The Impact of Breaks on Forgetting When Performing A Repetitive Task

The Impact of Breaks on ForgettingWhen Perfornling A Repetitive Task

ii

1.•

SHLOMO GLOBERSONCullege of 8usiness Administration

)Northeastern University (Visi[(ng\istrat;onזess AdmiחGraduate School of Busi

iversityחTel Aviv U

NISSAN LEVIN IGraduate School of Business Admirjistration

Tel Aviv Univcrsity

AVRAHAM SHTUBring~Department of lndustrial Engi

el Aviv Universityז

-Abstract: Commonly used leaming curve models assume that a repetitive task is performed continuously, disre-secutive repetitions. Since these breaks generate forחk periods between co;ce of possible breחg the existeחgardi

.g curve rnodelsחance will be inferior [0 the perforrnance forecasted by typical learniחg, actual perfoחgettit dcsigned [0 investigate the nature of forgetting inחalyzes a laboratory ex.perimeחThis paper desc,ibes and a

at the degree of forgetting is a function of thett indicateחt. The resu!ts of the experirneחrneחg enviroחa workid the level of experience gained prior (0 the break. The study investigated the irnpact of breaksחbreak length a

ce deterioration due to the break.s v,.·asjust a few percentageחnarge of one to eighty two days. The perfoחwithin a ratage points for the longest breaks. A power cu[\'e was identifiedחd up to 70 perceחts for a single day break aחpoi

break length, perforrnance time before the breakחas a proper forgetting model to depict the relationsrip betwee..and the degree of forgetting

))T(lt pa.rameter (aחg poiחira - a st.aomists haveחd ecoחeers aחgiחFor more th<lnfifry years. e•task perfom1-חחסg the e[t-ect of task repetitioחstudyiחbee

alyticalח modeL iחance times. Wright [13] developed an a-cחwer fupced as aחce time is represeחwhich task performa-of the cumulative number of repetitivns, with the folחtio

:g formחlowi

m - a curvature parameter.

de\'eloped since theחMany mathematical models have beeg \vork of Wright, with the aim of improving theחpioneeri

iclesתce times. Survey aהnaתability to forecast future perfo-d Belhaoui [2], deחby Yelle [14], Hancock & Bayha [6], a

.umber of these modelsחscribe and analyze ae modeLs stems from theףOne difficulry with most of L

ttent manufacruring system isחfact that although an rnteconsidered, in which the leaming process of a given t.ask

-terחpted as other tasks are perfonned, theeffect of imteiis-'ruptions is either ignored or it is assumed to be in accord

boutדownkceprua1 models. Little isחce with some coחagth (theחof break leחctioחature of perfonnance as a fuחthe

Iforgening process).

(1)T(s) = as~

where:

sחumber .of repetitioחs - the cumulative

T(s)זjזioח - the performance time of the sth repe

Rcccivcd-ח Novcmbcc 1987: rcvis<:d April 1988. Handlcd b)' Ihe Dcpa.jcasurcm.ntl,\,jclhodsJErgonomics~rkחmcnt of W

.cember~sן989 Dחsactioח\x;r 4, IIE TraחUI21,אVo\umc:••OOxOO @ 19R9 ·'IIEב.X/89ISךRIך4-חO376Page 162

two consecutive repetitions and the perfonnance level?achieved prior to the break

This study concentrates on the use of two models, the•power model which has been described above, and the Stan

ford B model (see Garg & Milirnan, [4]). The Sta1lford B:cl is of thc following formו>וחU

(2)T(s) = a(s+L)--

whereמ the me.aning of (1, s and m is the sarne as in equatio1) abo\·e. The additional pararneter, L, may be defmed as(

•previous experience, expressed in terms of equivalent number or repetirions previously performed. L = 0 indicates

.the taskמסprevious experienceסthat there was nEven though the use of Lrequires an estirnate of three

-rather than two parameters, to its advantage this rnodel incorporates forgetting into le.aming, as demonstrated by the

wing exarnple: suppose that a certain task was repeatedסfoll11:, tirnes before a break took .place, and after the break it

eמwjls repeated again for aoother set of 16 repetitions. Omay count the first repetition of the second set as the firstrepetirion, disregarding previous experience, that is, assume

