{\MXED VS. UNMIXED LISTS {N TRANSFER AND REWOACTION Thesis for ”:0 Degree of M. A. MR: {CAN STATE UNZVERSITY Phyllis G. Peters 1961 LIBRARY Michigan Stan: University AIXJD VS. UNLIXSD LISTS IN IRAKSFSR AED RETROACTILN By Phyllis G. Peters 9» F: :1 [*J D) H U) Submitted to the College of Arts and Sciences of Lichigan State University of Agriculture and Applied Science in partial fulfillment of the requiretents for the degree of MASTER OF ARTS Department of Psychology 1961 -, ...L_‘_(‘,ll 5 0 (-1 4. r _ . -\ : (5 (pr '_ fl ‘ t .\ ‘ I .v 1 .j I \ u 0 Av] lg“; LIL) in; L1 ‘4‘ L ’L "J i. ”L l v“ i it: Cid“, i...) “j Li' \ l. o I ill k.) L 3 _ 1 _IL I! 1 f l J J K; {J L: .1. : (‘1 ('fi {fr-3 '4 II )II‘ F4 (s i 11‘, L0] 1"._1l’-‘“ __, r._ rd k I)? f‘ 4‘": ‘ _ .1" 'l k [w‘J ' v 5...;- v ’ a v _ .-. -U v v v - u \y ) - «I &J t v _ ; '{*. g__>x ‘1’] fig- 5 L’ a [it .3 .1 ‘( t'; '1‘ -‘ J‘ '1 Q —~ kl "- {\J‘ 3 7 1 1Hi ‘ (J. .1 ‘1 J— QI Ci u.‘ j s- ’ ‘J L L x if n7 r»- r- L L) C‘. i ) \l i} L; L f r A ‘ j ‘ 1 V‘ a I ‘1 _. "v v .824 Lt, lav-7L 4:3 or: sillixi‘ t” slum)». -\ ’b 1 r‘ '1 ' ‘\ A “ " . .., ‘ ‘ . ‘t ' n ' 1' . t ‘- - ‘ W "‘1 " ' ‘ -1 ‘ ’ " - +“ l". (.1 ii 'Sb 4‘. L)“ U i. C L.—.\_' 75 ' k~ '."H./U~" U .L L; L.‘ C; C(J :1 -1. K.’ V (J U ‘2; b (J _L ’fi‘s ~. - ,7“ . ., A V ' _' l -. 'i _ ‘ 7- . A. ' , ° :_ _f'\ . ‘ r beOOd s Pcb;QHmH .Jiiiiiitj uiuenbion .iln loft; ".%:,._ o u w-' ..'L ' ' ' ‘~ lbi _ Q11 L)L, L'e “lllveli‘l by bbkt‘ifl, “U bl.“ UC’C ' JLL_i‘ 01; J « 1 A ' f . -v _0 - 'F F‘J I._I .. CT PM ,- t: L I I 0‘ C7 r—d (F O K; ).- , Cd 0‘ ,. l./ '1 !J < l (r O C 1—- I x Q r, L. J D m US )0... : {3.3 I l .. «v« V :x; :r”\ I 1 UL cri t 1'73 , 1. .A 1 7“" -L QM h v a, . A C PI. ”.1 i .-_ 13 O _I 1' 0 5:? < L; L- r—x Le I. L Lb U. 3 .1. , (9.7 d 32 .L a! .. .1 ...IIU 0 Lb 1L .1 A“. .7” a|\r. J C v 1 n .i r u C. i L r . I. . L. L A. . .3 i“ H. 0 Ar. nu 1L 3 NJ. .2. .1 Q A. , s c L. W. ‘ f‘,“ p p L.-\ C; 7 Omani/96 / 1V ‘Dz‘rmnrmd ((7.. 2J7/RCULCZ I“ 11 ‘V‘L‘l‘rff: LJ ~ fl If. a i c; H. D: A 1 | l‘ L ~\ ”’1! _'h -‘s D“ L 0 DO 'V'V‘TTI‘ \--\-..,.,‘,__‘ A‘* .L‘» 1m: 1 . ~.T,,,l -"-""’ V J...¢.v- O _. _ --__\ .‘ ‘I‘ \.I‘ -~——J...-. ' , . . t“ F"“"P"n~ 'L'-‘\"I=L'>. L~.\—v~ av . . l1 C‘ l‘.‘ k ‘ _. A u u: e: I‘\ 7'1 ‘ ’: '- .» u, .L.‘J , L21 Li: .J '1— . ,. ‘ J- . ‘ ;xr:LJ;‘g)r1Q:L,].‘ F‘ o . . o . . o . . o . . V9 Eff6u0& [TM T: *‘r ~—/ “Lg ‘J '~ :v‘Nl-L V-‘ C . O . .~-—--~.— . >:— *7 )4 L -.',.'..‘.‘ All. 0 o o o o ' ‘15.“. “H‘ ""‘IV‘fi “vf‘ 0' 1‘ ‘.I< ',, ‘ "IND 7‘ -‘~ -I. ' P 1. , a: “I. U i .L o o o o o -1 ‘ V" I" '(.‘ L. -' ,\ . l O #- Lw U )7 L‘ 'J I} ,1. ‘V"<[< In.) . . .“"‘fi1>—-~~,fi K . [ll'v ‘H, . v-1 1.. 5“. Fg~> w- ‘1} . ’4. 'b— L) L ‘\ \‘J m r- 4:- VJ +4 \ B: 1 v .s i") \J .J J \i) - " ‘ - n , o-v... I """ -"' ‘ -‘ H'C‘ J4.~.h _ _'-.- ‘ [I .\ f.“ . . "7'33 (:1 3 , l - v“- .-x j 3 - _I1_ 1 \ V. V.” _. _ . -- 0 LI .1 \I vr‘ L .'-; ()J ':(.L J__ ‘1 k’t‘le e 2‘ (Jr-5 ‘y-l '11“‘[1 t -! rt 1 . ‘— d J -- 2: 0. p 4 -, .3-7: ~- . 4‘ 1" ~ _. .- , - 1V, . “’9 --— VAL) 11‘... .Mrlili' "7-) '3 " ‘HL: I~.>"'. r-' >- O i . - \J 1- LA .1- d V l V‘ L, '1" ‘ ) ‘ ' I k) ‘V LI .‘0 f: ‘v 1 .1 " l - ' - ~ '— “i ‘ ’ 7" L") ‘7". . —'- ‘ I ‘ ‘ --- “ “*4 i”tv*l01vbb¢ 4«uurrals. . . -‘w wmw l. ~1 .‘ 0 fig- I .LQ th/Lg LAI 1140(4le l chLIJLlJiftkc ,_ NLLQVFCI w , u i U _, ' L.\ . _; J.._;_L1;:) - .uezn Hummer of trljlb to IL critericq for lGCs ari u “IXGU lists for each block of IObr augectlves 1p; blc“ s .itnin sizi- F v-' ‘ Lean rumcer O tj 0.) .4 . 5": F-“ P ' N [.J- } J (‘0 Fl. C?- O (\U c W I 1‘ :xe lets f r elcl k 0 sfigcctives ani blocas “Ithlh si»ilar y relatiwus C:-bifled . . . . . . . . alysis of variance . l terion on Il task for mixed lic m ‘ . x, J. {I saris" iirst six I. trials for mixed lis Anal s”: of Variance 01 trials to cri- ' ” for aug'xed lists. . . l wordsanfi- the lirjt six IL tr 1. US. Mean number of words correctly recalled in the one recall triel_®r mixed and unmixed ills—E15 O O O O O O O O O 0 O 0 10. Analysis of variance of number of words ll. recsllei for mixed lists. . . . . . . V A factorial analysis of variance of words &A recallei for snmixefi lists . . . . . vii ll l8 20 {K} v4 [\3 \J‘ M) v7 [\J U1 ._.| -“4... ,‘VT__.. ‘N z ‘i' u' ' I! “.o \ 3.. _' .L J \. Alma '1 7" ' ’1 _ I E. 4:11 :Au 1. Recall scores after inter;olated 153),:11 Ilillc') o o o o W 2. Mean uuqbe: of Worl; cooreccly anticjgatei in Ltt six IL trials fwr nixed uni upmlxe lists. . . . . El 5. humher of U033: cor ~ctly reiallei for mixed ani Hui; ; lists . . . . 