c m Mllullfli ‘ \ l D «H ”L i w \H I, i ($6?) “H THSF EHE EV} WEE CE HE STEMLUS, PROWDWG i‘v’IEDEATGRS, ANS EAERED-ASSOCMTE LEARNING I ... p r‘. “(13:A37r5‘3 “T 3?: H il b I 1 - ,ff ‘ 4 ‘ U I) 'U ;.\'S U 0 (“J ‘W WW” “f~.it\.‘.“=‘:£“{"‘ r3]? uhtuflu high”; :KJLRVHY pf: éiéZJ'E‘Hx RQLT tsFu:\ : LIPD/‘QY M'\.':J£‘_w fate L) Univeisgty ABSTRACT THE.NATURE OF TFE STIMULUS, PROVIDING NEDIATORS, AND PAIRED-ASSOCIATE LEARNIEG by Martin Bolt Although subjects report using mediators for learn- ing a paired-associate list, there is little convincing evidence that mediators actually facilitate paired- associate.learning. The present study was designed to test the hypothesis that.providing mediators will facil- itate the learning of a paired—associate list. A second hypothesis was that the extent of the facilitation de- pends on the.nature of the stimulus. A 3 x 3 factorial design.was utilized in which the first variable was the nature of the stimulus (trigram, bigram, or single letter) and the second variable was the appropriateness of the mediator (relevant mediator, irrelevant mediator, or no mediator). Varying the nature of the stimulus was expected to influence the mediators which the control.subjects would discover and utilize, and thereby affect the amount of facilitation obtained.by providing subjects with mediators. A separate group was used for each of the nine conditions with each grOUp receiving a different paired-associate Martin Bolt list. The study-test method of presentation was uti— lized; the stimuli, mediators for mediation groups, and responses were presented on the study trials, and only the stimuli appeared on the test trials. Both the nature of the stimulus and the appro- priateness of the mediator significantly influenced performance. The interaction was not significant. Providing subjects with relevant mediators facilitated their learning a:paired-associate list while providing irrelevant mediators had a deleterious effect. Perfor~ mance improved as the number of letters in the stimulus increased. The reason for the failure to obtain the interaction was not clear. Since the nature of the stimu- lus influenced the number of mediators reported by the control subjects but not by the subjects provided rele- vant mediators, it seems apparent, at least with the present materials, that this measure is not necessarily directly related to paired-associate learning. Approved .EADYCSVX‘ WgMQ Date 5614 3. O} “‘1 (oil THE NATURE OF THE STIMULUS, PROVIDING MEDIATOR , AND PAIRED—ASSOCIATE LEARNING BY Martin Bolt A THESIS Submitted to Michigan State University in partial.fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1967 ACKI‘IOWLEDGKENTS I wish to express my sincere appreciation to Dr. Gordon wood, my committee chairman, for the guidance and assistance he offered throughout this project. Thanks are also due Dr. Norman Abeles and Dr. David C. Raskin for their helpful criticism and advice. ii TABLE OF CONTENTS Page IrImODUCTIONOO000.00.00.00000000000000000000000000Cl IK'ViETI‘IODOOO'CO.‘OOCOOOOO’OOOOO0......OOOOCOOOOOOOOOOOOOJ'L" RESULTSoooooooooooooooooooococoon...ooooooooooo.ooo9 DISCIISSIOI‘IOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOlLE' SUI'fl’iAR-Yoooooooooooooooooo-oooooo-oooo00.00.00...0.0017 BEER-ELICESOOOOOOOOOOOQOOOOOOOOO-OOOOOOOOOOOOOOOOOO019 iii LIST OF TABLES Table Page 1. Lists for the three conditions having a relevant, irrelevant, or no mediator and trigrams as Stimulioomo00.000.00.000.o.006 2. Mean number of correct responses on each trial for each.condition....................10 iv ~-.-...r.- nu —_ LIST OF TABLES Table Page 1. Lists for the three conditions having a relevant, irrelevant, or no mediator 811d trigrams as StimlfliOOQJQ0....=..-.......O000-06 2. Mean number of correct responses on each trial for eaCh condition. 0 O O O O O. O— O- O O. O O O O O 0' 0'10 iv .n-auan-n\~4 C I ‘I I Q 0 i Q D O I . LIST OF APPENDICES Appendix Page A. I: “BEER CORRECT OII EACH TRIAL FOR EASE-I SUEJECT..20 INTRODUCTION There is some evidence that gs learning paired- associate (P-A) lists utilize mediators or mnemonic de- vices for at least some of the items. But there is little convincing evidence that mediators or mnemonic devices actually facilitate P-A learning. Bugelski (I962) found that g; reported the use of mediators for 67% of the paired-associates and that many §s reported inability to learn particular pairs until they were able to provide mediators for these pairs. Bunquist and Farley'(l96H) instructed §s tc.produce verbal mediating responses for particular pairs of items and measured the latencies of their reSponse. The assumption was made that the strength of the mediator would be reflected in its la- tency. Therefore, if a.mediator had been used for learning a particular pair, the latency of reporting a specific mediator should be quite short as compared with the latency of report when a mediator was not used in learning. They found that learning the P-A list prior to the “mediation