestigate the possibility that some between-subjectsvariability in adaptation may be explained by takinginto account certain perceptual abilities of theindividual subjects. Specifically, the exact way that asu bject uses the available visual and proprioceptiveinformation should depend on his visual and handabilities, and so the characteristics of his adaptationperformance should also depend on his visual andmanual abilities. Assessment of the relevant abilitiesshould therefore allow a more complete explanation ofadaptation results.

Two experiments were conducted with separatesamples of college students. In the first experiment,terminal exposure was provided. The subject viewed avisual target through a wedge prism, pointed to thetarget. and received a view of his pointing hand onlywhen it had contacted the curved surface on which thetarget was placed. In the second experiment, thesubject received continuous exposure. That is, he sawthe visual target and could see his hand continuouslyas he reached out to point to the target. In previousresearch (e.g., Cohen. 1967). significant adaptation ofthe exposed hand has been found for both terminaland continuous exposure conditions, while transfer ofadaptation to the nonexposed hand has been foundonly in the terminal exposure condition.

Preceding the administration of the adaptationphase of each experiment. each subject performed abattery of eight pretests, which were identical for thetwo experiments. Since the hypothesis was thatindivid ual differences in adaptation performancewould be in part attributable to differences in relevantabilities. the pretests were designed to provide variousmeasures of eye abilities. hand abilities. andeye-hand coordination. Sets of pretest scores wereused as predictors in multiple regression analyses withseveral measures of adaptation as dependentvariables.

GENERAL METHODS

ApparatusThe apparatus was a modification of the localization equipment

used in many previous studies of adaptation. The subject sat withhis head in a chinrest and forehead restraint (the use of a bitebarinterferes with EOG recording of eye movements). With his left eyeoccluded. the subject viewed with his right eye through either clearglass or a 10° base right wedge prism. as the various conditionsdemanded. The subject made pointing responses under a shelfwhich contained a movable shield that could be arranged toprevent the subject's view of his hand altogether or to givecontinuous or terminal view of the pointing hand. A second shelf,parallel to the first, extended over the subject's head. and khaki

doth was stretched around the entire shelving in order to preventthe subject's view of the testing room. Visual targets were presentedalong the arc of a circle with radius of SO ern and centered at theright eye.

Provision was made for recording both eye and hand movementsin an ongoing manner. Eye movements were recorded by means of aBeckman direct current electrooculography system. Previousevaluations of the accuracy of this system in our lab have shown thatthe 95% contidence interval around a given judgment is ±.75°. Foreach pointing response. the subject placed his foretinger on a trolleythat ran out along a boom centered on an axle directly below theright eye. Five microswitches were spaced equally along the boom.The tirst switch was tripped as soon as the subject began to point,and the last was tripped as the subject's pointing tinger contactedthe arc on which the target was mounted. The axle of the boomextended to the 11001'. where it was connected to a potentiometersystem. The output from the potentiometer appeared on onechannel of the Beckman dynograph and thus provided a continuousrecord of the radial direction of the pointing hand with reference tothe center of the apparatus. The microswitch output appeared onanother dynograph channel. These two pointing channels thusprovided a record of the radial direction of the pointing tinger ateach of five distances from the body.

PretestsEach subject in both experiments performed an identical set of

eight pretests. The pretests were presented in the order in whichthey are listed. and they were performed with the left eye occludedand the right eye viewing through clear glass.

Pretest 1: Fixating target. The subject was shown a point of lightdirectly ahead of him in an otherwise dark environment and wasdirected to fixate the light for 30 sec. The score was the number ofeye movements with a lateral extent greater than 1°.

Pretest 2: Fixating an imaginary target in the dark. The subjectwas asked to maintain his fixation at straight ahead in a completelydark environment for 30 sec. Again. the score was the number ofeye movements greater than 1°.

Pretest 3: Pointing to visual targets. A small light was placed atvarious azimuth target positions, ±5° and ± 100 from the straightahead. The subject was asked to fixate the target. then point to itwith his unseen right hand. Each target position was used twice. fora total of eight trials. The series was then repeated with the subjectusing his Icft hand. The pointing data were used to assess theaccuracy of pointing to visual targets with each hand: meanunsigned error and standard deviation were calculated for eachhand. The eye-position data were used to generate the calibrationfunctions needed for assessment of eye position in later tasks.