•that s = 1. When the Stanford B model is used, it is expected that the pararneter L will therefore obtain the value

= 0 for total~o forgetting QC{;urs,and a valueחL = 16 if.etting~~for

-g section de.scribes the experiment. It is folwThe follo.lowed by an analysis and a discussion of the results

Thle Study

-A data entry office environment was simu1ated in a miistedםcrocornputer laboratory (Asseo, [1]). The work co

of key punching forms containing personal information onmpleteסa company's ernp1oyees. Subjects were asked to c

16 forms in the flrSt session and another 16 forms in themסvaried frחסsnd session. The break between the sessiסsec

120a single day up to 82 days. For statistical purposes thestu<Ientsס6 who participated in the study were divided int

.1nearly equal size groups, as described in Tab1e

me CategoriesTand Breake2ble 1. No. of Subjז

654321.00Group

202019201623No. ofparticipants54-8226-5316-258-153-72•1)ge (daysaBreak time r65.140.819.111.56.11.7)Average brea.k time (days

d prograrnmed. inapedסA sottware package was develPascal. The package included three modules: a regisuation

,rnodule in which each participant entered his namc. age-e partictaddress. etc.; an introdudory module in which;ipant was informed about the experiment and his role in it

fthe prograrn, which sirnulated tbe taskסand the main bodytest and tneasurementמrrectioסrmea both a data cסand perf

.f performance timeס

.דרך

nd:A literature search of the engineering. managementeconomics fields revealed that suggested forgctting models

n:g curve models. Sleedmחmi:are silnilar in nature to lefter a break. is:,rnemory:11ןג] suggested that the residu[

ce tirneחd pcrformaחn of the Iength of the brcak aסa functi1\VC11(')ז ;\11(II~lrc:lk t()()k 111:I<.:C. C:lrlןcli.lrc tl\c!רill\I\\cJi;llcly

developed a learning- forgctting-Ic<lrning rnodcl in which3נ[forgetting is rnodeIled by a curve sirnilar to the learning

nסtheir model, forgetting is a functi(סcurvc. Accordingtheסgth and the performance time prior tחf the break leס

-rgetting curve is assurned to be of an expoסbreak. Their f-ature. Sule [12] proposed a similar. approach, asחentialח

g that the same leaming curve may be used to modelחsumivernentסd forgetting. Learning causes rnחg aחboth leami

down the curve, whilebreaks in the learning process causet paper by Globerson an,iחt up the curve. A rcceחmoveme

Levin [5) presents a conceptual nlodel hypothesizing that-f several factors, inסthe forgetting process is a function-cluding tumover, cornmunication and docurnentation. Un

real data collcctedחסg models which are basedחIike leami•ceptualחinindustry. all relevant forgetting rnodels are co-f interסmonitor and analyze the irnpactסsince it is difficult t

.nsסrupti-of f)syחtioחThe issue of forgetting has attracted the atte-m~f m(סchologists for quite a while. Researchers in the area

differentiated between7(נ[.,ry decay (e.g., Hulse, et alסng-terrn-ח rnemory decay. Shoסןtetm mem.ory andח-sho

ry has a limited capacity. If the information storedסterm memit will be totallyסot retrieved within 30 seconds or Sחis

.forgotten-חס Typical studies concerning forgetting and shts were~~ry involved experirnents in which subjסterm memrepeatסasked to mernorize a list of nonsenSe sy11ables and !t

n a few~o more thחfסthem after specified time intervals-simסts. Such experiments are obviously not a gחmome

eiither theסulation of organizational reality with regard t.nature of the task or the break 1ength

-t to forgetחry research is rnore relevaסLong-term memtsחizational environment. Typica1 experimeחting in an orga

rea involve repetitions of a memorized Iist of words:in this.after a prespecified break length. typically up to a few days

ing that~Klatzky [8J reports the results of such a study shory decay is a power function of the break ler)gth. Forסmem

exarnple, subjects forget 55% ofthe words within a six hoursd 80% within 72 hours (three days). Whileחbreak time, a.