2“ INTRODUCTICN taller and Pilzecker (1900) began the work on retroactive interference when they discovered that learning new material during the interval between original learning and the recall trial nade the amount of material recalled less than the amount retained when simple rest filled the interval. The topic became a popular one for study in the field of learning. One of the factors found to be particularly significant in affecting the amount of interference or facilitation is the amount of simi- larity between the interpolated and original learning. Robinson (1927) was the first to clearly specify the effects of similarity in retroaction studies. He proposed the following relationship: "as similarity between interpolation and original memorization is reduced from near identity, retention falls away to a minimum and then rises again, but with decreasing similarity it never reaches the level obtained with maximum similarity. " (pp. 298—299). This relation— ship gives the well-known Skaggs-Robinson curve of retroactive inhibition. McGeoch and McDonald (1931) and McGeoch and LicGeoch (1937) failed to support the Skaggs-Robinson hypothesis and stated that the greater the similarity 1 in meaning, the greater the interference. Both the mcGeoch studies used Judged similarity as the definition of the meaningful relationship, whereas Robinson used identical elements. The method of identical elements defines the similarity dimen- sion by the proportion of elements common to both initial and transfer tasks. (For example, cef and def share two elements for the first degree of similarity; cef and deh share one element for the second degree of similarity; cef and doh share no elements for the third or most remote degree of similarity.) In the method of judged similarity, the judges must rate on some scale the extent of similarity of various words to the reference word. (For example, when "free" is the reference word and direct synonymity is excluded, "Open" is judged to have the first degree of similarity, "odd" to have the second degree of similarity, and "closed" to be the most remote.) Gibson (1940) in a theoretical article on verbal learning distinguished between the similar- ity of the reSponse in transfer studies. This suggested a reconsideration of Robinson‘s hypo- thesized curve in that he failed to distinguish between the similarity of the various parts of the task. The distinction between stimulus similarity and response similarity found eXperimental valid- ation in eXperiments by Gibson (1941) and Hamil- ton (1943) varying stimulus similarity. Both studied the effect of varying stimulus similarity in a paired-associate task with geometric forms as stimuli and verbal material as reaponses. They obtained Opposite results. Gibson reported that retroactive interference increased with increasing similarity of stimuli. Hamilton reported that retroactive interference decreased with increasing similarity of stimuli. However, Gibson used different and neutral responses while Hamilton used identical responses. Thus, Gibson varied both stimuli and reaponses while Hamilton varied only stimuli which further stressed the necessity for the distinction between similarity of parts of the task. Osgood (1946) varied the similarity of the response, holding the stimulus constant. He reported retroactive interference increased as reSponse similarity decreased. In 1949, Osgood proposed a resolution of the similarity paradox in his transfer and retroactive surface. In this surface, he relates the effect of varying stimulus similarity, response simi- larity and both types of similarity simultaneously to transfer and retroaction effects. He did not feel it was necessary to distinguish between the methods of identical elements and judged similarity in their effects on transfer and retroaction. The significant features of the surface can be stated in three empirical laws, based on then available experimental evidence. Of particular interest for this study is the second law which deals with response variation: ”Where stimuli are functionally identical, and responses are varied, negative transfer and retroactive interference are obtained, the magnitude of both decreasing as similarity between the responses increases. " (Osgood, l949,p.135) (See Figure 1). His 1946 experiment was a major experimental support for this law. Osgood regards the contrary results of the McGeoch studies, previously cited, to be due to the simultaneous variation of both stimulus and reSponse similarity. Bugelski and Cadwallader (1956) made a com- prehensive attempt to test all of Osgood's surface in a single eXperiment. Their results confirmed two of Osgood's empirical laws: (a) Law One, dealing with stimulus similarity, and (b) Law Figure 1. Recall scores after interpolated learn- ing. A comparison of Osgood's theoretical values with empirical data. The theoretical points were plotted on the basis of a control group perform- ance for the zero effect level and a possible peak score of 1}. Recall of original learning material is shown after learning interpolated material consisting of identical stimuli and identical, similar, neutral and opposed responses. (Bugelski and Cadwallader, 1956) 13 H Osgood's 0--¢ Bugelski's No. of Words Recalled Reaponse Continuum Three, dealing with stimulus similarity and response similarity simultaneous variation. Contrary to the second law, Bugelski and Cadwallader found that when stimuli are identical, negative transfer is greatest with the most similar responses and least with opposed reaponses, (See Figure 1). Their results approximated very closely the Skaggs-Robinson hypothesis. Several differences in procedure might be reaponsible for the obtained differences. (1) Osgood used letter pairs (such as c.m.) as his