Pretest 4: Looking at felt hand position. The experimentermoved the subject's right forefinger to a target position andinstructed the subject to look, in the dark, at where he felt his handto be. The same target positions were used as in Pretest 3. After theright hand was used as a target, the series was repeated with the leftforefinger as target. The mean unsigned error of aligning the eyes tothe felt position of the hand and the standard deviation of thelooking responses were calculated separately for each hand.

Pretest 5: Eye and hand tracking of a moving visual target. Asmall light was turned on 100 to the subject's right of straightahead. The subject looked at it and pointed at it with his righthand, then tracked it with both eye and hand as it was moved by theexperimenter through straight ahead to 10° to the subject's left,then hack to lO" to the right. The light was moved at a rate ofapproxiruatclv So sec. The same procedure was then done with theleft hand. with the target moving from 100 left to 100 right andback. Six measures were taken from this task, the number of timesthat each hand stopped in tracking the continuous movement of thetarget, the summed unsigned errors of pointing at the midpoint andthe endpoint of the movement for each hand. and the number of eyediscontinuities <the number of times the eye broke its smoothtracking of the target) separately for each hand's trial.

Pretest 6: Eye tracking of the unseen hand. The right. then the

UNDERST ANDING PRISM ADAPTATION 339

48)

Mean SD

7.1 5.514.8 10.3

5.4 3.26.4 3.95.3 deg 3.4 deg6.1 deg 4.1 deg

3.5 deg 1.7 deg3.4 deg 1.8 deg2.9 deg 1.3 deg3.0 deg 1.3 deg9.9 3.08.9 2.98.0 deg 4.5 deg7.6 deg 5.0 deg3.4 deg 1.5 deg4.0 deg 1.7 deg4.1 deg 2.3 deg4.3 deg 2.1 deg

6.9 deg 3.7 deg5.4 deg 3.6 deg5.4 deg 3.4 deg4.4 deg 2.2 deg

15.1 5.814.8 6.130.8 deg 16.9 deg28.9 deg 13.7 deg

Number of eye movements (target)Number of eye movements (no target)Number of eye movement episodes (R hand)Number of eye movement episodes (L hand)Accuracy of reproduced eye positionVariability of reproduced eye position

A. Lye measuresI25577

13. !land measures3 Accuracy of pointing (R hand)3 Accuracy of pointing (L hand)3 Variabili ty of pointing (R hand)3 Variability of pointing (L hand)5 Number of hand stops (R hand)5 Number of hand stops (L hand)5 Hand error (middle + end) (R hand)5 Hand error (middle + end) (L hand)H Accuracy of reproduced hand positions (R hand)H Accuracy of reproduced hand positions (L hand)H Variability of reproduction (R hand)H Variability of reproduction (L hand)

C Eye-hand measures4 Accuracy of eye to hand (R hand)4 Accuracy of eye to hand (L hand)4 Variability of eye to hand (R hand)4 Variability of eye to hand (L hand)6 Number of eye discontinuities in tracking hand (R hand)6 Number of eye discontinuities in tracking hand (L hand)6 Eye-to-hand error (begin + middle + end) (R hand)6 Eye-to-hand error (begin + middle + end) (L hand)

Table 1Group Results on Pretest Measures (N

--------- ---- ---------------------------------------Pretest

lett hand was used as the target in a completely dark visualcnvinuuucnt. The subject's hand was moved by the experimenterthrough the same path used in Pretest 5. The subject's task was toeve-truck the hand as it moved. The measures taken from thispretest were the number of discontinuities in the eye-movementtruck. separately for each hand target. and the eye-to-hand errorssummed over the starting point. the halfway extreme point, and theend point of the movement. again separately for each hand,

Pretest 7: Replicating an eye position. The experimenter placedthe target light at a target location, and the subject tixated it for asecond, Then the light was turned off and the subject was instructedto look oil to one side, then to look back at the place where the lighthad been, The same four target positions were used as in previouspretests. The mean unsigned error was tabulated, as was thestandard deviation of localizations.

Pretest Il: Replicating a hand position. The experimenter movedthe subject's right hand to a target position, held it there for asecond, then moved it off to one extreme. The subject's task was tomove his hand back to the target position. The visual environmentwas completely dark. Each of the four target positions was usedonce with each hand, and mean unsigned error and standarddeviation were calculated for each hand.

