Brain (1996), 119, 1199-1211

Throwing while looking through prismsII. Specificity and storage of multiple gaze-throw calibrations

T. A. Martin,1 J. G. Keating,1 H. P. Goodkin,1 A. J. Bastian2 and W. T. Thach134

'Department of Anatomy and Neurobiology, 2The Program Correspondence to: W. T. Thach, MD, Department ofin Physical Therapy, iDepartment of Neurology and Anatomy and Neurobiology, Washington University SchoolNeurological Surgery and 4The Irene Walter Johnson of Medicine, 660 S. Euclid Avenue, Box 8108, St Louis,Institute of Rehabilitation Research, Washington University MO 63110, USASchool of Medicine, St Louis, USA

SummaryHuman subjects threw balls of clay at a visual target whilelooking through wedge prism spectacles. In studies of short-term adjustment, subjects threw in the direction of theirprism-bent gaze, missing the target to that side. Within 10-30 throws, they gradually adapted with a wider gaze-throwangle and hit the target. Immediately after removal of theprisms the wide gaze-throw angle persisted and throwsmissed the target to the opposite side, the so-called 'negativeafter-effect'. Repeated throws were required to adapt backto the normal gaze-throw angle and hit the target. Theadaptation was specific both to the body parts trained and

Keywords: throwing; prism; motor adaptation; learning

the type of throw trained: training with the right hand didnot generalize to throwing with the left; overhand trainingseldom generalized to underhand throwing. In a study oflong-term adjustment, two subjects threw with the same hand(right) and the same type of throw (overhand) alternately,with and without prisms, over a period of 6 weeks. Theygradually learned to hit the target on the first throw, withand without prisms. The two gaze-throw calibrations (prismand no-prism) were retained for >27 months. The long-termadjustment was shown to consist of a coordinated relationshipof eye-in-head, head-on-trunk and trunk-on-arm angles.

Abbreviations: AC = adaptation coefficient; LC = learning coefficient

IntroductionThrowing objects at visual targets is an acquired skill requiringpractice and adjustment to be accurate. In skilled throwing,gaze fixates the target for one or more seconds (Vickers,1994) and the subsequent throw is in the direction of gaze.Gaze direction is determined by the relative positions ofeyes-in-head and head-on-trunk and throw direction (in thehorizontal plane) by the positions of trunk in space and arm-on-trunk at the shoulder. A skilled throw requires coordinationof the positions and movements of these body parts.

We have used prism adaptation (Kane and Thach, 1989)to study short- and long-term adjustments of the gaze-throwangle. First, we examined the specificity of adaptation withrespect to the body parts involved. We asked whether theadaptation was private to the trained hand (right) or whetherthere was carry-over to the untrained hand (left). Next, weexamined the specificity of adaptation for different types ofthrowing. We asked whether the adaptation was specific fora trained overhand throw or whether there was carry-over toan untrained underhand throw. Last, we examined the storage

© Oxford University Press 1996

capacity for adjustments of eye-hand coordination. We askedwhether subjects could store two gaze-throw calibrations andwhether they would be able to use them interchangeably.

These results with those of the previous paper (Martinet al., 1996) suggest that the plastic relationship betweengaze direction and throw direction is controlled by thecerebellum. Gaze-throw adjustment would be a particularinstance of the more general cerebellar role in coordinatingmovements of many body parts. Portions of this work havebeen presented previously in brief (Kane and Thach, 1989;Thach et al., 1992a, b, 1995; Martin et al., 1993).

MethodsSubjectsControl subjects were healthy, unpaid adult volunteers withno history of neurological injury and were naive to thepurpose of the experiments. Ten normal subjects (mean

1200 T. A. Martin et al.

B100

5 0

3o

I§ -50rroi

-100

PC - 3.2

• #

1

BEFORE

A C - 5 . 0

iC'AtCuuo'iJ

F °PRISMS

;*

Oo

III IV

AC - 8.9

AFTER

V

TIME —

3.9 + -39.0*e(x/5 0) 5.4 + 29.0*e'"'8 9)

. THROW DIR. I \ . / 1

j-GAZEDIR. , „ ,„ | V v

Fig. 1 Prism adaptation test, control subject. (A) Eye-hand positions after adaptation to base-right prisms. The light path is bent to thesubject's right, showing more of the right side of her face. Her gaze is shifted left along the bent light path to foveate the target in frontof her. Her hand position is ready for a throw at the target in front of her. (B) Horizontal locations of throw impact points displayedsequentially by trial number. Deviations to the left are negative values; deviations to the right, positive. With the prisms (gaze shifted tothe left) the first impact is displaced ~60 cm left of centre. Thereafter throws approach the target (0). After removal of the prisms, thefirst impact is ~50 cm right of centre. Thereafter throws approach the target (0). (C) Gaze and throw directions are schematized witharrows. Inferred gaze (eye and head) direction assumes that the subject is foveating the target. Roman numerals beneath the arrowsindicate times during the prism adaptation experiment (see B).

age±l SD = 32.9± 10.1 years; range 25-54 years) partici-pated in the right hand-left hand experiment; another 10normal subjects (mean age± 1 SD = 30.6±7.5 years;range 22^44 years) participated in the underhand-overhandexperiment. Two other controls (aged 22 and 24 years)participated in the skill acquisition study over the course of6 weeks.

The tasksPrism-adapted throwing and measurement of the 'negativeafter-effect' (Weiner et al., 1983) has been described in thepreceding paper (Martin et al., 1996). Subjects participating inthe right hand-left hand, underhand-overhand, and alternatingprism-no prism paradigms viewed the target binocularly(Fig. 1A) through 30 diopter, Fresnel 3M Press-on plasticlenses (3M Health Care, Specialties Division, St Paul, Minn.,USA) mounted base-right in safety glasses. As in the previouspaper (Martin et al., 1996), subjects were tested while they

threw clay balls at a target (8X8 cm2 square drawn on alarge sheet of parcel paper) centred at shoulder level 2 m infront of them.