-such break Iengths between consecutive sessions m:).ybe reIt to organizational environments. the nature ot· the taskחeva

be used~heless. the results of such studies maחt. Neveסis n:. which~tors (lf tllC n<llure of thc forgclting curv:<lSindic

.rmסof a power fסis alsAlthough the results of the psychologicaI studies are not

st engineering purposes. the researchסmסoo relevant t!

f a contro!led laboratory experiment could beסmethod-priate forgetting model. Takסadopted in dcveloping an appr

ing this approach, a Iaboratory experirnent was designedforecastסn: Is it possible tסswer the fol1owing questiחto a

n ofbreak Iength betweenסtask performance time as a functi

<nחn"".-tiהeC"p.mN-r IQRQ. TTF. Tr[

,,~..•

j,.ו

Page 163

-in order to avoid the impact of fluctuations caused by red 17) their valueחts 16 aחts (poiחferring to justtwo data poi

.should bc cstim:Jted by smoothing the random fluctuations.ed as fo110wsrSmoothing was perfo

-f\. power mode1 was fitted to each data set of the two sesS. Thc modc1 W:JSthcn uscd to calculate the relevantךSiOI

-was oncc calculatcd from thc model deד)parameters. T(ld thenwasחFigure 1), aחrived from the first session (F i

-g the mode1' derived from the second sesחrecalcu1ated usi).1Figureחiגsion (Q

gחtation of the forgettiחeral represeחThe following is a gefu.nctio:ח

(3)Q, = j(F. D)

where:

ance time of repetition 17 (firstrepetitionrperfo-גQ

after the break), calculated through fitting a power.32ance times of repetitions 17 tormodel to perfo

ance time of repetition 17 usingrF - caIcu1ated perfo.e1 derived from the first sessionסthe power m

D - break time between repetition 16 and 17.

estimateסels can beused tסDifferent mathematical m-f F and D. A stepwise regresסctionחas a fuגf Qסthe va1ue

r evaluating the predictabiliry powerסsion analysis was used f.2els. The results are sumrnarized in Tableסof possib1e m

able 2. Rtting Possible Forgetting ModeLsז)C•• C,. C, are parameters(

•C,C•cנAdjuste<l RModels.No

.1.2136.890.551C,F+Co:נa1

1.281.225.100.6508, •• C. + C,F + C,D2

1.291.24-0.652B, - C,F + C,O3

.0.902.360.613\a ••• C.FC4

0.090.901.870.750B, •• C.FC\OC25

0.091.02-0.741",,'FC\ OC2~a6

,a,ס1.001.00104.80.687 _ C.' C. F . C7

From the analysis one can draw the fo110wing conclusions:

ce time has a higher impact than theחarus perfo1ס. Previ-break length (first variable to enter the stepwise regres

).sion

'e, with an R5ח) is the best oס.cl (n2ס. The power m).0.86=0.75 (R

rm relatingסEach particip:Jnt rcceived 16 forms. each fcluded dataחe form iןןן.dividual employeeחto a specific i

.J to bc kcycJ cX:Jctly :JS:t h:S tllןןS of two typcs: itCIןןitCIed firstסIhe forln and dala that had 10 be cחסIhey appcar

g 10 a predefincdcoding t:Jble and then keyed intoחrdiסaccC()IIlpurcdcach ficldןXY.S!CIlןTIlc uata c(>rrcctit)Iןן.Ihc SystCI

.d by the subject with the cxpectcd cntry into that field~terחecrcd. Al1ןחd 10 bc ree:n error in :1 ficld. it h:In case,of

,s possible:ts were motivated 10 perform as wellחparticipace Ihe faster Ihey completed the work Ihe sooner theyחsi

icip:Jnts. prizesrher mOlivale thc parorder to fuחן.could quitwere given to the first three who comp1eted 'the job in each

.grouptחtai-ח was a data base coחperimeגThe outcome of the e

-g 120 files - a fil.e for each participant. Each file con~ח

iח(s aחeach of the 32 repetitioחסed performance timeחtai.the break time between the two sessions

Figure 1 charts dala for a typical participant (see notatioll).itions given belowחdefi

zQt:...wa..wa:

35

~ °1a..