In all, 26 subrneasures were taken from the eight pretests.Complete pretest analysis was not possible for 2 of the 50 subjects.For the remaining 41l subjects. the means and standard deviationstor each of the 26 sub measures appear in Table I.

I he purpo,,' 111 condul'llng tile pretests was to measure subjectabilities iluu might he impurtant in accounting "1I' variance inmeasures III .ulapt at ion . lhc suhmcusurc-, were clnssificd into eight~lhilit~- call'guriL".

II) FIe Accuracv. Mean unviuncd error and standard deviationlnun Prell',t -:. . "-

12) I',\e Smoothlll'SS, Pretests I and 2; eye discontinuity measures

lrum Pretest :-; eye divcnntinuuy measures from Pretest 6.(,1) I{ight-Halld Pointing, Mean unsigned error and standard

dl'\i"ti,," of pointing by the right hand in Pretest 3; sum ofullsiglled crror-, of right-hand pointing in Pretest 5,

H) Left-Hand Pointing. Parallel to Right-Hand Pointing.1:-) Right-Hand Proprioception. Mean unsigned error and

sl'llldard deviation of directing the eyes to the right hand inPretest 4; mean unsigned error and standard deviation ofright-hand performance in Pretest Il; summed unsigned errors ofdirecting the eyes to the right hand in Pretest 6.

(h) Left-Hand Proprioception, Parallel to Right-Hand Pro­prioccpt ion.

(7) Right-Hand Smoothness. Number of right-hand discon­tinuities in tracking visual target in Pretest 5.

(K) Left-Hand Smoothness. Parallel to Right-Hand Smoothness,'I hus there were three types of categories, proprioception

(representing afferent abilities). response (representing efferentahilitics). and smoothness, It should be noted that the EyeSmoothness and Hand Smoothness categories were parallel, butthat the Fye Accuracy category included both afferent and efferentcomponents, while the afferent and efferent components wereseparated into distinct categories for the hands. It should also benoted that several of the pretest submeasures could have beend,,,silied in other ways than they were. The most difficult decisionwas the clasvification of Pretest Il. replicating a hand position,which involver! both proprioception and pointing components, ThePretest K suhmcasures showed a stronger pattern of correlationswirh the other proprioceptive sub measures than with the otherpointing submcnsurcs. and so the pretest was included in theproprioccpt ivc categories, It may he noted from Table 2 that thecorrcl.uion-, of the pointing with the proprioceptive categories werenot signiticant. For other submeasures which involved twocomponents and thus might have been classified in either of two

340 WARREN AJ.'lD PLATT

Table 2Correlation Matrix for Eight Ability Categories (N= 48)

EyeAc EyeSm RHPt LHPt RHPro LHPro RHSm

Eye accuracyEye smoothness .31*R hand pointing .01 .15L hand pointing -.01 -.01 .54**R hand proprioception .37** .04 .11 .11L hand proprioception .17 -.16 .01 .24 .40**R hand smoothness -.15 .28* -.10 .21 -.34* -.19L hand smoothness -.16 .19 -.02 .05 -.18 -.05 .79**

ability categories. the decisions were typically not difficult becauseone component was clearly more contributory to performance thanthe other. For example. in Pretest 3. pointing to a visual target. anerror could be due to misperceiving the target location because ofinadequate eye movements. or to inaccuracy of the hand in pointingto the target. The Pretest 3 submeasures were included in thehand-pointing categories on the grounds that the errors in directingthe eyes to a small visual target are trivial when compared to theerrors of pointing to a visual target.

Table 2 shows the pattern of correlations among the eight abilitycategory scores. Several points are evident from the data in thetable. First. the Eye Accuracy and Eye Smoothness categories werenot independent of one another. Second. there were strongcorrelations between the right- and left-hand measures ofsmoothness. pointing. and proprioception. Third. the correlationsbetween smoothness. pointing. and proprioception categorieswithin the right- and the left-hand divisions were generally notsignificant, suggesting that these various categories might beexpected to account for separate portions of variability in adependent variable on which they are regressed.

Data from the 26 pretest submeasures were converted to z scores.and an ability index was formed 1lH' each subject for each categoryI" summing his z scores for the submeasures in that category.lhus , tor each subject. there were eight indices. one representinghis performance in each ability category. It was hypothesized thatthe ability indices. as assessed in the pretest phase. wouldcontribute to variability in the dependent measures taken in theadaptation phase of the experiments.