Right hand and left hand throwingIn an experiment designed to test for possible transfer ofprism training from one arm to the other, subjects executedthe following paradigm, (i) A pre-prism set of throws withclay balls was recorded for each hand, (ii) Subjects then puton 30 diopter, base-right, Fresnel prisms and threw with theirright hand until the point of impact repeatedly fell close tothe centre of the target, (iii) Subjects then removed the prismsand threw with the untrained left hand, (iv) Still withoutprisms, they again threw with the trained right hand. Subjectshad an unobstructed view of the target at all times and wereasked to throw where they saw the target. They were askednot to look at their hands when they were handed the clayballs and during the throwing. To see if there was transfer

Specificity and storage of eye-hand coordination 1201

of the training effect from the adapted right hand throw tothe untrained left hand throw, we looked for a statisticallysignificant negative after-effect in the latter. A Mann-WhitneyU test (special tables for small V ; Darlington, 1975) wasused to calculate the significance of any deviations of thefirst three left hand (and right hand) post-prism throws fromtheir respective pre-prism throws (the last eight throws madewith the same hand before donning prisms). As in thepreceding paper (Martin et al., 1996), we shall hereafter referto this simply as a 'significant after-effect'.

Overhand and underhand throwingIn a similar paradigm, underhand and overhand throwingwith the same hand were used to test for possible transfer ofprism adaptation from the one task to the other. First, (i)pre-prism sets of overhand and underhand throwing wererecorded, (ii) Then, subjects put on the prisms (30 diopters,base-right, Fresnel lenses) and adapted their overhand throwsuntil the point of impact repeatedly fell close to the centreof the target, (iii) Subjects then removed the prisms andthrew underhand. Finally, (iv) still without prisms, subjectsagain threw overhand. To see if there was transfer of thetraining effect from the adapted overhand throw to theuntrained underhand throw, we looked for a significant after-effect in the latter. A Mann-Whitney U test was used as aboveto calculate the significance of any post-prism deviations ofthe first three underhand (and overhand) throws from theirrespective pre-prism baseline throws.

Long-term training with and without prismsTwo subjects each threw tennis balls over a distance of2.5 m back and forth to each other each day, for 4 days perweek. Each subject wore 30 diopter base-right Fresnel prismsor no prisms in alternate 50 throw sets, until each had thrownand caught 200 times with and 250 times without prisms.The one starting with prisms finished with two sets of 50throws without prisms; the one starting without prismsfinished with one set of 50 throws. The first to wear prismseach day was alternated. On the fifth day, performance wasmeasured by recording 25 throws without, 100 throws with,and 50 throws without prisms. This routine was continuedfor 6 weeks: weekly totals were 900 throws with and 1075throws without prisms; totals for the 6-week period, 5400throws with and 6450 throws without prisms for each subject.

B

BEFORE AFTER

Fig. 2 Specificity of prism-adapted right handed throwing: notransfer to left handed throwing. (A) One subject threw lefthanded (filled circles on left) and then right handed (empty circleson left) before donning prisms. After donning prisms the subjectshowed normal adaptation. After removing the prisms left handthrows showed no after-effect. Subsequent right hand throws didshow after-effect. (B) Throwing accuracy with each hand beforeand after prism exposure for each of the subjects. The filledcircles represent left handed throws, the empty circles righthanded throws. Diagonal lines contact the means of the fourpopulations before and after prism adaptation of right handedthrows; vertical lines represent ± I SD from the mean.

Kinematics of gaze-throw coordination afterlong-term training.At the end of the 6-week period the horizontal angles ofeyes-in-head, head-on-trunk and trunk-on-arm throwing withand without prisms were calculated. Each subject wasvideotaped from overhead at 60 fields (30 frames) per secondwhile throwing before, during and after donning 30 diopter,

base-right Fresnel prisms. Reflective markers were placed onthe front and back of the head and on the tops of bothshoulders. The video was digitized and analysed using a PeakPerformance Motion Analysis System. Reflectors marked themajor axes in the horizontal plane of head (sagittal) andshoulders (transverse). Just prior to each throw, when thehand was at its most posterior position and the wrist cocked,these axes were measured and used to calculate relative

f i

. HORIZONTAL DISPLACEMENT (cm)

g 8 o S

m O^ HORIZONTAL DISPLACEMENT (cm) _ . HORIZONTAL DISPLACEMENT (cm) HORIZONTAL DISPLACEMENT (cm)

sI

~o

CO£co

OI

( ^

3

. HORIZONTAL DISPLACEMENT (cm)

8 8 o S 8

HORIZONTAL DISPLACEMENT (cm)

I 8 o S §HORIZONTAL DISPLACEMENT (cm) HORIZONTAL DISPLACEMENT (cm)

00

Specificity and storage of eye-hand coordination 1203

changes in each of the three angles eyes-in-head, head-on-trunk and trunk-on-arm while throwing before, during, andafter wearing prisms.

Knowing (from above) the position of the head-in-spacewith respect to the target and the angular deviation of theoptic path by the prisms (with subjects foveating the target)we calculated the horizontal positions of the eyes-in-headbefore, during and after prism use. The head-on-trunk anglewas calculated from the intersection of the line connectingthe two head markers and the line connecting the two shouldermarkers. A 90° angle between these two lines was normalizedto a 0° value for the head-on-trunk angle. The trunk-on-armangle was calculated from the shoulder-to-shoulder line andthe shoulder-to-hit location line. A 90° angle between theshoulder-to-shoulder line and the shoulder-to-target centrewas normalized to a 0° trunk-on-arm angle.

