134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

download 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

of 7

Transcript of 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    1/7

    THE RELATION OF SPEED OF LEARNING TO AMOUNTRETAINED AND TO REMINISCENCEBY HAROLD J . LEAVITT

    Brown UniversityIn a recent paper (6), Leavitt and Schlosberg reported an experi-ment designed to study the relative retention of verbal and of motorm aterials. Fo rtun ately the design of the experiment was such th a ta further treatm en t of the data shed light on two add itional problem s,(1) the problem of the relation of speed of learning to the amountof material retained, and (2) the relation between absolute andrelative am oun ts of reminiscence shown by individual Ss. I t is tothe first of these two problems that primary consideration will begiven in this paper.The procedure used to obtain the data for the present analysis ispresented in detail elsewhere (6). Briefly, four groups of 12 Ss each

    were given 10 30-sec. trials on (1) a list of 15 nonsense syllables, and(2) a modified K oerth pursu it roto r. The four groups were recalledfor retention tests after 1, 7, 28, and 70 days respectively, a differentgroup returning after each interva l. Scoring was in term s of thenumber of correct anticipations for the nonsense syllables, and interms of total duration of target contact for the pursuit rotor.THE PROBLEM

    Th ere is almost universal agreement among investigators (3, 4,5, 8) that the slow learner retains less of what he has learned thandoes th e fast learner. The results are the same whether practice iscontinued to a given criterion of mastery, or to a given time limit.To the writer's knowledge one of the very few studies that is at allequivocal is one of Luh (7). Like the great majority of investigators,Luh had concluded that the slower learner was probably the poorerretainer, but he noted also that "with the recognition method, andpossibly also recon struction , th e correlation between speed of learningand the amount of retention tends to change from positive to nega-tiv e" (p . 81), with the passage of tim e. Thu s the learning-retentionrelationship might not be so clearcut as the earlier experiments,using short retention periods, would lead one to believe.

    RESULTSIn the present case our measures of retention were neither recog-

    nition nor reconstruction, but our results are similar to Luh's.134

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    2/7

    RELATION OF SPEED OF LEARNING 135From Table I, below, it will be seen that for both materials, thenonsense syllables and the pursuit rotor, the rank order correlationsbetween score on the last learning tria l and score on the first relearningtrial pass from positive to negative as the length of the intertestperiod increases. Although not all of the individual correlationscan be considered significant, the consistency of the trend is apparent.Taking the individual correlations for what they may be worth, ourfindings, like those of most other researchers, are of a high positivelearning-retention relationship after short intervals. But as theduration of the retention period is extended, the correlations decreasethrough zero to a negative value.

    TABLE IRANK ORDER CORRELATIONS BETWEEN SCORE ON LAST LEARNING TRIAL AND SCORE

    ON FIRST RELEARNING TRIAL

    NonsensePursuit . .i day

    .84.o5*54-i47 days.67.io .o6.29

    28 days .26.23-.68.io

    70 days-.I7.2S-73-8

    * Standard errors.The prediction to be made from Table I is simply that the fastlearner will retain more material after short intervals than will theslow learner. As the period between learning and relearning in-creases, the superiority, in terms of amount of material retained,shifts from the fast to the slow S.To approach the problem from a slightly different angle, considerthe rank order correlations in Table II. Here the correlations arebetween amount learned and relative amount retained. The scoreon the first relearning trial has been converted into percent of thescore of the same S on his last learning trial.

    TABLE IIRANK ORDER CORRELATIONS BETWEEN SCORE ON LAST LEARNING TRIAL AND RELATIVE

    SCORE ON FIRST RELEARNING TRIAL

    NonsensePursuit1 day

    .o8.2824.237 days.24.23

    -.oi.3O

    28 days- 4 S . I 7 . 46 . i6

    70 days .4O.l8- 4 I . l 8

    Relatively, too, the fast learner, after short intervals, tends toretain more than the slow learner. But again, at longer intervals,the trend is reversed and the fast learner retains a smaller proportionof what he had originally learned.