EXPERIMENT I:TERMINAL EXPOSURE CONDITION

Method

In the terminal condition. a visual target subtending .50 wasplaced at the 1000 position. so that it appeared at straight ahead. or(}(f. when viewed through the base-right prism. The subject wasinstructed to point to the target by placing his right fore linger onthe trolley and pushing it out until his linger contacted the Plexiglasarc on which the target was mounted. The experimenter instructedthe subject to take about 1.5 to 2 sec to make the pointing motion.then to retract the hand after I sec. The movable shield blockedthe, subject's view of his hand and the boom until his lingercontacted the Plexiglas arc. The subject was not allowed to correcthis response while the tinger was in view. He pointed repeatedly atthe target for a maximum of 20 trials. If the subject made liveconsecutive responses within 20 of the visual target position. he wasthen given three additional trials and the adaptation phase wasterminated. On each trial. the hand boom was started alternately10° to the subject's right or 10° to his left of the previous response.

The adaptation task just described was the third in the series ofsix tasks needed to assess adaptation. Each of 25 subjectsperformed the six tasks in the same order. In Task I. the subjectpointed with his right hand at each of six visual targets. The shield

was arranged to cover the hand completely. so that the subjectreceived no information about the correctness of his response.Task 2 was the same as Task I. except that the subject used his leftha 11(1. Thus. Tasks I and 2 provided measu res of the accu racy ofpointing with right and left hands before adaptation. Task 3 wasthe adaptation task. already described. Task 4 consisted of fourtrials pointing with the left hand. with no visual feedback.Comparison of Task 4 with Task 2 thus provided a measure of theshift in pointing of the left (nonad apted) hand. Task 5 was includedto regain any loss of adaptation of the right hand suffered duringthe delay involved in Task 4. It consisted of a maximum of 12right-hand adaptation trials. with terminal feedback as in Task 3.Task 6 consisted of four trials pointing with the right hand. with nofeedback. Comparison of Task 6 with Task I thus provided ameasure of right-hand adaptation. (Following these six tasks.additional tasks were administered to assess the course ofadaptation back to conditions of zero visual displacement. but thesedata will not be discussed here.)

ResultsGroup results. A graphic representation of the

group results appears in Figure 1. The ordinate is indegrees: 90° is really straight ahead, but the subjecthad to point to his right, to 100°, to hit the target. Themean response shifted rapidly over the early trials.Criterion was defined for each subject as the first ofany set of four trials on which the response was within2° of the target. The mean trials to criterion was 10.3,standard deviation 6.6.

Another way to assess adaptation is to compareperformance in no-feedback pre- and postadaptationconditions of pointing to visual targets. Thehorizontal arrows at the ordinate indicate thepreadaptation performance for right and left hands(Tasks 1 and 2). The heavy vertical arrows representthe net change in pointing from pre- topostadaptation for right and left hands (i.e.,Task 6-Task I. and Task 4-Task 2). The mean shiftby the right hand was 7.8°, standard deviation 3.0°.while the mean shift by the left hand was 3.9°.standard deviation 3.8°. The adaptation as assessed inthis way was significantly greater than zero for bothhands, and the right-hand shift was significantlygreater than the left-hand shift.

Multiple regression analyses. Stepwise multipleregression analyses were used to assess the relativecontribution of each of the eight ability indices to thevariance in each of several dependent measures of

adaptation. Multiple regression as a data analytictechnique can be used to assess the suitability ofdescriptive models. The stepwise procedure firstselects the predictor that is most highly correlatedwith the dependent variable. In each subsequent step,that, predictor is found which makes the next bestprediction, that is, explains a new source of variancenot overlapping those already determined. Theimportance of each predictor's contribution to theexplanation of the dependent variable is bestrepresented as the "R2added" (~R2) by the predictor.

The dependent measures included the number oftrials to criterion in the adaptation phase. the amountof right-hand adaptation, the amount of left-handtransfer, and the proportion of left-hand shift relativeto right-hand adaptation. The measure of right-handadaptation was the ratio of demonstrated adaptationto the total amount possible, given the individualsubject's right-hand preexposure performance orconstant error. That is, right-hand adaptation =()(T6 - )(n)/(loo° - )(n). The measure of left-handtransfer was the ratio of demonstrated shift to thetotal amount possible, again taking the subject'spreadaptation performance into account. That is,left-hand adaptation = ()(T4 - )(1'2)/(100° - )(1'2).The following ratio was calculated to represent theproportion of left-hand shift relative to the righthand's adaptation: left-hand proportion()(T4 - XT2)/()(T6 - )(n)·

The multiple correlation of the five pretest abilityindices (excluding the left-hand indices) with the trialsto criterion measure was .39. The indices accountedfor only 15% of the variance in the measure. Clearly,the pretest categories did not bear much relevance tothe speed of adaptation.

