This article was downloaded by:[BIUS Jussieu/Paris 6] [BIUS Jussieu/Paris 6] On: 6 June 2007 Access Details: [subscription number 772812101] Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Ergonomics

Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713701117

Control of Performance in a Multi-element Repetitive Task a

a

C G. Drury ; E. N. Corlett a Department of Engineering Production, University of Birmingham.

To cite this Article: Drury, C G. and Corlett, E. N. , 'Control of Performance in a Multi-element Repetitive Task', Ergonomics, 18:3, 279 - 298 To link to this article: DOI: 10.1080/00140137508931462 URL: http://dx.doi.org/10.1080/00140137508931462

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. © Taylor and Francis 2007

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Control of Performance in a Multi-element Repetitive Task C. G. DRURY* and E. N. CORLETT Depart~nontof lingineering Production, Universit,~of IJirrninghnm This rcsearch is the latest part of a series of studies of the functioning of higher-levcl control loops in the motor-control hierarchy. Previous experiments had shown that this control level could be studied by analyzing and modelling the sequential dopendcncies betwcen pcrforrnance on successive cyclas of a repetitivo motor-control task. The presont rcsearch applies the previously developed mcthods to a repetitive motor task having four clemcnts per cyclc, rather than one or two clemcnts a s studied prcvionsly The effccts of visual vs blind control and practlce a t tho task arc examincd. Tho previously-doveloped first order autorcgressive modol again fittod the results wcll and the effects of vision and practicc arc examined in terms of parameters of this model. Tho sequential dependoncics found in this task were g e n ~ r d ylower than those found in previous studies. I t was also fc~undthat soqucntial dependencies docreascd with both visual control and practicc a t tho task. Conclusions from both tho current and prcvious rescarch aro prcscntcd.

.

1. Introduction The hierarchical nature of skilled human performance has becn noted and cliscusscd many times (e.g. Welford 1968, McCormick 1970, Reishon 1967). Approaches to thc modelling of human skills are based on the assumption thibt the human operator has a hierarchy of control systems, each taking its input (or cotntnancl signal) from the level above and in turn controlling its oaput a t this value by issuing command signals to lower levels of the system. One particular level which has received widespread attention is human control of accuratc movemcnts. Enough is known about how an operator makes discrete tnovemeuts (e.g. Welford 1968 op. cit.) and performs continuous tracking tasks (c.g. ICelley 1068) to produce quantitative performance models which, cvcn if less than pctfect, are readily applicable to the design of human tasks (e.g. Fogel 1 963). Howcver, there is very little data which relates this level in the hierarchy to acljacent Icvels-according to Reishon (1967, op. cit.) a necessary condition for producing a complete description of human tasks. In particular thc relationship between this ' motor control ' levcl and the next highest level has received little study. Welford (1966, op. cit.) quotes four such studies which show some tnodification of the performance of the motor control level by higher levels in the hierarchy. !Phe experiments describcd in this paper are part of a series conducted a t the University of Birmingham which studied explicitly the modification of performance on successive cycles of repetitive tasks. hevious stuclies ( h u r y 1969 a, b, 1971) have shown that in a, wide variet.y of repetitive tasks requiring accurate movemcnts there was successive ~nodificationof nerformance from cvcle to cvcle of the task. Whether the performance measure sttessecl in the experiments was speecl or accuracy, serial correlations between successive eerformance measures showed that the subiects were controlling their level i f performance from cycle to cycle. &rial .?

Now a t the Depnrt~nentof Industrial Engineering, Statc University of S c w York nt Buffalo, 13uffal0, Now York, U.S.A.