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    3/7

    136 HAROLD J. LEAVITTA third method of considering the results lies in savings scores.In Table III the correlations are between scores on the last learning

    trial and the percent of trials saved in relearning.TABLE IIIRANK ORDER CORRELATIONS BETWEEN SCORE ON LAST LEARNING TRIAL AND

    PERCENT SAVED IN RELEARNING

    NonsensePursuiti day

    .O4.2937-l97 days

    .IO.27- . 2 5 . 2 328 days

    - . 8 4 . o s-42.i870 days

    ~-.87i.04-.42.i8

    In this case the finding is th a t th e fast learner, after short inte rvals,can relearn to his original criterion more rapidly than can the slowlearner. Once again, however, the lengthening of the period betweenlearning and relearning is accompanied by a shift in superiority fromthe fast to the slow learner. After intervals of seven days or morethe fast learner requires more trials to relearn than the slow learner.At this point, before entering into a discussion of results, weshould like briefly to consider the second of the two above mentionedproblems, the problem of absolute versus relative reminiscence.In a recent paper (1), Buxton predicted: "With a low degree ofmastery few items have risen near enough to the threshold, at thetime of the rest interval, to be available for reminiscence; with ahigh degree of mastery few items are not already above threshold.Both stages of mastery should be inferior in absolute amount ofreminiscence . . . to moderate mastery. The relative amount ofreminiscence . . . might, on the other hand, grow progressivelysmaller as the initial level of learning rises" (p. 329).Buxton later tested this hypothesis (2), and his results substanti-ated it. As original learning increased, he found that the absoluteamount of reminiscence tended at first to increase and then todecrease. The amount of reminiscence relative to the amount ofmaterial originally learned consistently tended to decrease.Although Buxton's technique differed in several respects fromour own, the results of the present experiment may serve as a partialcheck and possible source of elaboration of Buxton's findings. Bux-ton held amount of material to be learned roughly constant, whiletime spent in learning was a variable. In the present case timespent in learning was held constant. Furthe rm ore, no S everreached 'high' levels of mastery (more than 50 percent of the totalpossible material) during the original learning period.Thus far no mention has been made of the occurrence of remi-niscence in our results. Actually, on the pursuit rotor, 11 of the 12 Ss

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    4/7

    RELATION OF SPEED OF LEARNING 137in the one-day group showed reminiscence, i.e., gains in their firstrelearning trial over the score on their last learning trial . For th eseven-day g roup, 7 of the 12 Ss reminisced. We shall limit ouranalysis to these two groups because these were the two groups inwhich more than 50 percent of the Ss reminisced.For the one- and seven-day groups retention and reminiscencemay be considered practically synonymous. Th e two are almostsynonymous because our measure of retention included the gainwhich we have chosen to call reminiscence. For exam ple, in Tab le Iabove the correlations are between score on the last learning trialand score on the first relearning tria l. Score on the first relearningtria l is one of our measures of retention . B ut since this score on thefirst relearning trial was often a higher score than that obtained bythe same S on his last learning trial, the score on the first relearningtrial also serves as a measure of gain or reminiscence. If we nowconsider Table I as a table of correlations between amount learnedand absolute amount reminisced, our results are similar to Buxton's.There is a high positive relation between am oun t learned and am oun treminisced after one day (.54), up to moderate levels of mastery.After seven days there is essentially no correlation ( .06).Similarly, if Table II is considered, what was previously a set ofcorrelations between amount learned and relative amount retainedis now a set of correlations between amount learned and relativeam oun t reminisced. B ut this time the findings do no t agree withB uxton 's. Instead of a negative correlation indicating th a t remi-niscence is decreasing as the learning level rises, we again find apositive correlation, indicating an increase in percent reminiscencewith an increase in level of learning. For the one-day group thecorrelation is .24; for the seven-day group there is again essentiallyno correlation, .01.Inasm uch as 11 ou t of 12 Ss reminisced, and inasm uch as bothabsolute and relative retention show a positive relation to originallearning, it follows that there is a positive relation between originallearning and pure absolute reminiscence.Finally, if the correlations between absolute and relative remi-niscence are determined , they are found to be .91 =fc .03 for theone-day group, and .83 .06 for the seven-day group .Before leaving the problem of reminiscence it may be of somevalue again to note that, although few Ss reminisced after intervalsof more than seven days, the majority of those who did reminisceafter those intervals had original learning scores below the mediansof their respective groups. Of the 24 Ss at the two longest intervals ,11 showed some reminiscence. Of these 11 Ss, 8 had learning scoresbelow their group medians.