For the right-hand adaptation measure, the abilitycategories accounted for about 29% of the variance inthe dependent measure (R = .54). The mostimportant category was Right-Hand Smoothness (~2= .15). followed by Eye Smoothness (~R2 = .07) andEye Accuracy (~R2 .04). The pattern ofcorrelations between the pretest submeasures and thedependent variable showed that the better the handabilities were, the less adaptation occurred, while thebetter the eye abilities were, the greater theadaptation. The results make good sense if weconsider that the subject is confronted with a problemto solve (missing the visual target) in the earlyadaptation trials. and that he has two sources ofinformation available (visual and proprioceptive) tohelp him solve the problem. The better his eyeabilities are and the worse his hand abilities are, thegreater the subject's tendency is to solve the problemby adjusting his hand behavior to fit the visualinformation. and the more adaptation he shows.Conversely. the better his hand abilities are, the lesstendency he has to solve the problem by adjusting his

UNDERSTANDING PRISM ADAPTATION 341

98

OlC>

~ 96

wenZo 94~

en Rwco: 7.7 0

l92Z 3.5

0

-cw

:i 90 l

erit.10.3

88 R

2-3 ~-s 6-7 8-9 10-11 12-13 1~·IS 16-17 18-19

TRIAL

Figure I. Growth of adaptation, terminal condition.

hand behavior to suit the visual information, and theless adaptation he shows.

For the left-hand dependent measures, all eightability indices were included as predictors, since itseemed likely that some right-hand abilities mightintluence the amount of left-hand shift. For theleft-hand adaptation measure, the ability indicesaccounted for 33% of the variance (R = .58). Theimportant indices were Eye Accuracy (~R2 = .10),Left-Hand Proprioception (~R2 = .06), Left-HandPointing (~2 = .05), Left-Hand Smoothness (~R2 =.03), and Right-Hand Pointing (~2 = .03). Thedirections of these relationships were the same as inthe analysis ofleft-hand proportion, discussed below.

The left-hand adaptation measuredisregards whatmay be an important determinant of left-handadaptation. namely the amount of adaptation shownby the right hand. It may be unreasonable to equatethe absolute magnitudes of left-hand adaptation fortwo subjects, one of whom showed a lot of right-handadaptation and the other of whom showed very little.A more appropriate dependent measure for theleft-hand effect is one that takes into account themagnitude of the right hand's adaptation. Theleft-hand proportion measure served this purpose byrepresenting how completely the right hand's effect(however large or small it was) transferred to the lefthand. The ability indices accounted for 60% of thevariance in this dependent measure (R = .78). EyeAccuracy (~R2 = .40) was the greatest contributor tothe measure. followed by Left-Hand Proprioception

342 WARREN AND PLATT

(AR 2 = .08) and Left-Hand Smoothness (AR2 = .07).The pattern of correlations between the pretestsubmeasures and the dependent variable showed thatthe better the eye accuracy was. the less transferoccurred. and th'at the smoother and more accuratethe left hand was. the more transfer occurred.

The results from the analyses of transfer apparentlycreate a puzzle. The hand and eye abilities wererelated to the dependent measures in an inversemanner. as was the case for right-hand adaptation.However. the directions of the relationships werereversed. Good eye abilities were related to a largeright-hand adaptation but to a small left-handtransfer. Good hand abilities were related to smallright-hand adaptation but to a large left-handtransfer.