ResultsPrism adaptation of the gaze-throw angleFigure IB shows the horizontal location of the points ofimpact before, during and after prism adaptation in relationto the target. Figure 1C is a diagram of the gaze-throwrelationships as the subject undergoes prism adaptation. (I)shows the normal gaze-throw calibration with the throw inthe same direction as gaze. (II) shows gaze shifted by theprisms to the left with throws in the same direction as gazeand points of impact missing the target to the left. (Ill) showsgaze still shifted to the left but the throw has adapted in arightward direction towards the target. (IV) shows gaze on-target with prisms removed, but the adaptation persists withthrows still to the right of gaze, missing the target to theright—the 'negative after-effect' (Weiner et ai, 1983). (V)shows gaze and throw both on-target; the calibration hasreadapted to normal.

Right hand and left hand throwing: specificityof adaptation to the trained handBefore wearing prisms subjects threw first with the left andthen with the right hand (Fig. 2A). Donning prisms, subjectsthrew with the right hand. Initial impacts were in the directionof the prism-displaced gaze but with repeated throws shiftedgradually toward the target. After removing the prisms thesubjects threw with the untrained left hand. The impact pointswere all close to the target centre, without significant after-effect. Finally, the subjects threw again with the trained righthand, without prisms. There was a significant after-effect,which had persisted despite use of the left hand and which

required repeated right hand throws before they graduallyreturned to the target.

Figure 2B is a summary of impact points across subjects,hands and prism conditions. Circles represent the means foreach subject for the last eight pre-prism throws beforedonning, and the first three post-prism throws after removing,the prisms for each hand. Filled circles represent left handthrows and empty circles, right hand throws. Lines connectmeans of pre-prism throws to means of post-prism throws.Vertical bars represent ± 1 SD. There was a significant andlarge after-effect for all post-prism right hand throws (emptycircles) and no significant after-effect for all left hand throws(filled circles).

The after-effect for the prism-trained right hand wassignificant across all subjects (Mann-Whitney U test). Therewas no significant after-effect for the prism-naive left handin any subject. For the right hand, the difference betweenthe mean of the pre-prism throws and the mean of the post-prism throws was significant for all subjects (one-tailedStudent's t test; t = 5.49, d.f. = 18, P < 0.0001). Thedirection of after-effect displacement for the right handthrows was always consistent with expected deviation. Forthe left hand, there was no significant difference between themean of the pre-prism throws and the mean of the post-prism throws (one-tailed Student's t test; t = 0.767, d.f. =18, P > 0.05). In summary, there was no transfer of adaptationto the untrained left hand.

Overhand and underhand throwing: specificityof adaptation to a particular trained actOn donning prisms all subjects adapted the overhand throw.For five subjects (four are shown in Fig. 3A-D) the subsequentpost-prism underhand throw showed no significant after-effect (Mann-Whitney U test) while the subsequent post-prism overhand throws did (Mann-Whitney U test). Theoverhand significant after-effect persisted despite interveningpost-prism underhand throws. Overhand throws thenreadapted with repeated throws. In another three subjects(two are shown in Fig. 3E and F) results were similar exceptfor transfer of the overhand adaptation to the first subsequentunderhand throw. Two of these three subjects showedsignificant after-effect in the underhand throws (Mann-Whitney U test). The transferred after-effect disappeared withthe second throw. Yet in these three subjects, as in thefirst five, prior overhand adaptation survived interveningunderhand throws and appeared as a significant after-effectin subsequent overhand throws (Mann-Whitney U test).

Fig. 3 Relative specificity of prism-adapted overhand throwing in individual subjects. (A-D) There was no transfer to underhandthrowing in five out of ten subjects (four shown here). (E and F) There was transfer to the first underhand throw in three out of tensubjects (two shown), and (G and H) greater transfer in two out of 10 subjects. In all graphs the throwing order is pre-prism underhand,pre-prism overhand, prism-overhand, post-prism underhand and post-prism overhand.

1204 T. A. Martin et al.

The overhand after-effect readapted only after repeatedoverhand throws.

Two of the 10 subjects (Fig. 3G and H) showed nospecificity of training. They showed significant after-effectin both underhand and overhand throws (Mann-Whitney Utest). The underhand after-effect disappeared with repeatedunderhand throws. Yet the overhand after-effect persisted. Inone subject (Fig. 3G), the overhand after-effect thoughsignificant was small. In the other subject (Fig. 3H), theunderhand and overhand after-effects were approximatelyequal in magnitude. Both subjects required subsequentoverhand throwing to readapt.

Long-term training with and without prisms:acquisition and storage of a second gaze-throwcalibrationWith subjects practicing over 6 weeks the first throws withprisms and the first throws without hit progressively closer tothe target, ultimately hitting it. Figure 4A shows one subject'sinitial prism training session; adaptation and significant after-effect are apparent. After 6 weeks' training the subject's initialprism and post-prism throws are almost indistinguishable frompre-prism throws (Fig. 4B). The subject had stored a secondgaze-throw calibration ('known-prisms' calibration) and coulduse it instantly and interchangeably with the 'no-prism' normalcalibration. Figure 4C shows the gradual reduction of throwingerror while alternating prisms and no-prisms in both subjects.The subjects' pre-prism throws moved to the left until theywere centred about the target by week 5 (last 15 throws fromeach subject before donning prisms are represented by smallblack dots; the means± 1 SDof the pre-prism throws from bothsubjects are represented by bars). The initial deviation to theleft upon donning prisms (average of first three throws upondonning prisms for both subjects; triangles) and the negativeafter-effect (average of first three throws upon removing prismsfor both subjects; large filled circles) both moved closer to pre-prism performance until falling close to the target by the fifthweek of practice.