wבt=

5

ON NUMBERןזןזREPE

tחFigure 1. Typical Data Set 01 a Par1icipa

Results and Analysis

e firstד•to two major partsחThe analysis is divided igחo the :leamiןחg iחrgettiסf fסnסuctiסdeals with the intr

d natureחitude aחd part estimates the magחcurve. Theseco)1(el described by equationסof forgening. The power m

cl. Its validation as aסwas used as thc basic learning m0[r this study was established by fitting itסel fסper mסpr.s for e:Ich of thc 120 subjcctsןחpoi:ו:ןוf 16 d(Ihe first $ct-ted through perחg may be represeחThe impact of forgetti

ample.let us assumeגk. For e:formance time after the bre200n would beסth repetiti1דf theסthat performance time

secondsחrepetitioח if no break wcre 10 take place hetwee-but when there is a break of, say, 20 days beדן,t 6 and

280ac{ua1 performance time is1,דdחWeen repelition 16 a{cre35eחiחforgetting thereis aסsecond.s. Therefore. due t

•ance time. Howeverחof 80 seconds (280-200) in perfo

~ךן

.f,·

.:•

,

ג1':.<}."'.i·ךן~:.::

)1::4

,~

~'i

1:1-,"i.'נ -,I.l

j;j';i ~.;

'-,Iז._

'!.;: .,!,~,נ'ה"'.,~;~

Page 164

Table 3. Comparison 01 the FOfecasting Ability 01 aing-Forgetting Modetחd a learaConventionat leaming Modet

MMADMADThe Model

26163.39m-.•T a als

-12156.3-4E 8,(L+S)-m,T

,(1ADח Mean Relative Deviatio(ןvAbsolute Deviation-or, term (M). All three paramerd the mean eחMRD), a(-lel's are derived from the error distribution, that is, the dis

d the forecaSt givenarvation~actual obחl;Ince belween a,el. The optimum value for M is zeroסby the appropriate m~d the obaunbiased oneas that the model isחas it mea

s are equ.alJy distributed between the two sides ofחservatio-ts due to theuse of theleamingםthe curve. The improveme

57%el are 13% in MAD, 1% in MRD, andסforgetting m.Mחi

10The fact that the MRD improvement is so small leadsthe-ם hypothesis that the leaming curve ofthe second set co-verges rapidly to the leaming curve of the fLrStset. There

fore, the impact of forgetting may be ignored when dealingciently large batches, an issue which is discussedfwith sU

.-.in the next paragraph:g may be evaluated on rwo levelstThe impact offorge-d the cumulative one. Evaluating forgetatheinilial Ievel-itial level is an estimate of performance' deחthe iחסting

cline inunediat.ely after the brealc, while the cumulativc:: Ievel-rformance decl.ine in a batch production c::n~deals with a

'..vironmentThe initial level may be measured by the ratio described

byח)6:( equatio

,(6)F-גCoP:'DC

FFF-גQ =--=IF

where IF is the initial impact of forgetting. IF = 0 mea.nsd F are defined as before. IF is defmedaגno forgetting; Q

ion of additional tirne required (0 completepסrזas the prothe first repetition aftc::rthe break, as comparc::d to the timc::

-that would have been taken without a break. The initial forgening was found 10 be a function of the. break length, as

.2shown in Figureponential inגis eםFigure 2 indicates that the phenomeno

-shape. The greatest impact was c1early obtained for the long-est break time group. reaching the value ofIF = 0.65. Max

imum forgetting is obtained .,.·hen performance time of thefirst repetition after the brc::akcquals the very first repetition

ance time of the firstחin the first session. Average perfod average value fora544=k was Qזrepetition of this

ce time of repetition 17, using theחaחF (calcu)ated perfoel of thc: first session) was F = 212. SUbstituting thosc::סm

values in equation (6) we obtain

544-212IF_u = 212 = 1.57 .

379

.. '.