There are two distinct points to be considered inexplaining left-hand transfer. One is the amount of,adaptation shown by the right hand. Then, if thisamount of adaptation is considered to be available fortransfer. the second question is how to account for theamount of adaptation that actually does transfer tothe left hand. A study by Cohen (1967) bears on thetirst question. Cohen found a signiticant intermanualtransfer of adaptation under conditions of terminalexposure. although the magnitude of the transfer wassigniticantly less than the magnitude of theadaptation shown by the exposed hand. Cohensuggested that the adaptation shown by the exposedhand consists of two components, one a shift in theproprioceptive position of the exposed arm, and theother a shift in visual direction of gaze. The pro­prioceptive arm shift component of the adaptationis local to the adapted arm and therefore is notavailable for transfer. but the visual shift componentis general and may exert an effect on the behavior ofthe contralateral hand. In the present study, as inCohen's. it was unfortunately not possible by directmeans to separate the right hand's adaptation intoproprioceptive arm shift and visual shift components.Less directly, however, it may be argued, on the basisof Cohen's formulation. that if it is only the visualshift component that is available for transfer, thenthose subjects who showed more left-hand transfermust have acquired more of the visual shift componentthan the subjects who showed very little left-handtransfer, The results do provide some support for thisformulation, The correlation between the EyeAccuracy index (on which a high score means pooraccuracy) and left-hand transfer was .63, indicatingthat those subjects with better eye accuracy showedless left-hand shift.

With respect to the second question, it should benoted that there are other factors besides the amountof visual shift adaptation that may affect themagnitude of shift shown by the contralateral hand.Specifically. the abilities of the left hand should be

involved. In fact, both Left-Hand Proprioception(AR 2 = ,08) and Left-Hand Smoothness (AR 2 = .07)contributed substantially to variability of left-handtransfer. As already noted. both relations showed thatthe better the left-hand abilities were. the moretransfer occurred. How may these results beexplained'! Consider the implications of the visualshift that is hypothesized to develop during right-handadaptation. The effect of the visual shift is to producea new frame of reference. one that is shifted somenumber of degrees laterally (to the subject's right, inthe present situation). Thus. in Task 4, where thesu bject saw a visual target and had to point to it withhis left hand. the apparent position of the target wasshifted laterally from its apparent position beforeadaptation. The better the subject's left-hand abilitiesare. the better he should be able to point where hewants to. namely to the position of the visual targetwithin the new, shifted frame of reference, The factthat those subjects with better left-hand abilitiesshowed more left-hand transfer thus makes sensewithin this formulation.

It should be noted that these explanations are basedon post hoc reasoning. Logically, the weakest part ofthe argument is that there was no basis,independently of the left-hand transfer results, toassess the magnitude of the visual shift component ofadaptation acquired during adaptation of the righthand. Further experimentation. in which the visualshift component is independently assessed, isnecessary.

EXPERIMENT D:CONTINUOUS EXPOSURE CONDITION

Method

In the continuous adaptation condition, the experimentalsituation was identical to that described for the terminal conditionexcept in Tasks J and 5. In Task J. the adaptation exposure task,the shield was removed so that the subject could see his hand duringthe entire pointing movement. with the exception of the initial'J--t in .. where the subject's hand was hidden by the prism mount.Six blocks of trials were given in Task J. A block consisted of foureontinuous exposure trials, followed by two test trials where theshield \\as arranged to prevent visual feedback completely. Thesubject was instructed to take 1.5 to 2 see to point. then pull hishand hack completely. then point again with a delay of only asecond or t wo. The no-ICed back trials in each block followed in thesame timing sequence. since the experimenter could slide the shieldinto place during the short delay. In Task S, two blocks of trialswere given. Tasks I. 2.4. and /) were the same as in the terminaleondition.

Tw cn tv-Iivc naive college vtudcnts served a'> subjcctv. None hadparticipated in the previous experiment.

ResultsGroup results. A graphic representation of the

group results appears in Figure 2. Again. the ordinateis in degrees, with 90° straight ahead. The subject had

UNDERSTANDING PRISM ADAPTAnON 343

I"ij:ure 2. Growth of adaptation, continuous condition.

Multiple regression analyses were conducted usingboth left-hand dependent measures. The abilityindices accounted for only 14% of the variance in theleft-hand adaptation measure, and for only 17% ofthe variance in the left-hand proportion measure. Thefailure of the categories to account for substantialvariance was not surprising in view of the lack of asignificant shift by the left hand.