stimuli; Bugelski and Cadwallader used geo- metric figures from Gibson (1941). (2) Osgood used a four-second presentation rate for his stimuli; Bugelski and Cadwallader used a two-second rate. (3) Osgood used mixed lists; Bugelski and Cadwall- ader used unmixed lists. Of the three differences, the one involving mixed and unmixed lists seemed the most likely to explain the results. The difference in stimuli was not considered reSponsible for the difference in results because of the marked agreement between Osgood and Bugelski & Cadwallader on Osgood's first and third laws, both of which involved stimulus 7 similarity variation. (also, since Osgood used a weighted reSponse measure which involved giving extra credit for correctly anticipating within the first two seconds of the stimulus presentation, the importance of the timing difference is less than would appear at first inspection. In the unmixed list design each interpolated list represents only one similarity variation. Thus, one group of fis has an interpolated list in which all the items are similar to the original; another group has an interpolated list in which all the items are neutral in relation to the items of the original; a third group has an interpolated list in which all the items are opposed in meaning or feeling tone. Since the same relation holds throughout the whole interpolated list, §.might discern the relation, thus contributing to positive transfer and confounding the experiment. (Slamecka and Ceraso, 1960) The mixed list procedure attempts to avoid or reduce the possibility of § develOping such a special set for the transfer list by incorporating more than one similarity relation into the inter- polated list.(Osgood, 1946). Thus, one-third of the items in the interpolated list are similar to the original responses, one-third are neutral and one-third Opposed. mixed list designs are also preferred be- cause: (1) Fewer subjects are needed since it is not necessary to have a separate group of subjects for each condition in the experiment. (2) Each subject serves as his own control, thus making for more sensitive statistical tests. Recently, Twedt and Underwood (1959) reported no significant differences in transfer effects as a function of mixed vs. unmixed list design. They compared mixed and unmixed lists in three conditions: (a) identical stimuli - neutral reSponses, (b) identical stimuli - identical but re-paired responses, (0) neutral stimuli - identical res- ponses. However, the point of greatest disagree- ment in the results of Osgood and Bugelski occurred in the variation with identical stimuli - similar responses. This variation was omitted in the Twedt and Underwood study. The condition of identical stimuli - Opposed responses was also omitted. Thus, it was still feasible to hypothesize that the differences in the results of Osgood's work and the Bugelski and Cadwallader experiment might be due to this difference in eXperimental design. The purpose of the present experiment is to study transfer and retroaction as a function of mixed and unmixed list design, using a paired- associate verbal task, which varies reSponse similarity along the dimensions used by Osgood, (1045) while holding stimuli constant. Excerimental Design. This study followed the traditional retroactive interference design (A - B, a - K, A - B) in which the eXperimental variable is the meaningful relation of B and K reSponses in paired-associate learning. Three degrees of meaningful relations were used: similar, (3), neutral, (K), and opposed (0). All groups had the same list for original learning and for the recall trial. For the unxixed list design, three groups had an interpolated list uith only one degree of similarity - all twelve responses being either similar, neutral or opposed to the responses of the original list. For the mixed list design, three groups of subjects learned a twelve item interpolated list with a block of four responses similar, four neutral and four opposed to the responses of the original list. There were a total of six experimental groups. Lists. The lists were taken from Os;ood (1946) and consisted of letter pairs as stimuli, and meaningful adjectives as responses. Table l 10 Table 1. ll Adjectives employed in the experiment in terms of the meaningful relations between original and interpolated materials. 0L IL Similar Eeutral Opposed f.s. tense hard basic soft g.l. free open odd closed h.v. dainty clean curious dirty j.y. robust solid long flimsy k.t. neat clear numb hazy d.m. skillful quick sour slow q.r. slender airy daily solid r.h. noisy excited equal calm t.g. drowsy dead necessary alive w.p. boorish rough near smooth x.n. rounded graceful lucky clumsy y.b. pale sickly similar healthy 12 presents the adjectives used in the experiment in terms of their meaningful relations. Three of Osgood's stimulus-response pairs were removed from each list to reduce the time required to complete the experiment. Table 2 presents the particular combination of adjectives used in the three mixed lists. Three mixed lists were used so that each response adjective used in the three unmixed lists would appear in the same