The location of these initial impact points, plotted againsttime and number of throws, approximated exponential decaycurves similar to those seen for adaptation (Martin et al., 1996).As in the adaptation curves, the rate of change of slope (thedecay constant) represented the rate at which the subjectacquired the behaviour. For the two subjects in Fig. 4, thisvalue was 1.4 weeks for acquisition and 2.4 weeks for thenegative after-effect. To distinguish this rate measure from thatof adaptation (AC; Martin et al., 1996), we have called the ratemeasure for the second type of curve a learning coefficient(LC).

Both subjects were tested for retention of this newlyacquired skill at 9, 18 and 27 months after the end of the 6-week training period without any other intervening exposureto the prisms. Both subjects showed nearly complete retentionof this ability (Fig. 4C).

Cuing context for retrieval of stored calibrationsWe next asked how the subjects accessed the no-prism andknown-prism calibrations and whether the prism gogglesthemselves provided the contextual key. After the two subjectshad practiced for 6 weeks with 30 diopter base right prisms(the known prisms), we substituted 'novel' prisms for theknown prisms. The novel prisms were 30 diopter, base left,in the same type of frames as the known prisms. The subjectsdid not indicate at the time or afterwards that they recognizedthe switch had taken place before their first throw. The resultsof their throwing under no-prisms, known-prisms, and novel-prisms conditions are shown in Fig. 5. At (A) subjects threwon target while wearing no prisms (empty circles). At (B)with novel prisms (squares), the gaze was directed ~17° tothe right, and the throw 34° (~100 cm instead of theappropriate 50 cm) to the right. This suggested that the novelprisms deflected the gaze appropriately, but called up thenow-inappropriate known-prisms gaze-throw calibration.During the adaptation with subjects wearing novel prisms (Band C), a break in the curve was seen for both subjects atabout the same half-way point. Both subjects required anumber of throws to adapt this deviation but finally threwon target (C). A negative after-effect of 50 cm leftward shiftwas seen both at (D) when the subject threw without prisms(one test-throw; filled circle) and at (E) (one test-throw andsubsequent, repeated throwing; triangles) when the subjectthrew with the original, familiar prisms. Both subjectsrequired a large number of throws while wearing knownprisms to readapt the novel-prisms after-effect seen in boththe no-prisms and the known-prisms single test throws. Afterreadapting (F) both subjects retained the learned ability toshift back and forth between no-prisms and known-prismsconditions and hit the target in the centre (G and H). In thesetwo subjects under these conditions of training, readaptingthe novel-prisms after-effect while wearing known prisms(Fig. 5E and F) served also to remove the novel-prisms after-effect from the no-prisms calibration.

The two gaze-throw calibrations that have come to bestored simultaneously are indicated by the arrows below thegraph in Fig. 5. At (A) the subject was using his normal no-prisms gaze-throw calibration (throw in the direction ofgaze). The context of donning prisms (B) caused the subjectto access his known-prisms calibration (throw 17° to theright of the direction of gaze). The known-prisms calibration,however, added to the deviation in gaze produced by thenovel prisms causing a 'double' deviation to the right ofthe target.

Kinematics of throwing after having acquired asecond gaze-throw calibrationFigure 6A and B shows for each subject the angular positionsof head-in-space, shoulders-in-space and spatial location ofeach impact for the 25 throws before donning prisms, thefirst 25 throws after donning known prisms, and the first

A100

NAIVE

PC-12.6 AC-11.1 A C 17.3

BEFORE AFTER

Specificity and storage of eye-hand coordination

B1205

inn

(cm

)C C

Z 50uuo

5a. o»Q

JTA

L.

HO

RIZ

Of

3

O

PC - 3.0

BEFORE

6 WEEKS

PRISMS

# *

AFTER

AFTER: LEARNING COEFFICIENT - 2.4 WEEKS

PRISMS: LEARNING COEFFICIENT = 1.4 WEEKS

NAIVE WEEK 1 WEEK 2 WEEK 3 WEEK 4 WEEK 5 WEEK 6 9 MO. 18 MO. 27 MO

SESSION-POST-TRAINING

Fig. 4 Long-term training and acquisition of two gaze-throw calibrations. (A) Throwing by a naive subject, showing normal prismadaptation. (B) Following 6 weeks' training (see Methods), the subject's first throw was immediately on target upon donning andremoving the prisms. (C) Learning a second gaze-throw calibration over 6 weeks' practice in both subjects. Large filled triangles showthe average of the first three throws upon donning prisms for both of the subjects. Large filled circles show the average of the first threethrows after removing prisms for both subjects (negative after-effect). Small filled circles show the horizontal deviation of pre-prismthrows (15 throws before donning prisms for each subject). The mean± 1 SD of these control throws is shown by the large bars. Thesubjects changed their gaze-throw calibration over time and practice until they threw near the centre of the target for the first throw afterdonning and removing prisms. The subjects still retained the two calibrations 9, 18, and 27 months after the last training session. Therate constants of exponential decay functions fitted to the data serve as a measure of the speed of acquisition of the skill.

25 throws after removing them as measured at the end of6 weeks. In Fig. 6C and D the adaptive changes in angularposition of eyes-in-head, head-on-truhk and trunk-on-arm areshown. These values were calculated knowing the angularposition of eyes-in-space (calculated from known foveationon the target and the prism diopter) and the position of targetimpact (= angular direction of throw; see Methods).