Despite the fact that the left-hand transfermeasures did not show effects significantly greaterthan zero for the continuous exposure group, therewere some su bjects who did show evidence of transfer.In fact. left-hand transfer, expressed in degrees ofshift. ranged from -1.7° to +4.5° (+ indicates a shiftin the direction of adaptation). In an earlier paper(Warren & Platt, Note 1), we suggested that soinesubjects in a continuous exposure condition mighteffectively be in a terminal condition. That is, despitethe fact that the hand is visible during its pointingmovement, the subject might retain fixation on thevisual target and not really gain the advantage of thecontinuous exposure. Such a subject might beexpected to show intermanual transfer. To test thisformulation. we measured the frequency of lateral eyemovements during the pointing movements in the tirstblock of four adaptation trials in Task 3. theadaptation phase. The correlation of this eye­movement measure with the amount of left-hand shift(in degrees) was -.49 (p < .02. df = 20). That is,those subjects who made fewer movements during theearly adaptation trials showed more transfer ofadaptation than those whose eyes were more active

TEST TRIAL

9-10 11-127-8

cn t.5.3

3-4 5-61-2

98

CI..~96wV)

Z094Q.

V)...a:

L

to point to his right. to 100°, to hit the target. Eachdata point represents the mean of the two no-feedbacktrials in a block in Task 3. The position of the firstdata point shows that considerable shift had occurredas a result of the first four adaptation trials. Criterionwas defined for each subject as the first block onwhich the two no-feedback trials averaged within 2° ofthe target position, if the corresponding trials on thesucceeding block were also within 2°. Mean trials tocriterion was 5.3. standard deviation 4.2.

Magnitude of adaptation was also assessed bycomparison of Tasks 6 and I. as in the terminalcondition, and left-hand transfer was assessed bycomparison of Tasks 4 and 2. The horizontal arrowsat the ordinate represent the preadaptationperformance for right and left hands (Tasks I and 2).and the heavy vertical arrows represent the net changefrom pre- to postadaptation for right and left hands.The mean shift by the right hand was 6.8°, standarddeviation 3.0°, while the mean shift by the left handwas 0.7°, standard deviation 1.8°. The mean left-handshift was not significantly different from zero. Thefailure to find a significant left-hand effect providescorroboration of the results of Cohen (1967), whoreported no transfer of adaptation to the contralateralhuud under conditions of continuous exposure.

Multiple regression analyses. The same abilityindices were calculated on the basis of pretestperformance as for the terminal condition. The abilityindices were again used as predictors in severalstepwise multiple regression analyses, one for each ofthe same dependent measures used in the terminalcondition analyses.

The multiple correlation of five ability indices(excluding the left-hand indices) with the right-handadaptation measure was .54; thus the indicesaccounted for 29% of the variance in the measure.Right-Hand Pointing contributed the major part ofthe variance (dR 2 = .17), followed by Right-HandProprioception (dR2 .(5), Eye Smoothness(dR 2 = .(3), and Right-Hand Smoothness (dR 2 =.(3). The direction of the correlations indicated thatsubjects with better right-hand abilities showed lessadaptation, and that subjects with better EyeSmoothness showed more adaptation. Theserelationships were parallel to those found in theterminal condition, in that better eye abilities wereassociated with more adaptation. while better handabilities were related to less adaptation.

Similar results were found for the trials to criterionmeasure. The five ability indices accounted for 58%of the variance in this measure (R = .76). The mostimportant indices were Right-Hand Smoothness(dR2 = .30) and Right-Hand Pointing (dR 2 = .26).The direction of the relationships indicated thatsubjects with better hand abilities took more trials toreach criterion.

344 WARREN AND PLATT

during adaptation. The question also occurs whetherthe relationship between left-hand transfer and eyemovements is one that is specific to the adaptationsituation itself or is more general. Would the amountofleft-hand transfer be correlated with an eye-activitymeasure taken independently of the adaptationsituation? A frequency-of-eye-movement measure wastaken from Task 1 of the adaptation phase, where thesubject pointed with his right hand to a visual targetwithout seeing his hand. The correlation of thismeasure with the magnitude of left-hand shift was-.54 (p < .01, df = 22): left-hand shift was greater forsubjects who showed relatively few eye movements in asituation independent of the adaptation phase.

It is unwise to draw causal conclusions from theseresults, since it is possible that some third factorinfluenced both amount of left-hand shift andfrequency of eye movements. However, the data donot contradict the notion that subjects who show lesseye activity during adaptation are effectively in anadaptation condition that involves terminal ratherthan continuous exposure. Further research, in whicheye activity is influenced by instruction or some otherexperimental manipulation, might provide adefinitive test of this notion.