·1

sity and itחforgetling inteחסimpact3ח. 8reak lime has aelסadds around 14 % 10 the explanatory :lbility of the m

).0.137=0.613-0.750(

i"1,II

I

I-ely dcscribes thc for;ן!odcl 5 :lpprorri/'ז'!g thatחAssunii

ooclןןןgחrl1i:cd it il1tU1I1CIc:ןג\lCill!cgr\)ח.gc!ting pl1CI1UI11Clltto improve its predictability power. The Stanford B modelequation (2» \\las selected for this purpose. The parameter(

tsthe impacI of forgcning. The leaming-forgettingחL represe:g stagesחniodel is developed through the followi

e15 in 'Table 2. Thatסa. Estimate the value of G1 using m.is

Co FC' DCl=גQ

(4)1.87Fo.9°Do,o9 Qג.

ed break time is D = 25whileחFor example, if the planF = 200. then substituting those values in equation (4) weobtain

Ql = 1.87*200°·9*25°·09 = 294

17of repetitionחce time deterioratioחIn this case performaequals

.dstF = 294 - 200 = 94 seco-גQ

elסford B mtg the Sחuc::of L. Usiab. Calculate the vis theגue of Qaing of the second set, the vחfor the begin

-value of the firsl repetition in that set. Substitutipg the values in equation (2), we obtain

Co FC' DC2 = Q(l +L)-- .

Thus:

O(ג~--1(5) FC' DC~L = (

:0keeping in mind that L should be within the range of16. L = 0 denotes total forgening, meaning that the~L

ce time equa1mfirst repetition of the second .sct has a perfo-o forחin the first set. L= 16 denotes10ח the first repetitio

.g at allחgettic. Substitute the value of the parameters Q, m. and L into

ting the first repetition of the second .setחequation (2), cou.cחnumber oחas repetitio

el andסming-forgetting m:bet,veen the IcחA con1pariso.el was performedסal leaming curve mחthe traditio

g the abilicy ofחcemiחTable 3 summarizes thc::results co-els to forecast the performance time of the secסthc two m

el is a conventiona1 po-סs. The first mחond sct of repetitio.el. The second one is the leaming-forgettingסwerleaming m

.el developed in this pa.perסmare the MeanחThe three measures used for the compariso

Dccember 1989. IIE Transactions

Page 165

gctting paramcter, which is dc:fined as fo11ows:

(י)TLO

FP = TLA

whcrc

. "-FP - the forgetting param~ter

""-

TLO - time 10st (see Figure 3)

TLA - time of the last repetiti,on in the first set .

-uiv~pressed byגthe time 10stenס<.es.The parameter FP deed. Forחalent number of times..of the !ast repetition perfo

" ,3=ample, if TLO = 630 and TLA = 210, then FPגewhich me.1ns that the time 1051equals three times the last

.performance tirm:-Since expressing the time lost byusing the forgetting pa

rameter is a re!ative rather than an absolute me.asure, it maynship betweenסn purposes. The relatiסbe used for eyaluati

•4FP and the bre.ak length D is presented in Figure

:3

fpFO<Qcttinq,PQrQm~ter

2

O~~~~~~~~~~~~0~--מ~~~ס20°~נ~~~h<חq81ו",k L

Figure 4. The Forgetting PClameter (FP) as a Function 01 BreakLength

Figurc 4 cleMlyiThe naturc: of thc:relatiooship presentednסnstrates thc: decre.asing impact of the break: lengthסdem

.rgetting costסthe fimum value of FP i.s ob<ained after a vcry longaThe m

break, resulting in a 1earning curve of parameters identicalto those obtained for the session before the break:. For this

.0.325=study the average values were a == 544, and m:foUows8דכ(st in thi.scase i.scakulated by equation!סTime

(8))22: as--.

I~

TLO = 2:as--",1-1ון

"L. __ , nnt;n. •.•...- T"'\כ,..nc:rTןח::.

--.t.~,

j ~~:.;ב.~

.זן11i1!~,~li...

~1.

~.,.

~

ff~![f

Jr(~t.'.

IF

InilialFOfgetting

.60

.50

ס".