0

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

cor~dationsin pcrfornii~~ices had bcen observed earlier by Laming (196'7) and Abrrmzi ( I 952, 1!15(i) altllougli Liming used a probabilistic task rather than a ~,cl)ctitivcone itntl Abrmzi only meiwurcd serial correlations where inspection r l f t h c d i h suggested they may occur. Ellis and \\lade (1968) calculated the l i d two autocorrelation coefficients of scores in a rcpctitive tafget-aiming task itnd found them generally positive. They concludctl t h a t the subjects used inform:~tion only from t h e previous trial t o control perforniancc. Tlic tcchniquc used i n the present serics of studies mas based on the ci~lci~li~tion of autocorreli~tion fimctions for the sequence of pcrformancc ~)itri~mctcrs obtnincd on successive cycles of a repetitive task. The i ~ u t o uorrclt~tion function is ;I, serics of Pmrson product-moment correlation cocfficicnts measuring the corrclntion between pcrformance parameters sclmfittctl by particular numbers of cycles. !I!hc number of cycles separating tho l~crformi~nce pttrarneters is known as the lag. Thus tlic autocorrelation I ~ n c t i o nis it serics of correlation coefficients mcasurcd at cliffcrcnt lags from ;I lily of I (rcprescnting correlation between s~~ccessivc performance pimmietcrs) w l s . :I:n thcsc stutlics it maximum 1;~gof I0 has been used because little :~utocor~clntion was found in Iitgs greater than 10. '.l%e previous studies cximiitled tlic autocorrelation function of pcrformimcc ~ ) i ~ r i m c t e in r s ;I nunibcr of tasks atid found that in each case t h e subject bclii~vcda s if lic were using his estimate of performance on one cycle of the titsk to control his 1icrform;wcc on the next cyclc. 'Chis nicthod of control can be tlcscribcrl m n t h c m a t i c i ~as ~ ~ n~ Firs!-Order Auloregressive proems. Let tlic subject's estimate of his ]?erforrnancc on thc d l cyclc of a task be S, and tlittt on t h e prcvious cycle Sf,-,. If thc subject attempts t o reproduce his on t h e current (nth) cycle but his reproduction ~wcviousperfor~nimcc(S,-I) 1)roccss is snbjcct t o sonic random error Z, this can bc written tlic subject's dcgrcc of 1.cIia11cc 011 previous \vhcrc X is it ~ o n s t i t n tin~1ic:~ting pcrforn~;tncc. 8uj)lwse further tlii~tthe snbjcct's estimate of his perfor~niuicc on ci~c11cyclc, XI,, is not tlic sitme MS tl~itttnci~suredby the exlrcrinicnter, Y,,, but is snbjcct t o sonic ri~~icloni error, Un, thus:

.

Illhcsc cqnntions dcfinc ;L ilfod$ed Autore~gressive l'rucess (Cox and Millcr I ! ) :St can be slio\vn thirt for such a process the imtocorrc1;ttion function i ~ :L t Iitg o f s (tlcfincd i ~ t,lic s product-moment correlation cocfficient betwcen itnd is oihSf0~S > I Autocor~dittiona t lag of (s) = r(s) = 1 fors=0 ) I , ,

\vlicrc t l ~ ct\vo rlimcnsionlcss constants, a and X can he cxl)ressccl in tcrrns of 1 . 1 1 ~v;~ri;~nccs of the three p~~occsscs, S,U i ~ i dZ its:

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Co?~trol of I'erformcmce in a Ndti-eloneitt Repetitive Taslc

281

Tlie autocorreli~tion functions found in the prcvious studies \wrc all i ~ d e q ~ ~ ; t t efitted ly by equations of the form given in Equittion (3). This supportecl the conclusion of :lCllis ant1 \Vacle (op. cit.) t h a t only infornlation from the immediately preceding cycle is used to control current performance, but took t,his assertion n stage further by giving a quantitative motlel of performance. The previous results showed t h a t the two parameters a and X mere cliffere~rtially affected by changes in experimental conditions. \\'liilst the valuc of X remained substantially constunt (at approximately 0.8) between conditions within experinients, and between experinients, the value of n varicd over a range of 0.1 to 0.6, accounting for most of the vitriance of'the autocorrelation functions. Tlie model indicates t h a t X reflects the extcnt of tlic subject's reliance on previous performance (Equi~tion( 1 ) ) whilst a represents the precision with which lie estimates this performance (Equation (4)). One important finding was that in a11 accuratc niovement task with two cletnents pct cyclc, autoconelation functions for each element and cross correlation functions between elements were of the same size. 111 conjunction with the abovc moclcl tlris implied tliat the subjects used the performc~nce parameter from either elcmcnt of the previons cycle to control performance of the currcnt cycle. Autocorrelation thus measures t l ~ csystenr which controls the level of ejJorl involved in performing the elements of the ti~sk. This is a higher lcvel in the hierarchy of control than thc control of muscular response ~~suitlly examincd in tracking and terminal control studies. 'I'his was tlie original reason for asserting t h a t a higher levcl in tlrc hierarchy than t h a t examined in most normal skills str~clies\vas indeed thc one bcing exitmined by tlris set of studies. r7 .I lie present study \vas designcd to achieve a number of objectives: ( a )Confirm t h a t auto- and cross-correletion functions bct\veen clc~ncnts within cycles in a multi-cycle task are of the same magnitude. (0) Ilctcrmine wllethcr extending the study to a, more realistic work cyclc of four elerncnts per cyclc itffected the size of the auto- and crosscorrelations. (c) 'I!o determine how the control lcvel bcing studied responds to increased practicc on the task. ( d ) To clarify a, suggestion from it11 eitrlier experiment (Drury IWi9 a ) t h a t visually controlled movements sho\ved less autocorrelittion t l ~ a n non-visually controlled nlovcments. 2. Methods