DISCUSSION

The group results from the two prism-adaptationconditions reported here did not differ markedly fromthose of other prism work. In the terminal condition,the adaptation by the exposed hand averaged about60% of the amount possible, while the unexposedhand showed an average adaptation of about 40%. Inthe continuous condition, the exposed hand showedan average adaptation of about 65%, and that of theunexposed hand averaged less than 10% and was notsigniticantly greater than zero.

In recent years, several investigators have arguedthat it is oversimplified to consider adaptation toprisms simply as a shift in felt hand position or a shiftin some aspect of visual direction. Rather, adaptationconsists of at least two separate components whichdepend on different sources of information and whichmay develop at different rates. The presentexperiments corroborate the complexity of theadaptation process. In fact, the results demonstratethat the strength of the various components ofadaptation depends not only on different sources ofinformation, but on different abilities of subjects tomake use ofthe available information. In the terminalexposure condition, more adaptation was found for

subjects who had good eye abilities and lessadaptation was found for subjects who had good handabilities. Apparently. subjects with better eye abilitiestended to solve the problem of missing the visualtarget by depending more on the visual information.and they therefore showed more adaptation. Thesubjects with better hand abilities did not have such astrong tendency to rely on the visual information andtherefore did not adapt as much. The phenomenon ofintermanual transfer of adaptation is also clarified bythe present results. The subjects with better eyeabilities apparently acquired less of the visual shiftcomponent of adaptation, the component that isthought to be available for intermanual transfer(Cohen, 1967). Furthermore, the amount of theavailable adaptation that actually did transferdepended partly on the abilities of the left hand.Those subjects with better left-hand abilities showedmore transfer.

The results of the continuous exposure conditionare also useful. The magnitude of initial adaptationby the right hand showed a pattern of dependence oneye and hand abilities that was similar to the patternfor the terminal condition, although the indices of eyeabilities were less important in accounting forvariance than they were in the terminal condition.Detailed analysis of the eye-activity characteristics ofindividual subjects supported the hypothesis that theoccurrence of intermanual transfer in the continuouscondition depends on the extent to which the subjectactually makes use of the continuous view of his handin the initial adaptation. Intermanual transfer ofadaptation did occur for some subjects in thecontinuous condition. but it occurred for thosesubjects whose eye-behavior characteristics made thecontinuous condition effectively a terminal condition.

Most of the work on adaptation to prisms hasimplicitly taken the approach that a singleexplanation may be found that applies equally well toall subjects. This approach has been characteristiceven of theories that consider adaptation to be amulticomponent process. We suggest not only thatadaptation is a multicomponent process, but also thatthe relative importance of the components varies fromsubject to subject. not in a haphazard way but in away that may in part be explained by reference tovarious perceptual abilities of the subjects.Adaptation to prisms may indeed be primarily a resultof shift in felt hand position for some subjects, butprimarily a result of shift in visual direction for othersubjects. Adequate experimental treatment of thiscomplexity requires consideration of not just thecharacteristics of the experimental situation, but alsoof individual differences in ability characteristics thatexist independently of the experimental situation.

REFERENCE NOTE

I. Warren, D. H., & Platt, B. B. The subject: A neglectedfactor in recombination research. Paper delivered at the Conferenceon the Recombination Procedure as a Tool for the Study ofVisual Perception. November 1973, University of Kansas.

REFERENCES

COHEN. M. M. Continuous versus terminal. feedback in prismaftereffects. Pcrccptuul and Motor Skills, 1967, 24, 1295-1302.

HARRIS, C. S. Perceptual adaptation to inverted, reversed.and displaced vivion. Psychological Review, 1965. 72.419-444.

HAY, J. Coo &: PICK, H. L.. JR. Visual and proprioceptiveadaptation to ptical displacement of the visual stimulus.

UNDERSTANDING PRISM ADAPTATION 345

Journal 01 Experimental Psychology, 1966.71. lSO-158.McLAUGHLIN. S. C., & BOWER. J. L. Selective intermanual

transfer of adaptive effects during adaptation to prism.Psychonomic Science, 1965. 3. 69-70.

McLAUGHLN. S. c.. RIFKIN, K. I.. & WEBSTER. R. G. Oculo­motor adaptation to wedge prisms with no part of the body seen.Perception & Psychophysics, 1966. I. 452-458.

WELCH. R. B., & RHOADES, R. W. The manipulation ofinformational feedback and its effects upon prism adaptation.Canadian Journal of Psychology, 1969. 23. 415-428.

(Received for publication June 20. 1974;revision received November 22.1974,)