meaningful relation in a mixed list. Thus, the three sets of four similar ad- jectives in the unmixed similar list were used each once, the first set in the first mixed list, the second set in the second, etc. This was also true for the neutral and Opposed conditions. The order of presentation of the 8-H pairs Of the mixed lists during interpolated learning was arranged to min- imize runs of reaponses Of the same degree of similarity. apparatus. For each list learned, three different orders of the S-R pairs were shown on a Lafayette memory drum, with a two-second pres- entation rate, and a two-second inter-item time. Inter-trial time was four seconds. The original list was shown in window one (left side when facing 13 Table 2. Particular blocks Of adjectives em- ployed in mixed lists in terms of the meaningful relations between original and interpolated mat- erials. OL IL mixed 1 2 3 sisilsr £22123; eggpseq f.s. - tense hard basic soft g.l. - free Open odd closed j.y. - robust solid long flimsy w.p. - boorish rough near smooth figgtgal O nosed Similar q.r. - slender daily solid airy t.g. - drowsy necessary alive dead x.n. - rounded lucky clumsy graceful y.b. - pale similar healthy sickly Qppgsed Similar Neutral h.v. - dainty dirty solid curious k.t. - neat hazy clear numb d.m. - skillful slow quick sour r.h. - noisy calm excited equal 14 machine) for all groups. The unmixed lists - similar, neutral and Opposed - were in windows two, three and four respectively. The three mixed lists were shown in windows two, three and four respectively. Two memory drums were used; lists were interchanged daily to control for variation due to machines. Procedure. The general procedure was as follows: (1) The § was given instructions (presented in Appendix A). The anticipation method was used for all lists. (2) s3 learned the original list to a criterion of two successive correct repetitions. (3) A two-minute rest was given in which the § left the machine, and was engaged in conversation by l H (4) The interpolated list was learned tO a cri- erion of two successive correct repetitions. (5) A two-minute rest was given as in (3) above. (6) The g was given one recall trial on the orig- inal list. The recall trial was not begun until the s understood that he should anticipate the first time through the list. Three female experimenters ran all §s. The 15 author ran one—half of the subjects kept in the analysis. Subjects. Subjects were volunteers from the introductory psyChology course, who were given research credit for participating in the eXper- iment. A total of 124 subjects were run. Twenty- eight were not used in the analysis, leaving a total of 96 gs. gs were eliminated for the following reason“: one s for not following instructions, fifteen §s for not reaching criterion on the original list in 50 trials, twelve gs for not reaching criterion on the interpolated learning at the end of their experimental session. There were eight females and eight males in each of the six groups. An order of conditions was prepared before the experiment began with each of the six conditions represented once in each block of six fis, and the order counter-balanced from block to block. §s were assigned in order of their appearance at the laboratory. All gs were naive in verbal learning exper- iments. This section is divided into two parts: the first part presents the analyses of the transfer task, the second presents the analyses of the retroactive effects. Bugelski and Cadwallader (1956) reported a high negative correlation (rho:=-.85) between the trials taken to learn the interpolated list, and the number of items of the original list recalled in the one recall trial. Therefore, he presented only the results of the recall trial in his analyses. Osgood (1946) reported the results of both his transfer and retroaction measures. In the present study, r=-.62 for the three unmixed lists, «.48 for the three mixed lists, and -.54 for all lists combined for trials to criterion and items recalled. All correlations were sig- nificantly different from zero, (p (.05), but somewhat less than expected. Therefore, analyses are presented below for path interpolated learning and the recall trial. Original learning was fairly uniform and an analysis of variance revealed no significant differences between groups. (F=1.03, p>.05). 16 17 Transfer Three measures of interpolated learning were used to ascertain the amount and kind of transfer. (1) Number of trials to the criterion of two successive correct repetitions of the twelve items. (2) Number of trials to the criterion of two successive correct repetitions of blocks of four adjectives common to the mixed and unmixed lists. (3) humber of correct reSponses in the first six trials of the interpolated list for blocks of four adjectives. (All gs had at least six trials of interpolated learning.) a. Lixed vs. Unmixed Lists Table 3 gives the mean number of trials to criterion for each block of four adjectives for each of the similarity conditions in the mixed and unmixed lists. Each block of adjectives contains the four adjectives that are found in the mixed lists and that are in the corresponding unmixed list. The difference between the mixed and unmixed list design was tested for each block of four adjectives by means of t tests. Only one of the nine values of 3 reached the .05 level of signif- Table 3. adjectives combined. H (0 near number of trials for mixed and unmixed lists and blocks qy.