.30

.2ס

סו.

oסן

Rgure 2. Inilial Forgelting as a Funclion 01 Break Length

The difference between this va1ue and the maximum ot>tained-in this study (IF = 0.65) shows that there is still a signif

icant potentia1 for forgetting and that the present impact ofimum value. Recovery fromag is onJy 41 % of its mחrgettiסf

.he slower it is theז.of great importanceסforgetting is a1s,reסgth. Therefחt it is (0 reduce the break leחmore importa;ot sufficient just to identify the forgetting intensityחit is

theסone should also re1ate to the cumulative time 10st due t-nal to the shaded area in Figסbreak. This time is_prop<Jrti

n-ureח 3, since without the break: it is expbcted that the perfo.ance will follow the initial learning eurve

;'it.;

ifI:J'~

~~iב.

גz.!0

~ i=t=1' wa.wa:

t a:~ wa.

t w::E~f.

ן355.

)ION NUM8ER (SזןזREPE

Figure 3. Descriplion 01 Time LO$t Due 10 a 8reak belween the..wo Setsז

anceסmדThe time lost is obviously a furiction ofboth perf-time of the last repetition before the break (that is, repeti

tion 16 in our s.ludy) and the break length. To investigate-this issue \et us introduce a new parameter called the for

380

Page 166

CESאREFERE

-rgcning Curvcs. Analysis ••nd CombiסAss<:o. R.• "Lcaming and F[ון.1987•ns." M. 5<:. thesis. Tcl Aviv Univcrsiryסnali

.Wcstport.א Ckנ.rum Booסg CUr\V!. Quחir2) Iklhaoui. A .• 1M ua(.1986

•••.rgctting Cosl3ס] CarlSOfl. J. G. andRowe. A. J .• "How Much Docs F[.1976•40-47.9. pp8•.םא.g. Volחri~~g;חIndu.slrial E

-iסAircraft Proccss Curve - M"דnc••i>a41.ת Garg. A .•• nd Milim[•g. Volת,ri~r IXsign Changcs." tournalo! Indu.slrial Engjסficd f

.'1961•23-28.1. pp12•.םא-L.c,arחIסrporaling Forgclung i"Iתc5,א•.כז) Glob<:rson, S. ar.d Lcvin[

rand Productioחo/Opt:ralioסwווtןגiona1 Jrt:ןחן,"ning Curve.s.gחfonhcomirוl.cgtmttM

aתd-c Lcaming Curve." in H','6,:חד) Hancoclc, W.M., Bayha, F. H(•1982•Wi!cyסןhת.),vcndy (cdS.al Engint:tring. Gtbook o/lndus•gחH.• 1M Prychology o! uamidd Egca.•(7) Hulse. S. H•• Dce.se,I

.1975,nסiliU4~-ז•tcGraw Hillן.Mtmory - SlrucrurtJ and Procls.rt:J, Wutzky. R. L•• H8ג] KI[

.1980•y. San Fra.ncisco~H. Frecdnun ar.d Com.ring~1M ProcזG. R. ar.d Loftus, E. F., Hwnan Mtmofזus,9] Lo[

.1976,o! [n!ormarion, 1. WillcyEconomic Lxiing Effect10ב) /<.futh, ]. E. ar.d Sprerrunn, K., "Lc[

•1983•264-269•2. pp29,.סא•Scitnct, VoIזntזJ.1gtaSizing." MNJJrtrCurvc:: Theory." lmman, 1., "Somc: Improvcl111 S[

•1970,189-205.3, ppס.uaiOfl RtJtarch, VoI. 8. NrJournaJ o! Poז-Fac Pcriods of Lca.mingcEffect o{ Alt''זlוc,12) Suk. D. R.(

,Manufacturing Qu4nt.ity," A.lIE Tr=iOlt1mn Eronoסgctting1978.,338-343•3, ppסן.םא,•VoI.

uCost of AirpIJ1!lC3," Jotg13ח) Wright, T. P., "Fac!OI"S Affecti(.1936•302-328•1'1',10..r, Vol~urical Scitnctro~o! A

d Compre-aing Curvc: Historical Rcvic:wL'דhc:.•e, L. E14ז] Ycl[.<jgnca, VoI. 10, pp. 302-328, I~ci.riofl Scil,"ivc Survcy~5ח

ry mcmbcr of the Gt'"3od= Schoolaג.סIProfes.sor ShJomo Globcrson i.s a fg~}ס prta visiaivc:rsiryםAdministntioo at Tc:I-Aviv Uחof B\fSi

consulucioQ::חס watlt =archrivcrsiry. His cuחbrthc:astcm Uאateחc:ntivc:SYiand-נ c:ntcnpcrfoגnסtiirganסfסthe arcas!arc i