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    5/7

    1 3 8 HAROLD J . LEAVITTD I S C U S S I O N

    No final at tempt to explain the f indings outl ined above wil l beoffered a t thi s t im e. W e shall co nt en t ourselves w ith sugg esting afew possibi l i t ies for explanation, and with point ing out also that nocompletely sat isfactory explanation has ever been offered for thosefindings which do not agree with ours.As possible bases of understanding of the relat ionship betweenspeed of learning and amount of retention the fol lowing may beconsidered:(1) O verlearn ing. I t m ay well ha ve been tru e th a t th e slowerSs, having learned less, learned i t bet ter than did the faster ones.Such an hypothesis would account for the long term superiori ty ofth e slower Ss on th e serial tas k, non sense syllables. I t wo uld n o tserve to explain the results with the pursuit rotor , nor would i t shedm uc h l ight on th e short te rm sup eriori ty, especially th e supe riori ty inrelative reten tion and rem iniscence, for th e faster Ss. T his sh ortterm success of the faster Ss would be difficult to explain in terms ofoverlearning even if i t were admitted that interfering tendencies, orfading of weak traces, did not effectively begin until after severald a y s . How could an overlearning hypothesis account for the fasterlearners ' relatively greater re tent ion and relatively greater reminiscenceafter an interval of one day?(2) The negative correlations may be a function of a floor set onlearning . As G il let te (3) pu ts i t , the slow learner is at an ad va nta gein th a t , hav ing learned less, he has less to forget. On the pu rsu itrotor the slow learner 's advantage is part icularly great ; he needsonly a single lucky period of contact during his relearning trials inorde r to equ al or surpass his original score. T h e si tuat ion m aydemonstra te another example of the s ta t i s t ic ian 's ' regress ion towardthe mean . '(3) Differential forgett ing the ory , as espoused b y M cG eoch (8) ,is a possible source of exp lana tion. A ctu ally McG eoch did no t seemto apply his differential forgett ing hypothesis to this part icularproblem . Ins tead , he s imply w rot e: "T his high posi tive re la t ionbetween individual scores in learning and retention is to be expected,of course, from the fact that learning and retention are continuingprocesses. . . The introduction of a relat ively long interval betweenm easu rem ents should no t gre at ly al ter this relat ion. I t seems th enthat the 's low' learner gains no retentive advantage from his slownessand the ' fast ' learner is at no disadvantage from his fastness" (p. 388).I t would seem, however, that differences in the rate at which correctand incorrect responses are forgotten might very well result indiscrepancies between short interval and long interval learning-