&1l.q', \rIlL'. 4.1» Similar Relation to IL criterion for each block of four :,~~. .' . .—-W. -,\- Simllafltj Pelatluhs Blocks Block 1 Block 2 bloc” 3 Combined Lixed 13.1 15.7 14.7 14.5 Unnixei 13.2 16.1 12.3 14.0 Heutral Relation Blocks 7 Block 1 Block 2 Block 3 Combined mixed 12.1 13.6 17.4 14.4 Unmixed 17.4 13.9 3.7 13.0 Opposed Relation Blocks——) Elocx 1 Block 2 Block 3 combined mixed 12.2 13.6 17.6 13.1 Unmixed 12.2 13.4 12.6 13.4 19 icance: Block 1 of the neutral condition, (t=2.21, p_<.05). Die difference between mixed and unmixed lists was also tested for each condition with blocks combined. The three 3 tests were not significant. Table 4 gives the mean number of words cor- rectly anticipated in the first six trials of L); interpolated learning for mixed anx unmixed lists for each block of shared adjectives. Figure 2 presents the condition means for mixed and unmixed lists graphically. Line 3 tests were calculated on the nine blocks testing for differences between mixed and unmixed design with this measure. Only one value of t approached the .03 level of significance: Block 1 of the neutral condition had a t of 3.71, (p<.01). Three 3 tests were performed on the means of the conditions with blocks combined. lone were significant. b. mixed Lists A simple analysis of variance was performed comparing trials to criterion on the transfer task for the three mixed lists. Lists were not sig- nificant, (F =1.0€, p_> .05). Table 4. 20 Kean number of words correctly anti- cipated in first six IL trials for mixed and unmixed lists for each block of four adjectives and blocks within similarity relations combined. Similar Relation Blocks Block 1 Block 2 Block 3 Combined mixed 8.9 10.8 9.5 9.7 Enmixed 10.6 9.7 11.8 10.7 Keutral Relation Block 1 Slock 2 Block 3 Blocks Combined nixed 10.8 7.7 7.7 8.7 Unmixed 5.7 6.1 7.8 6.5 Opposed Relation Blocks Block 1 Block 2 Block 3 Combined Mixed 11.4 10.0 8.0 9.8 Unmixed 11.8 9.6 10.1 10.5 21 Figure 2. Mean number of words correctly anti- cipated in first six trials for mixed and unmixed lists. 12; 10 - \ Lie an Correct . p—-Xixed List xn1Unmixed list Rs Rn R0 Similarity Relation This was an over-all check of the equivalence of the three mixed lists. Ho difference was expected as the only difference in the construction of the three lists were the particular words used for the similarity condition. The similar condition was compared with the neutral (S/N) and with the Opposed (5/0), and the neutral condition with the opposed (N/O). Three matched 3 tests were completed on the number of trials to criterion for the three blocks of four adjectives with the combined totals for a condition from all three lists. No value of t approached the .05 level. Osgood also performed these tests in his analysis, and found the same results for the S/N and N/O comparisons. However, he found the opposed condition took significantly more trials to reach criterion than the similar condition. Since the number of men and women were equal in number in each group, a factorial analysis of variance was used to increase the precision of the statistical tests and to compare the results of men and women. A "Type 1" analysis after Lindquist (1953) was used. Analysis of variance of trials to criterion is shown in Table 5. ihile the three conditions of similarity were not significantly different from 23 Table 5. Analysis of variance of trials to cri- terion on 11 task for mixed lists. Source df M.S. r Between-Subjects 47 Sex 1 529.00 4.73 * error (b) 46 111.83 Within—Subjects 96 Similarity (Sim) 2 8.50 .47 Sex x Sim 2 27.91 1.53 error (w) 92 18.17 Total 143 Table 6. Analysis of variance of number correct during first six IL trials for mixed lists. Source df M.S. F Between-Subjects 47 Sex 1 100.55 2.96 error (b) 46 33.97 Within-Subjects 96 Similarity (Sim) 2 17.06 1.23 Sex x Sim 2 76.29 5.71* error (w) 90 13.36 Total 143 * .05 level of significance ** .01 level of significance 24 each other, females took significantly fewer trials to criterion then males. Table 6 gives the aralvsis comparing the number of words correctly anticipated in the first six IL trials. Keither sex differences nor similarity con- ditions were significant. However, their interaction was found significant. Individual comparisons revealed that men and women were significantly different only with the condition of "opposed" similarity. c. Unmixed Lists The mean number of IL trials to criterion for the similar, neutral and opposed lists are 18.2, 19.8 and 18.3 respectively. Analysis of variance of number of trials to criterion for men and women for each of the similarity conditions yielded nonsignificant results for the similarity variation. The women took significantly fewer trials than the men. The interaction was not significant. (See Table 7) However, no significant difference in trials to