.sharing oncsicspccia1ly garזu~tc

ns Rcsc:arch from Tcl AvivסtiriS$an Lcvin rccci"ed his Ph.D. in Opcא1 :1tIdttly an indcpcndcnt consulחUrlivcrsity, Israc:I, in 1982. He is currc:

riccpcrations rcsu.rch in c:1סfסhis rC:Scuch intcrcsts includc applicationsuons managcmcnt.ruulitif'$ ar.d in 0PC:

mcntג oC Industrirthe Dcpaib is a scnior lecturcrזvraham Sh~Dr.ring at Tc:I-Aviv Univcrsity, 1500. Hc rcccivcd his Ph.D. dcgrecמEngi

-rom the Univcrsity o( Washington, his MBA dcgrc:c:: from Tcl·Aviv Uni(-echזofnנ~ion - Isracl Itutiזc:chחtmסfrrhis B.Sc. dc:gaזvcrsi

aysisגin thc: an:.as of anrvolvcd iniםnolog;y. He has bec-ning and control, projבua:ion plruction systcms, prcla.s.sirlCation of p

Dr. Shtub is a 5<:nioc Mc:mbcr oCם.facilitics dcsigagcmc:ntוcct•MSחmc:mbcr ofנa•IIE

381

\.

'"!n (8) represents the time rcquiredסhe first term in equatiד

-in cquaהpcrform the first 16 repctitions. The sccond teסt-completethe repcסI.Sthe time requiredtחtion (8) represe

titions during the second session when no break too1c placeg the average valuesחs_ $ubstitutiחsessioסbetween the tW

-thc folחiבwc ()bt(זן),חסti:nd m == 0.325 in cqu:544=a:g valuesחwiסן

1)16

.1725=s·5גנ - E544*s-O·גנLO == }:544*s-Oד_,1ון_,

I~.

LAדcalculatc thc forgening parameter FP, thc value ofסד.cu[atedauld bc: cסfor rcpc::tition 16 sh

btain1ס) wc(חSubstituting thc propcr values in equatio:thc following value

LAגנ·5=221. ==544*16-oד

-d TLA in cquaaLOדg the va[ues ofחAnd fina11y substituti:7) wc find il.S maximum value(חtio

FP = 1725 = 78-- 221 ..

f FPס~g parameter obtained the valuחcc the forgcttiחSie may say thatחaVC:r.lgc:brc:alc of 65.1 days o2.8ח attc:r a=

fthe-ח pott.:סut2.8/7.8·(100=35.9%ס(g that pc::riodחduri.Iace?חg has .a1ready tak.eitial forgcn

nסConclu.si

.productivity is weU documentedחסgeoflcThe impansח i.סterruptiחt be said about iסחHowevcr, thc sarne can

d the re1ationship between the,lengthחg process aחth.e learniance immdiiatelyהs, the ]evel of perfoחof such interruptio

'a 1aoo-חסd productivity, Basedחaחbefore the interruptiory expcriment this relationship has becn cstab1ishcd forסrat

-rgetting model has b«n deסa specific taslc and a learning-f:_pc::dסveIidepictסne tסA power model was found to be thc best

g modclithe forgetting phenomenon. Thc fact that the forgenature to a learning mode1 may indica.te thatחis similar in

.or images of each:omerrthe two processes may in fact be mi-wn that the potential impact of forgetסt has also been shז•t after a break of more than 65 d.aysating isstill signific

tia1חum pott:af m35.9%סg that period onlyחsincc duri-hcr s[udies shou1d conזg occurred. Thercforc. fuחforgetti-well as tasks of a dif:ger brcak periodsחloחסtratcחce_tחckחut whether forgening is task dcpeסferent naturc: to find

future rCSC3.fchסtribution of this study tחסThc immediate cep-סiרCpriatc models for cסrc: apprסf mסd use~is thc: propo

.g and forgettingחiery studies dea.ling with lסboratבןrסru.al

ransactionsדDecembcr 1989. IIE

Page 167