An appartttus wi~sco~istructeclt o enable a pre-prograninic sequence of lights t o appear on a vertical boitrd. The eight lights were wranged a t 10' intervitls on iln arc of a 10" radius circle. li:wh light WLS '$! in diameter and their order of lighting was controlled by a tape in an S channel titpc reader. At a distance of 3" bclow eitch light tliere was a, 4" diamctcr brass target its shown in Figure I . Tllc subject's titsk mas t o move his right hitntl to each new light its it appeared and wait tlicre until the n e s t light came on. I n the " with-vision " condition tlie subject held ;I pointer contitining a plioto-

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

282

C . O. D r w y and E . N . Corlett

diode which switched a timing circuit when tlie pointer left t h e old light and \vlicn i t rci~checltllc new light. I n t h e " without vision " conclition movements wcrc m d e bot\vcen the brass targets rather than the lights ancl a touch-sense s\vitch wi~sused in place of the photo-clioclc; a canvas screen was fixed as sho\vn in :Figure I. t o prevent tlie subject from seeing either the targets or his own hand and arm. The two mcthocls of recording are shown in Figure 2 : recordings of thc timcs of the targets appearing, t h e subject leaving the old tawget nnd going to the new target were recorded on punched paper tape using ;In i~utomatictimer and rccorcler (S.E.l17.A.R.)so t h a t Reuclion Time ancl ilfovcn~cnt7'itt~cfor each element of tlic cycle could be measurccl. STIMULUS LIGHTS

POSITION OF CANVAS SCREEN

TARGET BUTTONS

Figuro 1.

Figuro 2.

Layout of apparatus.

Ncthorls of rcsporlding.

Although tlic apparatus s constructed SO tliat cigllt lights could be switcllctl on in any order, for this experiment only the middlc four lights (Nos. 3, 4, 5 ancl 6 from the subject's left-hand side) were used t o give a work cycle of four clements. They appeared in the order 5, 3, 6 , 4 in each cycle a t 2 scc intervals and one experimental trial consisted of 50 cycles, taking about 8 win. 'I1\vo experiments were conducted-' with vision ' mcl ' \vithoub vision '. I h c h usctl eight subjects \\rho performed two experimental trials per day on

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

e d of~three consecutive days. Irour of t h e subjects from the first experiment werc d s o subjects in t h e second. All t h e subjects werc staff or resenroll students of t h e Department of Engineering l'roduction whose ages ranged from 20-31 years. I n all, eight male a n d four female subjects were used. p\To pre-training was given because one of t h e objectives of t h e study mas t o find t h e variation in autocorrelation with practice.

3. Results 3.1. A?zcdy.sis of Elemenl Times

Although t h e autocorrelation findings are of major interest in this experiment, the overall results should first bc compared with other, similar, strdies. Both experiments basically measured reaction times and movement times for different lcngths of movement and can thus be compwetl with studies such a s those reviewed in Fitts and Posner (1967). Average reaction time and movement time were calculated for each movement and plotted against Ilrelford's Intlez of 1)ifficulty (\trelforct 1960) for each movement in Figure 3.

INDEX OF DIFFICULTY, BITS P E R MOVE

Figure 3.

Variation of mean reaction Liims and movement times with index of difficulty.