criterion was found. The substantial correlation (r=.67) between original learning and IL suggested the use of analysis of covariance. Analysis of variance of the number of words correctly anticipated in the first six IL trials for men and women for each of the similarity conditions, 25 Table 7. Analysis of variance of trials to criterion of IL task for unmixed lists. Source df M.S. F Between 5 Similarity (Sim) 2 11.52 .21 Sex 1 744.18 14.72** Sim x Sex 2 64.18 1.27 Within 42 50.54 Total 47 Table 8. A factorial analysis of variance of the number of words correctly anticipated in the first six IL trials for unmixed lists. Source df M.S. F Between 5 Similarity (Sim) 2 793-75 5.46* Sex 1 1564.08 10.75** Sim x Sex 2 66.06 .46 Within 42 145.32 Total 47 * . .05 level of significance ** .01 level of significance 26 shown in Table 6, revealed both similarity and sex differences to be significant. The interaction was not significant. Individual comparisons showed the neutral lists required significantly more trials to reach criterion than the similar or Opposed lists. The similar and opposed lists did not differ signi— ficantly. Women learned the lists faster than the men. Retroactive Effects The retroactive measure used is the number of words correctly recalled in the one recall trial. The reader will remember that all groups have the same recall trial, but will be labelled with the relation name they had during interpolated learning. a. Mixed vs. Unmixed List Design Table 9 gives the mean number of words correctly recalled for mixed and unmixed lists for each simi— larity condition for each block of four adjectives. Figure 3 presents the mean of the totals for each similarity condition graphically. The difference between mixed and unmixed list design was tested for each block of four adjectives by means of Lgtests. Block 1 of the neutral condition again was the only comparison to reach the .05 level. Again, no significant differences were found between 27 Table 9. Lean number of words correctly recalled in the one recall trial for mixed and unmixed lists for each block of four adjectives and blocks within sim- ilarity relations conbined. Blocx 1 Block 2 Block 3 Blocks Combinea :-.LJ\ed 1.8 106 106 1.7 U11'ili3‘Z‘E’i 1.7 10? 1.7 1.7 Block 1 Block 2 Block 3 Blocks Combined Mixed 2.5 1.5 1.8 1.9 Enjr.ixe\1 1.2 1.7 106 105 Oppose: Relation Block 1 Block 2 Block 5 Blocks 0') 4209/1 {lxed 1.8 1.9 1.6 1.8 {TLALle‘C-i 1.9 205 L—o 2.2 . Ku_mber of words c Irectlv recalled for mixed and unljxe d lis 2.25 L I / I / / BOIL‘O p / Lean 1.75 r deerlled 1.50 r 1.25 '- ;.Iixed b—o ¥. Unmixedt—-‘ J.— l l Rs , Rn 30 H w }_J Similarity i atlon 29 mixed and unmixed lists when tested for each condi- tion with blocks combined. b. hixed Lists An analysis of variance of the number of items recalled in the recall list revealed no significant differences between lists. This test was a rough check of the equivalence of the mixed lists. An analysis of variance using the number of words recalled tested the effects of the interpolated simi- larity conditions and found no significant differences. ( as Table 10) (1) To check the possibility that the type of analy- sis used by Osgood might give different results, matched 3 tests were applied, comparing each similarity condi- tion with the other two. No value of t approached the .05 level of significance. This confirmed Osgood's failure to find significant differences on the first recall trial. 0. Unmixed Lists The mean number of words correctly recalled of the similar, neutral and Opposed groups are 5.0, 4.7 and 6.6 reSpectively. An analysis of covariance, adjusting for differ- ences in the original learning, revealed no differences \N C) Table 10. Analysis of variance of number of words recalled for mixed lists. Source df K.S. F Between-Subjects 47 Sex 1 2.01 .67 g 1,\ II" 9 "Q error (a) 46 _.9c . . . ~. \ Similarity (DIN) '1': ". ‘ , PI 1- Jltflln-Quu ' a U Sex x Sim 'ects \(7 ‘l [u Ad I error (w) 92 1.21 .e W " ’97 («>172th Jerj . 1 ~ A 3 fl .. -‘ 't '. "151 ‘ 3’ .f“ “'- 0 Table 11. a lactorlal anai:b.s oi valianne oi he: recalled fer unmixed lists. I, .. '“J Source if .S. F Similarity Sim) 2 16.14 2.53 Sex 1 70.09 11.2l** Sex X Sim 2 10.40 1.66 r. o , u '. ,- [— dltuln 42 0.2) Total 47 * .05 level of significance ** .01 level of significance \)J H . -- 1 9"“~ - - fi 1 l .. ‘. j 1 ‘ 1 g' 1 \ I m '\ -., . . 1 -.' :-- -‘(. . , lxirrniper 61 moles recalled. iLls is cintlary to ”7‘— ‘1 1 i -. 1 *A q .- 5: ['1 j l 1'1 ’\ . w ' r o ' " “’\ '. .1. - - W nugelbnl and Cadwdlladef's results, WMlCM found significant differences between each condition. Analysis of variance (Table 11) of words re— r._J i-J 1'4 Vfi U‘ LJ 5 1‘“) .1—" ' O Si? C.