In Fig. 6A, the first subject moved the head left in space

-8° and the shoulders cross-axis left in space ~6° afterdonning prisms. Impact locations while wearing prismsshowed ~3° of adaptation and -2° of negative after-effect.In Fig. 6C the prisms shifted gaze to the left by 17°. At theend of 6 weeks of practice, the angle between gaze andthrow was composed of 9° of eyes-in-head, 2° of head-on-trunk, and 6° of trunk-on-arm.

In Fig. 6B the second subject moved the head left in space

1206 T. A. Martin et al.

150

-100

AC = 34.4 AC - 8.5

= THROW DIR.

t - fGAZE DIR.

\ \ \ I \B D H

Fig. 5 Unexpected introduction of a 'novel' prism (30 diopters, base left, represented as squares) to asubject who had stored the gaze-throw calibration to 'known' 30 diopter, base right prisms (plotted astriangles). Throws without prisms are plotted as circles. The plotted throws are labelled according tothe following conditions: (A) throws before donning prisms, (B) when first donning 'novel' prisms,(C) after adapting to 'novel' prisms, (D) one throw while wearing no prisms, (E) when first donning'known' prisms, (F) after readaptation to previous behaviour while wearing 'known' prisms, (G) whilewearing no prisms, (H) one final throw while wearing 'known' prisms. Throw direction and inferredgaze direction are represented by arrows below the graph. Capital letters beneath the arrows indicateinstances in time during the prism adaptation paradigm described above.

-10° and the trunk cross-axis left in space ~5° after donningprisms. Impact locations while wearing prisms were morescattered than for the other subject, and there was little orno adaptation or negative after-effect. In Fig. 6D, the prismsshifted gaze to the left by 17°. At the end of 6 weeks ofpractice, the angle between gaze and throw was composedof 7° of eyes-in-head, 5° of head-on-trunk, and 5° of trunk-on-arm.

DiscussionThe terms 'motor adaptation', 'motor learning' and 'motor

skill' are often used, sometimes without definition ordistinction between them and usually without knowledge ofbrain parts or mechanisms responsible for their acquisition,storage and retrieval. It has been suggested that the cerebellumplays a role in motor adaptation and in motor learning,although such a role has been disputed. Those who acceptthat the cerebellum plays a role in motor adaptation and

motor learning do not agree on the possible mechanisms. Itis generally agreed that the cerebellum plays a role in thefine control and coordination of movement performance.However, it has not been agreed whether the mechanism isat the level of modulation of motor neuron and muscle firingrates so as to control a particular movement parameter or atthe level of novel combinations of downstream elements soas to create and initiate novel movements, or both (Thachet al., 1992a). In this discussion, we define and relatethese terms and concepts. We contend that they are indeedsystematically interrelated.

Prism adaptation: what is adapted?Helmholtz (1867) first described both the adaptation and theafter-effect in arm reaching movements in subjects reactingto prism-induced lateral visual displacement. Held and Hein(1958), Harris (1963) and Kohler (1964) have shown thatprism adaptation produces changes in proprioceptive, visual,

Specificity and storage of eye-hand coordination 1207

is5

£ 10

-610

f-10

_ 10

0

-6

-10

B

\ ^

BEFORE

•

PRISMS

•

AFTER

-s -

-6 -

as -

-20_ 10

5 -

-5 -

•10

•

• * * • *

BEFORE PRISM8

V

•

i

•

AFTER

QAZE

HEAD IN SPACE

EYES IN HEADM

<aSHOULDER IN SPACE

-20

HEAD ON TRUNK•KIII

ARM IN SPACEtoII

TRUNK ON ARM20II

TARGET

DQAZE

II

HEAD IN SPACE20IS

EYES IN HEAD•20

" ISHOULDER IN SPACE

20'IS

HEAD ON TRUNK20ISr

ARM IN SPACE h•201

TRUNK ON ARM

ISr

TARGET

Fig. 6 (A and B) Relative positions of head-in-space, shoulder-in-space and arm-in-space (equivalent to throw angular error) for each ofthe two subjects who participated in the skill acquisition paradigm at the end of 6 weeks of prism training; the 25 throws before donningprisms, 25 throws while wearing 'known' prisms, and 25 throws after removing the prisms are shown. (C and D) Relative positions ofeyes-in-head, head-on-trunk, and trunk-on-arm have been calculated for each of the two subjects. Relative changes in each of theseangles during the transition from the pre-prism to the prism trials are presented as values in boxes.

visuomotor and motor processes depending on severalvariables that include testing conditions and availability andtype of visual feedback. In addition, prism adaptation hasbeen shown to affect those processes which maintain normalhand-eye calibration (Held and Bossom, 1961; Welch, 1974;Prablanc et al., 1975; cf. Kornheiser, 1976). Kane and Thach(1989) described wedge prism adaptation of throwing objectsat visual targets. We continued these studies to obtain furtherknowledge of its mechanism.

It has long been questioned whether prism adaptation, i.e.changes in bodily movement resulting from wearing laterally

displacing prisms, is fundamentally visual ('felt direction ofgaze', Craske, 1967; visual perception, Kohler, 1964), pro-prioceptive ('felt position of the arm', Harris, 1965), cognitive('straight-ahead shift', Harris, 1974), sensorimotor (Held andFreedman, 1963) or motor. Multi-component models basedon several of these hypotheses have been supported by someinvestigators (e.g. Redding and Wallace, 1988).