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

,, lablo 1. Fit of nnto-r~grcssi\~c model to betwec~i-cyclcnatucorrclation functions

:I'arn~rrctcrfitted \\'it11 vision:

Kcnction tilno Jlovement tinro Cyclc timc

0.

h

0.502 0.320 0.309

0.685 O.i(i(i

.-

Correlatio~r betwcen model and data

0.G;

0.983 0.903 0.99!1

WITH VISION

- - - - .- .WITHOUT VISION

LAG

17iguce 4.

Avcrngc autocor~dntionfr~nctimrsof reaction times (lag is in cycles).

-

WITH VISION

- - -----.WITHOUT VISION

LAG

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

'I'hc ci~rlierstudies h:~dsho\\m t h a t h changes very little with experinlentill conditions i ~ n dt h a t N changes considerably. Similarly, in this experiment whcn tho meitn t~utocorrelation functions for cach subject were fitted by I k p ~ t i o n( 3 ) , the coefficients of variation for N and h were 93% a n d 14% (with vision) i ~ n d7fi% and 11% (without vision) respectively.

With Vision

- - - - -- - Withooi

,J0

1

2

I'ig~lro 0.

4

3 T R I A L NUMBER

Virion

5

6

,

l!!ffcct of prnctico on first autocorrclation coeilicicnt. 7

I ,

I he chi~ngcsin amount of ilutocorrclstion with increasing precticc iLre shown in Ifiigurc (i for total cycle times: t h e gra,phs of total Reaction Times and total ~llouententIl'in~esiwe similiw. The parameter chosen is the product of a: and h which gives the fitted value of the first autocorrelation coefficient. This was chosen rather than the mei~surcdfirst autocorrelation coefficient (r,) as the virlucs of N i d h arc c;~lculi~ted from ull thc r, values 1 1 for s = 0

(this rcprcscnts tllc mcan performance on the nth cycle of the task) wlmt is tllc antocorrelation of thc process {T,}? By definition rp.(s)=

Covariance (T,, T,,,) var (T,)

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

C0)~1ro1 of Performume in u 11Iulti-element Repetitive Tusk

291

Now, Covariance (T,, T,+,) = E ( T , T,+,)

1

=k2E

[Z C

i=l

( Yi,nYi,n+s)+

2

I

( Yi,nYj,n+s)

all i + j

as the { Y , , ,} are iudcpendcnt E( Yi,,Yj,,,+,)= 0 for all i # j 1

. . . Covariance (Tn,T,+,)= - E k2

2" (Yi,nYi,n+s)

i-1

1

=-

k2kl

1 --

-

k2 '

...

Covariance ( Yi,,Yi,,+,)

k . aIr2. d s

Cov(T,,T,,+,)= -. 16

(from ( 3 ) )

(6)

Also

Thus using results 6 and 7 in Equation (5) gives

Thus i.e. for I; indepcnclent elements, the autocorrclation of the cycle means is equal t o the autocorrelation of the elements.

APPENDIX 2 Given a process { Y , } such t h a t

E(Y,)=O var ( Y , ) = oY2 and autocorrolation at lag ( s )is

r d s )=

aXS

for s > 1

1

fors=O.

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

202

Then if

(this rcprcsents i~vcragingsets of k consecutive items) then required to prove that autocorrclation function of T, =(

s )3 (

for all k 3 2

8 )

Two cases arise, s = l and s > 1. I n the former case, s = 1

For s > 1 C O V ~ ( T ~ ,= T&[ahso,-9 ~ ~ + ~ ) ahs-'up2

+ ahy-'oyZ+ aX80y2]

+

= $ U ~ . ~ X ~ -A]. '[~

Also Var(T,,) = &E(Tn2) =P [ Y n P , ,

+ 2 y,, yn-1+ y,-1

+ ah).