- t_. O 1‘3 (D called found fehales reca words than males, but not significant differences between the similarity conditions. Osgood re orted in his 1:46 study, and algo stated in his second empirical law, that wwen stimuli p. are identical and rescoases varied, negative transfer and retroactive interference decrees as sinilarity between reSponses increases. Bugslski and Cadwalleder however, asserted that the law should be modified to read, "... the magnitude of both increasing and then decreasing as similarity between the responses increases." (1956, p. 356) The results of the present study did not confirn the effect of response similarity variation reported by either Osgood or Eugelski. Our recall neastre showed no significant differences. If one nerely inspects the recall data for trends, once again there is a lack of agreement. The mixed list groups suggest the least retroactive interference at the negtral condition, and no difference between similar and Op— posed Conditions while the unmixed list groups suggest the most retroactive interference at neutral and the least at opposed. Bugelski found the most retroactive interference at the similar point while Osgood found the least interference at this point. Bugelski obtained significant differences in his first recall trial and used these exclusively for :4 H C on analysis. Cs;ood analyzed both his recall measures U 2 \_)J \N \N (D found no sifinificant dif— q Ha i-vi and his IL measures. ferences on his first recall trial and only a weak trend using tne conbined results of tne 3 BL trials. Therefore, it is worth inspecting O' (P O C L. J U) (‘4- *3 $1 D U1 I fer res‘lts for ayreenent with ours. Osgood dif a significant difference at the .OS level, usiné trials to the IL criterion between the similar and oppose items. But, the results for the first and first two IL trials were far more sensitive indicators of trans— fer. Je also found number of correct resgonses early in IL to be a more sensitive measure of transfer than number of trials to the IL criterion. Thus, we are in agreement with Osg od as to where transfer can be best measured. however, he found the similar condition to have the:]xxmat.negative transfery annl;neutral and opposed to show about the sane amount. We found simi- lar and opposed to show about the sane amount and neutral to show the most negative transfer. The failure to replicate either the results of Osgood or Bugelski may be due to grocedural differences. Osgood and Bugelski both used the method of adjusted or equated learning. This procedure atteuots to equate the amount of learning on any given word, by removing words from the list that have met a criterion of anti— cipation. The present study kept lists intact until the criterion was reached. #4 Generally, results obtained from adjusted learn— ing procedures have been similar in kind to those obtained with unadjusted procedures. However, the par— ticular criterion of learning we chose may have accentuated the difference between adjusted and unad— justed procedures. Both Osgood and Bugelski had used a criterion of two successive correct anticipations for each stimulus-reaponse pair. In attempting to replicate their studies as possible we used the same criterion of two successive correct anticipations, but applied this requirenent to the entire list rather than to individual stimulus-response pairs. The mean number of trials to criterion reported by both Osgood and Bugslski is far less than the mean number of trials to criterion in our study. The mean nunber of trials to the IL criterion for Osgood's most difficult condition was 4.8. The mean for our most difficult nixed IL condition was 15.1 trials. 0\ most difficult IL task was .9 trials. The mean for our most difficult unmixed IL task was 19.8 trials. It is possible that such a difference in learning rate f oulation di may be due to Pox ferences in verbal ability. The size of the difference seems to point more direCt to the interaction between type of learrir; orocedure (adjusted or unadjusted) ard the criterion of learuifli used. If this interpretation is correct, our choice Wr- N7 1 Lie axii(;iL._luU (f $.14 fl "“.'!- 7]., f' 1‘. I ‘-I Vi CI: 3- oei loll, luv, L 8156‘ ,5 A_ 'Y‘i f ‘ ‘ J_. w fl - '1 .JTI‘C.V'- lV":3o iiidld lb b.) {*3 lbit 11 LG 'JE‘iJBV'B b "“u fir-3115' er and retroac ti on V.'_~.._z.""-'1.s L7": alloimt Oi UL: and IL Hrsstice is increaser. \icaeocn era lflOn, W"': ‘~. A _°Wfl‘|"1 _' n _ ' 3__- 3 ,‘ w .' C: "6: ' _"-"1, i;;l;. avaiiaoie inioriation is not VuiiiClELbiJ conplete to indicate the manner in which this variable would interact with a similarity dimersion. The fadlnime to obtain Sj'IflJUCXflLU (l‘EGFGJCES when resooi se Slill‘ ”V was varied su33este’ t““ the similarity relation itseli should be ‘hecked. note tlian fift eel vears have ela,‘ scaled his words and his Yale sophomore jud3es grobably ‘vere dr aMJn from a population with more verbal ability Therefore, a 3rouo of 2C (Lichi3t an State Univer— sity undergraduate) judges, similar to our learning of lie resm>o se words for the CL LJ. Cf fi" ~vr‘w '3. r be, was aiVen a o list and a list of all of -s3ood s similar words. They were asked to pair each of the original res;onse words with the word most similar to it from the list of >rocedure was followed with i, .L similar words. The sale another 3rouo of 2d judges for the ofgosed words with instructions to match up in terns of an cryosed rela— tion. This procedure is closer to tie actual learnin3 .1 procedure tnan Osso oo' own technique of juogin3 similar— .l C‘" D O H.) Q; U) i...’ :3 ity, since it allows the Ss to 88: all ‘ r‘» n" ’L- '. ' ' ‘ . I" I‘ ‘ l “ -'L‘1y' ’ ‘l“‘ Q ‘~-\ " '\ *u . \ fl . L‘cld biomblili/ JG.) 8010:; Dowel", wuel Bab L'EyJOOO '5. S IU made SihilfiritJ jud3we1ts with reference to only one key word at a time. Only four words in either the similar or the Opoosed lists were Haired up cg at least ten of our (1) .' r fi ,fi , ° - ,. .i 3,. .;_ .,'J...‘- \., .. A ., . "b ,V,.!_' ., , .1- ' jtd3es in déidtu It with CbUOvO s IBldothb. This lack of agreement was not due to the unreliability of 0 the judgments since t least sixteen jud3es a3reed on go pairin3 three words which were not paired by Osgood. The extent of disagreement between the two methods of jud3in3 similaritv relations (plus :ossible gopula- U tion differences) would seem to ar3ue that studies on transfer and retroaction with verbal material must first make certain that the methods for jud3in3 similarity are defensible and that they are in fact cross—situa- tional. Otherwise, studies reportin3 to vary similar- ity relations may not be varying this dimension, or varying it in a manner quite different from that intended. One possible wag of avoiding the uncertainties of O the judged similarity aoproach w uld be a procedure for building up different levels of stren3th between hairs of words in the laboratory prior to the test for transfer. This could be done by varying the number of joint exoosures of pairs of words or by settin3 levels D Oi anticipation or recall of a word in the hair wlen the other was presented. ~Q \‘q .1_‘- J_ 'ul; U difference between (D .- Lt (1 UL; It was hygothesised the Osgood and Bugelski studies might be a function of the mixed and unmixed lists design. Instead, the re— sults of this study supyort the finding of Txedt and i“) Underwood (1359 that there is no di ference between mixed lists design. The single significant difference that occurred in one block of four words of the neutral relation seems due to an interaction between list desij: and the nature of the particular four words involved. This interaction could not be tested directly due to ct the lack of an aggrobria e stati tical test. (P 1 77f] S g} It is still gossible that the use of the two de U) might affect transfer and retroaction if different meaninjful relations are used, suci as hijhly similar resbonses (synonyms) and antagonistic or ongosite resgon— ses. An interesting finding in the present study ii the sex differences occurring on each of the variables. 'Hoten did consistently significantly better than men in all analyses of unmixed lists and in trials to cri- terion of mixed lists. brevious This variable has been nejlected in . Jr, ,. _° 7 _ ._ r i ‘ o , , -i ,‘I ,- if“ .- ., 7v- -' .. ,7 ' .L_'» , n n L studies on verbal transier dud retroaction nitn adult subjects. The purpose of the yresent study was to stuoy transfer and retroaction as a function of mixed and unmixed list design, using a dairea associate verbal task, varying resgonse similarity alonj the dimer ions (I) used by Osgood (1946) while holding stimuli constant. Osgood (1946), using a mixed list design, reported that varying response similarity resulted in negative trans- fer and retroactive interference which increased as response similarity decreased. Bugelsxi and Cadwallader (1956), using an unmixed list design, and using the sane resgonse similarity dinension, resorted that this resgonse variation resulted in retroaction which decreases as similarity decreases. Recent y, Twedt and Underwood (1959) found no significant differences in anount of transfer between mixed and unnixed list procedures, but failed to fully checL all the resgonse similarity variations of Osgood. Ninety—six §s in six experimental groups learned an orijinal list and a transfer list, and were given one recall trial on the original list. The transfer task for the mixed list groups contained all three responses variations — sinilar, neutral and ongosed. The transfer list for eacn unmixed list group regresen— ted only one resoonse variation. \N \ .0 The yerformadce measures used were nunber of trials to criterion on the IL task, number of correct resyonses during the first six IL trials, and the num— ber of words recalled durin* the recall trial. Statis- tical analyses were gerformsf on the results for mixed and unmixed lists tagether, and on each sewarately. no significant differences were found between the .s— v v r- " W ~"’~~-«"}-.~'.\ . ‘ ,—. v, ‘ ' - r“ ~‘- . - l 1* (1‘ v ' vardrnn: Bijcllmkflltrl Lrougs tuiinndwxriaf bl QLC to (E'I?i_i',er’iC»n Surly; the 11 t3. (I ‘ W C 1'1 '- .3 ’3' :1; ‘ _) Q a ‘4 *1.) I 1 \ D *‘i (D Q 1:“ FJ H (73 (l- C" |_‘ J I... (L Pa (1‘ A x J (-. F I...J C‘L‘ "i H o P-' _" ° ‘ (1 95‘ 5 \ r "I' 1 '\ ‘ ‘. r “ Dial illCé-UILI ‘_-1_L_€l‘:}flce‘ ”’pi‘k‘f lvU".l ‘1 Cl! COML :lIlll; - ' a 1 ‘ ‘ ~ - . r - r 1“ ~ '. kennel r~.unl9es save s'sflli1cwflblg ole ueUALIVe trans— r‘ ' '. . u \ , ‘ . . 1 ' ' " i” ier twat sisilar on 03,0:ed P3:;H'%tb wvicn J10 not A N . V ‘ a 1 ‘ ' . x x ' N ' r\ -‘ r 1 dlf‘er iron eacn Ut"€£. Elie result J' ic_t:u: cs . f - ‘V .. 1 i l- 7. 9- . a ' v. . A re: cute Slmllflr tv dec-easwb, ““9 “PWAtlve 'WHV