Our observations are most simply interpreted as a changein the angle in the horizontal dimension between the directionof gaze and the direction of throwing. In throwing and hittingtasks where gaze has been monitored (Vickers, 1992, 1994)

1208 T. A. Martin et al.

fixation of gaze on the target is the rule. Our instruction tofoveate the target, and the demands of the task performanceto do so, require that the direction of gaze be determined bythe diopter of the prisms. If the adaptation were within thevisual system, one might have expected carry-over of prismafter-effects to all the movements tested. Yet there was nohand-to-hand transfer. If the adaptation were within theproprioceptive system, using the same limb and the same orsimilar sets of muscles, the change might have been expectedto transfer from overhand to underhand throwing. Yet therewas no systematic transfer.

We define motor adaptation as a modification of movementin which three criteria are satisfied: (i) the movement retainsits identity as being of some particular pattern of muscleactivation or end result (i.e., 'throwing') but changes withregard to some parameter or set of parameters (e.g. force,velocity, endpoint, or direction of the movement); (ii) thechange occurs only with repetition of the behaviour and isgradual and continuous; (iii) once adapted, subjects cannotretrieve the prior behaviour; instead they must change theadapted behaviour with practice in the same gradual,continuous manner back to the prior state.

The adaptation of throwing by laterally displacing prismshas these properties. The normal relation between thedirection of gaze and the direction of throwing is an adjustablecalibration: it is developed and maintained through practice.Subjects newly introduced to prisms must adapt their priorcalibration to hit the target. The adaptation occurs graduallyand continuously with practice. Following completeadaptation subjects retain the new calibration and cannotimmediately return to the old calibration. Instead, they mustpractice, and only then do they show a gradual change intheir gaze-arm relationship as they redevelop their priorcalibration.

The outcome of the performance and its adaptation arerevealed in the time-course of change of the spatial distancesbetween the target and the impact location of the thrownobject. In turn, this record is a measure of the direction ofgaze (always on the perceived target) relative to the throwingangle of the arm. The throwing synergy of the arm and handis a complex function across many muscles and joints. Thesynergy is relatively constant across trials (Becker et al.,1990). But what is actually adapted is simpler and is confinedto the relative positions of eyes-in-head, head-on-trunk andtrunk-on-arm. Under the conditions of our testing in whichthe subject initially faces (eyes, head and trunk) the target,all that need vary is the angle between direction of gaze anddirection of throw. This angle can be determined by eyeposition in head, or head position on trunk, or trunk positionon arm at the shoulder or by various combinations of these.Different individuals appear to use different combinationsalthough any one individual tends to use a constantcombination.

Is adaptation volitional or automatic?Volitional strategic corrections can and do occur, particularlyif instructions are not explicit and repeated. Even then,

100

Eo

LU

LUo<

CC

OI

50

-100

P C

-

TRIAL

3.6 AC -

• !*

13.4

TRIAL 19

V

AC

yv

-3.8

TRIAL 36

4 1 — - 3 9 + 20 3*e ( x / 3 8 )

Fig. 7 Graph of the impact locations for a subject who attemptedto estimate and correct for the lateral displacement. Again, throwslanding to the left of the target are plotted as negative and throwsto the right as positive. Prisms-on and post-prism data sets havebeen fitted with exponential decay functions (Martin et al., 1996)by excluding those throws while wearing prisms which lie outsidethe curve. A piece of the curve passing through the prisms-ondata has been transposed to the outlying data points todemonstrate the adaptive trend of these points.

occasional subjects will disobey the instructions. One suchsubject, after the first prism throw, perceived the error andwent against the instruction to 'throw where you see thetarget'; instead he threw to a calculated target location (Fig.7). This subject threw close to the centre of the target beforedonning the prisms and the first throw with the prisms (trialno. 13) hit -40 cm to the left of centre in the direction ofthe deviated gaze. However, the subsequent throw was exactlyon target. Nonetheless, ensuing throws hit progressively tothe right of the target up to trial 19. At this point, the subjectwas told again 'Throw to where you see the target, not towhere you think the target actually is.' The next throw hit-25 cm to the left of the target. Subsequent throws tendedto fall progressively rightward toward the centre of the target.This rightward trend was similar to the rightward trend ofthe throws before the above instruction was given (trials14-19). Evidence of a true adaptation again came when thelenses were removed: the first throw (trial 36) was almost30 cm to the right of centre. The subject required severaltrials to recalibrate his gaze-throw calibration back towardthe centre of the target.

The data suggest that the volitional strategic correctionand an automatic adaptive change may have been workingsimultaneously and independently. Subtracting the throwswhere the subject appeared to be making a volitionalcorrection (trials 14—19), taking the first throw with prismsand the rest of the adaptation data, and fitting them with an

Specificity and storage of eye-hand coordination 1209

exponential decay curve gave a result similar to that of thesubject in Fig. 1. Then, taking a piece of that calculatedcurve (continuous line) and superimposing it (as a dashedline) through the subtracted data points in their originalposition shows how well they lie on the curve that describesa gradual change parallel to that for the rest of the adaptationdata. In this analysis, the offset of the subtracted datarepresents the volitional correction, the fit of that data to thecurve represents the automatic adaptation, and the occurrenceof the two together without interaction illustrates the apparentparallelism and relative independence of the two systems.

However, the question of whether the behaviouralmodification is a volitional shift of strategy or an automaticadaptive process is best answered by the failure of all subjectsto hit the target centre after removal of the lenses. Instead,there is the overshoot in the opposite direction of approxi-mately the same magnitude as that displacement occurringafter donning the lenses. The unpremeditated andunanticipated after-effect is not a voluntary strategy. In fact,the first throw after removing prisms is the only post-prismthrow demonstrably without cognitive interference (Weineret al., 1983). The after-effect is therefore (for the most part)the product of an automatic, subconscious process. Thatprocess compensates for the error introduced by the prismsand can change back to the original calibration only withpractice. The process is adaptive by definition. Finally, thesmoothness, regularity and repeatability of the exponentialdecay curves fitted to the normal prism adaptation datasuggest that this behaviour reveals the working of somefundamental adaptive mechanism rather than that of an abruptvolitional shift of strategy.