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    6/7

    RELATION OF SPEED OF LEARNING 139rete nt io n relat ionsh ips. I f th e fast learner has pi led up m an yinterfering tendencies as well as correct responses, and if the inter-ferences fade at a different rate from the correct responses, one mightcertainly expect that t ime would be a cr i t ical var iable in the problem.(4) Since the Ss had some knowledge of their relative standingthere may have been mot iva t ional changes taking p lace be tween thelearning and relearning periods. B u t again, w hy th e shif t in supe ri-or i ty wi th the lengthened in terval?(5) Any final answer must, i t would seem, be in terms of someact ive process or combinat ion of processes which serve ei ther (a ) t ocause a rapid loss of what the faster learners have acquired, or (b )to cause a gradual gain over and above what the s lower learnersoriginal ly reproduced, or (c) bo th . A sat isfactory basis for un de r-standing the f indings may l ie in some sort of a decrement hypothesis(6). Such an hypothesis would assume that reproduct ion is not atrue measure of learning, but that the s lower ' learners ' are in real i tyju s t the poorer reproduc ers . Th e decrem ent be tween the am ou nt ofmater ia l ac tua l ly learned and the amount reproduced would begreatest for the s lowest learners and would be gradual ly dissipated.Taking into account the s imultaneous process of forget t ing, longterm gains by the poorer ' learners ' would be expected, while at shortintervals, before forgetting has set in, the faster learner might wellretain more mater ial than the s lower one.Although such an hypothesis may account for the change frompositive to zero correlations, it is difficult to see how it could be madeto fi t the continued decline of the correlations from zero throughnegat ive va lues . M oreov er a decrem ent hyp othes is , in i ts presentform at least , could not very well account for the results with nonsensesyl lables where no reminiscence occurred.No one of the above suggestions completely fi ts all the facts.But again i t may be pointed out that the evidence for one of psy-chology's most widely accepted dogmas, i .e . , that the faster learneris th e be t ter retaine r , is nei the r conclusive nor supp orted on an yentirely sat isfactory theo ret ical grou nd s. T h e general izat ion seemsat the m ost to be only par t ia l ly t ru e . For th e imm edia te presentat least , we are satisfied to reopen a question which may have beeninoppor tunely c losed.

    SU MMA R YFour groups of 12 Ss each were given 10 30-sec. tr ials to learn alist of 15 nonsense syllables, and a similar 10-trial session on apu rsu i t rotor . T h e rete nt io n of each m ater ial was teste d after I , 7 ,28, and 70 days, one of the four groups returning af ter each interval .

  • 7/28/2019 134-140 the Relation of Speed of Learning to Amount Retained and to Reminiscence.

    7/7

    140 HAROLD J. LEAV ITTIt was found that:(1) The rank order correlations between amount learned andamount retained change from positive to negative for both materialsas the retention interval increases. This generalization holds

    whether we consider absolute amount retained, amount retainedrelative to original score, or savings.(2) The correlation between absolute and relative amounts ofreminiscence on the pursuit rotor is high and positive.(3) Fast learners show the greatest absolute amount of remi-niscence and also the greatest amount of reminiscence relative totheir original scores, when the retention interval is short.(4) After intervals of 28 days and longer, only the slower tendedto show any reminiscence on the pursuit rotor.

    (Manuscript received July 10, 1944)BIBLIOGRAPHY

    1. BUXTON, C. E . The status of research in reminiscence. Psychol. Bull., 1943, 40, 313-340.2. BUXTON, C. E . Level of mastery and reminiscence in pursu it learning. / . exp. Psychol.,1943, 32, 176-180.3. GILLETTE, A. L. Learning and retention: A comparison of three experimental procedures.Arch. Psychol., 1936, 28, No. 198.4. GORDON, K. Class results with spaced and unspaced memorizing. / . exp. Psychol., 1925,8,337-343-5. HUNTER, W. S. Experimental studies m learning. In : The foundations of experimentalpsychology. Ed . by C. Murchison. Worcester: Clark University Press, 1929, 564-622.6. LEAVITT, H. J., & SCHLOSBERG, H. Th e retention of verbal and motor skills. J'. exp. Psychol.,1944, 34, 4O4-4I7-7. LUH, C. W . Th e conditions of retention. Psychol. Monogr., 1922, 31, No. 142.8. MCGEOCH, J. A. The psychology of human learning, an introduction. New Y ork: LongmansGreen, 1942, pp. 633.