= +or2(1

Thus for s = 1

yn-11

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Control of Performance in n Nulti-element Repetitive T a s k and for s > l

Thus

> 1 for O < a , A < l Thus assertion (4) is true for k = 2 . For proof by induction assume that assertion is true for some value k > 2, then an expression will be derived for the next higher value of k , that is k + 1 . Firstly,

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Threc cnscs arisc according as s > k , s = k , or s < k. For s > k all s - k > 0 and thus

for s = k

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Control of Pcrfortt~ntscein o ilftclli-elen~entRepetitive Task

295

for s c k , let s = k - j , ( j = 1 , 2 ,...,k-1) then

Thos, Equation (8) is evaluated for all conditions of s and k. Thus combining Equations (8) and (0)

we wish to prove t h a t given

then

i.e. Cov,(T,,T,+,)

+ A 2 dsVar,(T,)

+ aXS Varic(T,,)+ d B

(assuming Var,L+,(T,) #O)

B u t from (14)

Inequality (15) will be proved using the values of A and B derived from Equations (8), (9), ( l l ) , (12) a n d (13).

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

C. C.D r w y and E. N . Corlett

2!)G

:For s > k condition (15) implies

AS

(-)

and (1 - hk-I) me boot positive, condition (15) is true for s > E .

li'or s = k condition (15) implies

as both

(--)

1 -X k + l

I-h

and (1

-h8-l)are positive condition (15) is true for f = k.

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

Control of Perfornmme in n illulti-clement Repetitive T a s k For s < k condition ( 1 5 ) implies

replacing cach torn by the series of which i t is a sum, we have k-a 8+l k 1+aA z h i + a h z h i - a 2 h s + ' z h i 3 0 i=l

i=l

i=l

as a, h < 1 this is true if

zAi- 2 h i ) > 0

k-s

k

i=l

i=s+Z

All terms are positive and thus condition (15) is true for s < k . Hcnce under all conditions of s > 1, k > 2 condition ( 1 5 ) is truc and thus assertion (4) is true for k + 1 if i t is true for lc. As i t is true for k = 2 i t is true for all k > 2. Cot orticlo prdsonte la dcrnibro partio rl'uno s6rio do rcchcrches snr lo h ~ t t c t . i u t n ~ c z ~dcs ~c~~t buuclos do commanrle do huut nivcau dnns la hi6mrchio dos commnndcs ~nutriccs. ],as oxl,6rionccs ant6riouros ont tnontrB nrro co niveau do commando noat Btre 6turli6 cn analvarmt nnr dcs mo,lblcs lcs rL6penrlanccs sdqr~entiollcsentrc lcs porformancos arrx divers cyclcs snccessifs dana uno thchc r6pBtitivo do commando motrico. L a pr6sento rochercho appliquo les mOthodos 6labur6cs pr6c6rlomrncnt 6 uno tnche r6pdtitivc rnotrice comportant quatre dkmcnts par cyclo a n lieu du un ou dcux dont il avait 6t6 tenu cornpte ilans Ics oxp6riencos ant6ricores. On 6tudio les clTcts (lo l a connnando 6 vuc cornpar60 8. In cummanrlo sans vision, ainsi quo l'cntraincmcnt h la thcho. Lo modi?lc autordgrossif d u prcmicr ordro 61abor6 pr6c6dcrn1ncnt pcrmet n n bun ajostomcnt dcs do1m6os. Los oflots do la vision ot d e l'entreinernent sont 6tudi6s en t a n t que parami?trcs do co modblo. Lcs d6pendnnccs s6quontielles, dans cette tbchc, sc sont av6ri.c~6tro g6n6rnlcrncnt plus faihlcs quo d a m les exp6riences ant6rieures. On a Bgalen~cntrnontr6 que ces ddpenrlanccs s6quentiolles ont 6t6 dirninubes sous l'effet d e Is, commande par Iu v a c e t d e I'cntrainement. L a conclusion porto sur les r6sultats des recherches pr6c6dentcs e t do l a rochorche relatee dans cot article.