Privacy of storage sites for body part and taskPrism adaptation during throwing did not transfer betweentrained and untrained hands. Prablanc et al. (1975) showedthat in reaching, each arm can be prism-adaptedindependently, in opposite directions. Early studies by othershad revealed that hand-to-hand transfer during prismadaptation of reaching is present only under certainconditions: freedom of the head to move (Harris, 1963;Hamilton, 1964), viewing of the moving hand only at theend of a ballistic movement ('terminal exposure'; Cohen,1967), and spaced distribution of practice (Taub and Goldberg,1973; Choe and Welch, 1974). Although our task constraintsallowed head movement and the prism-exposed throwinghand was only briefly seen at the end of a throw, theadaptation effects were specific to the arm used during prismadaptation, as in the study of Prablanc et al. (1975). Oneexplanation for the discrepancy in results may be that mostarm reaching movements are slow enough to be controlledin a closed-loop manner (Jeannerod, 1988), while throwingis too quick for feedback control (open-loop) and must bepre-programmed. One author has suggested that there is adifference in locus of prism adaptation in slow

(proprioceptively controlled) and fast (pre-programmed) armmovements (Baily, 1972).

We also found that there was no overhand-to-underhandtransfer in half of our subjects. In three subjects there wastransfer to the first underhand throw only; in two subjectsthere was transfer to the underhand throwing like the carry-over to overhand throwing. One earlier study suggestedincomplete specificity for prism adaptation between twodifferent movements with the same arm (Freedman et al.,1965). These authors found that after-effects of exposurewith sagittal arm movements transferred almost completely(85%) to transverse arm movements. However, after-effectstransferred from transverse to sagittal arm movements to alower degree (62%). In our study, the adaptation appearedto be more specific to the adapted movement.

Adaptation of one versus acquisition of asecond gaze-throw calibrationIn this study two subjects learned to store more than theone gaze-throw calibration. They showed nearly completeretention after 9, 18 and 27 months. McGonigle and Flook(1978) also demonstrated learning of prismatic effects inhumans who reached with terminal feedback of hand positionto five different targets during prism exposure. However,their subjects learned the prismatic distortion more rapidlythan ours (five training sessions for their subjects versus 20for our subjects to reach pre-prism levels). This differencein time course of motor learning may be due, in part, todifferences in the tasks. First, although they do not analyseit in their paper, examination of their fig. 1 shows that initialdeviation for their subjects after donning prisms was 8 cm(reaching) versus 50 cm in for our task (throwing). Asmaller deviation in pointing errors may fall into pre-prismperformance values more quickly than a larger deviationwhich may require more time for total compensation.Secondly, compensation for prismatic deviation may be morerapid in a task that allows for feedback during a movement(e.g. reaching movement) than a task which is open-loop(throwing). More recently, Welch et al. (1993) have induced'dual adaptation' in subjects exposed to 10 alternating prismexposures on 1 day of training. However, their short, 1 -daytraining procedure and their failure to test for retentionof the new gaze-arm relationship place into question theequivalence of this study to that described above or to thepresent study.

Our 'trick' presentation of the novel prisms in place ofthe known prisms suggests that the context of donning prismsis sufficient for these subjects to access the stored known-prisms calibration. The no-prisms calibration guided thethrow in the direction of gaze; the known-prisms (30 diopter,base right) calibration guided the throw ~17° to the right ofthe direction of gaze. When the subjects donned 'novel' 30diopter base left prisms (which they assumed were the known30 diopter base right prisms) gaze was directed -17° to the

1210 T. A. Martin et al.

right of target, and the throw used was directed by the storedcalibration as if for the known prisms, and hence was ~17°to the right of gaze (34° to the right of target). This errorwas equivalent to the error that which would be induced by60 diopter prisms.

When the subjects adapted to the 'novel' prisms, the after-effect carried over both to the no-prisms performance and tothe known-prisms performance. Yet, readaptation of the onealso readapted the other. After the readaptation, each subjectretained both the no-prisms calibration and the known-prismscalibration. Short-term adaptation in these subjects appearedto influence a common site that was capable of carryingseveral independent long-term calibrations for the behaviour.With prolonged training, multiple calibrations may be storedsimultaneously with each calibration being immediatelyaccessible.

The pattern of the prism-acquired gaze-throw calibrationas revealed in the relationships between the eyes-in-head,head-on-trunk and trunk-on-arm angles was different fordifferent subjects and varied even within each subject. Giventhe variability and inter-dependency of the different parts ofthe body involved, the stereotypy of the throwing results iseven more striking.

The definition of adaptation, as presented above, is notadequate to describe the added capacities of (i) acquisitionand storage of a second gaze-throw calibration and (ii)immediate access to, and retrieval of, either calibration.For this increased capacity others have often used 'skillacquisition' (cf. Adams, 1987). According to Adams (1987)the acquisition of motor skill is a process in which an accurateand rapidly accessible motor performance is learned throughpractice and is often retained for years. This definition seemsappropriate here.

AcknowledgementsOn a suggestion from Richard Held, Nigel Daw and TorstenWiesel developed this task for teaching medical students. Weappreciate the aid of James Lu, Maurice Makram and BradGreger in these experiments. This work was supported bygrants to W.T.T. from the National Institutes of Health(NS12777) and the Office of Naval Research (N00014-92-J-1827), to T.A.M. and H.P.G. by NRSA 5 T32 GM07200and to A.J.B. by the Foundation for Physical TherapyResearch (94D-18-BAS-01).

ReferencesAdams JA. Historical review and appraisal of research on thelearning, retention, and transfer of human motor skills. Psychol Bull1987; 101: 41-74.

Baily JS. Adaptation to prisms: do proprioceptive changes mediateadapted behaviour with ballistic arm movements? Q J Exp Psychol1972; 24: 8-20.

Becker WJ, Kunesch E, Freund H-J. Coordination of a multi-

joint movement in normal humans and in patients with cerebellardysfunction. Can J Neurol Sci 1990; 17: 264-74.

Choe CS, Welch RB. Variables affecting the intermanual transferand decay of prism adaptation. J Exp Psychol 1974; 102: 1076-84.

Cohen MM. Continuous versus terminal visual feedback in prismaftereffects. Percept Mot Skills 1967; 24: 1295-302.

Craske B. Adaptation to prisms: change in internally registered eye-position. Br J Psychol 1967; 58: 329-35.

Darlington RB, Nathan RG. Radicals and squares, and otherstatistical procedures for the behavioral sciences. Ithaca (NY):Logan Hill Press, 1975.

Freedman SJ, Hall SB, Rekosh JH. Effects on hand-eye coordinationof two different arm motions during compensation for displacedvision. Percept Mot Skills 1965; 20: 1054-6.

Hamilton CR. Intermanual transfer of adaptation to prisms. Am JPsychol 1964; 77: 457-62.

Harris CS. Adaptation to displaced vision: visual, motor, orproprioceptive change? Science 1963; 140: 812-3.

Harris CS. Perceptual adaptation to inverted, reversed, and displacedvision. [Review]. Psychol Rev 1965; 72: 419-44.

Harris CS. Beware of the straight-ahead shift. A nonperceptualchange in experiments on adaptation to displaced vision. Perception1974; 3: 461-76.

Held R, Bossom J. Neonatal deprivation and adult rearrangement:complementary techniques for analyzing plastic sensory-motorcoordinations. J Comp Physiol Psychol 1961; 54: 33-7.

Held R, Freedman SJ. Plasticity in human sensorimotor control.Science 1963; 142: 455-62.

Held R, Hein AV. Adaptation to disarranged hand-eye coordinationcontingent upon re-afferent stimulation. Percept Mot Skills 1958;8: 87-90.

Helmholtz HLF von. Treatise on physiological optics, Vol. 3, Sect.29. Translated from the 3rd German edition by Southall JPC.Menasha (WI): George Banta Publishing, 1924-5.

Jeannerod M. The neural and behavioural organization of goal-directed movements. Oxford Psychology Series No. 15. Oxford:Clarendon Press, 1988.

Kane SA, Thach WT Jr. Palatal myoclonus and function of theinferior olive: are they related? In: Strata P, editor. Theolivocerebellar system in motor control. Berlin: Springer-Verlag,1989:427-60.

Kohler I. The formation and transformation of the visual world(translated by Fiss H). Psychol Issues 1964; 3: 1-173.

Kornheiser AS. Adaptation to laterally displaced vision: a review.[Review]. Psychol Bull 1976; 83: 783-816.

Martin TA, Keating JG, Goodkin HP, Bastian AJ, Thach WT.Storage of multiple gaze-hand calibrations [abstract]. Soc NeurosciAbstr 1993; 19: 980.

Martin TA, Keating JG, Goodkin HP, Bastian AJ, Thach WT.Throwing while looking through prisms: I. Focal olivocerebellarlesions impair adaptation. Brain. 1996; 119: 1183-1198.

Specificity and storage of eye-hand coordination 1211

McGonigle BO, Flook J. Long-term retention of single and multistateprismatic adaptation by humans. Nature 1978; 272: 364-6.

Prablanc C, Tzavaras A, Jeannerod M. Adaptation of the two armsto opposite prism displacements. Q J Exp Psychol 1975; 27: 667-71.

Redding GM, Wallace B. Components of prism adaptation interminal and concurrent exposure: organization of the eye-handcoordination loop. Percept Psychophys 1988; 44: 59-68.

Taub E, Goldberg LA. Prism adaptation: control of intermanualtransfer by distribution of practice. Science 1973; 180: 755-7.

Thach WT, Goodkin HP, Keating JG. The cerebellum and theadaptive coordination of movement. [Review]. Annu Rev Neurosci1992a; 15: 403^2.

Thach WT, Goodkin HP, Keating JG, Martin TA. Prism adaptationin throwing is specific for arm and type of throw [abstract]. SocNeurosci Abstr 1992b;18: 516.

Schematic model of short- and long-term adjustments of eye-handcoordination in throwing [abstract]. Soc N.eurosci Abstr 1995;21: 917.

Vickers JN. Gaze control in putting. Perception 1992; 21: 117-32.

Vickers JN. Gaze control in basketball foul shooting. In: FindlayJM, Kentridge RW, Walker R, editors. Eye movements, processes,mechanisms and applications. Amsterdam: North Holland, 1994.

Weiner MJ, Hallett M, Funkenstein HH. Adaptation to lateraldisplacement of vision in patients with lesions of the central nervoussystem. Neurology 1983; 33: 766-72.

Welch RB. Research on adaptation to rearranged vision: 1966—1974. [Review]. Perception 1974; 3: 367-92.

Welch RB, Bridgeman B, Anand S, Browman KE. Alternatingprism exposure causes dual adaptation and generalization to a noveldisplacement. Percept Psychophys 1993; 54: 195-204.

Thach WT, Martin TA, Keating JG, Goodkin HP, Bastian AJ. Received December 29, 1995. Accepted February 16, 1996