Downloaded By: [BIUS Jussieu/Paris 6] At: 10:04 6 June 2007

2!)8

Conkrol of lJerfortrznwce in u Nulti-element Repetitive Tuslc

Iliusu Vorschung ist dur lotzto Toil oinor ltciho von Untcrsuchungen der,Funktion von KontrullsL(hIuifc~iliiihoron C r u d o ~in dur Hiorarcliio motorischor Kontrollo. Vorhergehonde Exporimonto hal~ungwuigt, diiss ~licscrGrad dor Kontrulle durch Analyse und hlodoll-Darstollung dcr Abhiingigkoits-Su,luo~~zenzwischun Leistung und succossiven Kroisen oinor ropotitiveri motorischan Iluthutlcn auf oinu ropotitivu rnotorischo Aufgabo an, eino Aufgnbo mit 4 Elomonten jo Z y l i l ~mohr ~ ~ , nls IUIP cin o d w zwoi E1cm011to wie bei vurhcrgegangonen Stutlicn. Dio IVirkungon ~ ~ I I L . ~ V ~ B I I O I I C I I I < U IiL~~ ~Vcrgloich n~ O I I U zu blindcr und tlio Praxis bei dcr Aufgabo wurdon untorsucht. Ilns v d ~ u ontwickulto r anturugressivo hlodell orston Gradcs stimmto wicdcr g u t mit don ltesultrrtcn iiborui~~.I)io \\'irlamgux~ dos Schcns und dcr Prnxis wurdcn untersucht und Pnramctcrn dicscs Alo,lalls nusgdriickt. Ilio Abhii~igigkoits-Soq~~cnzon dieser Aufgabe lagen im allgomeinen nicdriger als (lio in don vorhcrgcgu~lgonun U~rtcrsuchungongcfunneden. 13s zeigta sich auch, dass dio ~\l~l~iingiglzoitssoq~~enzc~~ sowohl bei visucllcr IGmtrollo nls auch bci clcr Praxis mit dor Aufgabo ~ ~ I ~ u r i h ~ nI)iu u n . Folgonmgon nus clcr laufcndcn und fruhorcn 1:orschung worden vorgestcllt.

References Auauzzr, A , 1952, lVork dlaaaurononl. (Now York: C o ~ u n l n rUN~VEIISITY ~ PRESS.) r\ulluzzl, A,, 19.56, Ilrork, IVorkers and 1Vork ~lfeaaurcs&cnl.(Now York: C o ~ u . l l u 1U~~ l v E n s r T ~ 1'11.ss.) I j ~ r l s ~ r oIst ,. J., 1967, PruLlon~sof T~rnliDoscription in Proccss Control. Erpnomics, 10, 177-186. H. D., 1965, ' Thoory of Stochastic Procosscs '. (Now York: WILEY.) Cox, 1). .R., a ~ l d~IILLKII, l'p. 276-284. D I ~ U I ~C.Y ,C., 1968, Soqnontial offocts in rcpctitivo motor tasks. U ~ ~ p u h l i s h oPh.D. d Thesis, I h ~ m r l n & oof~ lI;'nginceri?~~ Prod~cclio~i. Uniucrsily of Uinningham, En,glund. I)~run~ C.~C.. . IUKI a , Solno factors affocting tho accuracy of control rnovcmcnts. I n Ergorimnics i n ~llnchincDcaign, 1, (Conavo, ILO). I)~iu~r\-, C. C., 1969 11. ' Soq~mitialControl of a Simplo Itapetitivo Task '. Paper lo IEBE.C~'I~IIIIS Inlcr~~ntional Sytnp. on J l a ~ ~ - l l l u c l i i nSyslo,ne, a Camhridgo, I3ngland. J ) ~ i u ~ i vC., C., 1!)71, ' Soquontial Control of ltcpetitivo Motor Tasks '. Paper lo I~dcrmlional Spnposiuv~ron Vista1 Infomalion P~occssing,Sofia. EI.LIY,31'. J., a t ~ lWAIIE, 51. C., 1963, ' Intor-trial ltosponso Strategy and Oporativc T i ~ n o ISstirnatiuns ', Peyc/~o,mnicScience I I , 6, p, 203. 1'1.r.r~. 1.'. hl., and ~'IITEIISOS,J. It., l!)64, Infurrnation capacity of cliscroto motor rcslmnscs. 67, pp. 103-112. Journul I ~ ~ p c r i ~ n e ~/'~ycli~lo!/,y, tla1 I'ITTS, 1'. >I., and ~'OSNER, 11. I., 1967, H t ~ m u nPerformance. (Bclmont, California: U n o o ~ s COLIP~ ' ~ ~ U L ~ S I ICOBIPANY.) ~NO I"ooa~, I,. J., 1963, 13iolcchiwlo~:Concepla and Applicalions. (New Jcrsoy: PILESTICE-HALL.) 1'. 21!)-272. I