test" decreased the laten- cy of the reported mediators and that the latency of mediators was shorter for those items on which §_ re- ported using a mediator during learning. I The above evidence, although suggesting that medi- ators may facilitate P-A performance, is not sufficient because, as Underwood and Schulz (I960) suggest, gs may simply report using mediators on the items they learn most rapidly. Another approach, which does not rely on the reports of S, may be more appropriate for ascertaining the role of mediators in P-A learning. A study by Dallett (l964) suggests a possible alterna- tive. Dallett attempted to alter Sis associative re- Sponses to stimulus trigrams by presenting a word during the first presentation trial which was an asso- ciate of the trigram and either related or unrelated to the response which S was to learn. For example, the word bacon (relevant mediator) or hag; (irrelevant mediator) was presented in parentheses between the trigram Egg and response gggg, Dallett found that if a relevant association was primed, S learned faster than if an irrelevant association was primed. However, a control group presented only the trigrams and responses did not differ significantly in performance from the groups receiving relevant mediators. The failure to obtain a significant difference might suggest that providing mediators does not facilitate P-A learning. However, Dallett suggested that the relevant mediating associations were available without priming for control group SS, and therefore control gs were also using .._ L.’ mediators. For example, in learning the P-A item hac—egss, control gs may have discovered and utilized the mediator bacon. It seems plausible that providing mediators for one group will not facilitate their learn- utilize the same or compa.able mediators. That is, if control gs are not able to discover and utilize effec- tive mediators, providing gs with effective mediators should facilitate P-A learning. Two hypotheses were investigated in the present study. The first hypothesis was that providing mediators can facilitate the learning of a P-A.list. The second hypothesis was that the na- ture of the stimulus will influence the mediators which J) control 3 discover and utilize, and thereby affect the amount of facilitation obtained by providing §s with mediators. Thus, there should be an interaction between the nature of the stimulus and the mediation treatment in that the nature of the stimulus should not influence the number of mediators available to those §s provided with mediators. " no— “1‘?!" METHOD Design. A.3 x.3 factorial design was utilized. The first variable was the nature of the stimulus (trigram, bigram, or single letter), and the second variable was the.appropriateness of the mediator (rele— vant mediator, irrelevant mediator, or no mediator). The assumption was made that varying the nature of the stimulus would influence the_likelihood of control.§s discovering a "relevant“ mediator. For example, the mediator bgggg_should be more easily discovered if the stimulus-response Eggygggggis presented than if B:§gg§_ is presented. The irrelevant mediation condition was included to control.for the possibility that the medi- ator may simply serve as a better functional stimulus. One could argue.that facilitation is obtained by pre- senting a mediator simply because it is easier to asso- ciate a word with a.word than a word with a trigram, bigram, or single letter. Thus, if facilitation results when a "mediator" not.related to the correct response is provided, then the facilitation obtained by providing a relevant mediator probably results from S using a word as the functional stimulus rather than the trigrams, bigrams, or single letters. That is, if a pre-established L, association between the "mediator" and response is not necessary in order to obtain facilitation, then it is probably more appropriate to say t-a the S has used a different functional stimulus (i.e., the mediator) than to say that the §_has mediated. A separate group of s; was used for each of the ins conditions with each group receiving a different P-A list. The prediction was that the relevant groups would do better than the irrelevant groups and that the performance of the controls, that is, those groups pre- sented no mediators, would fall between these two groups. Furthermore, for the control group presented the trigrams, a relevant mediator should be more easily discovered than for the control groups presented the bigrams or- single letters. Thus, an interaction was expected such that there should be a greater difference in performance between the single letter control and single letter relevant groups than between the trigram control and trigram relevant groups. Materials. Three of the nine lists utilized in the experiment are presented in Table I. Each list had the same stimulus trigrams and responses. The lists differed with respect to whether no mediator was presented, a relevant mediator was presented, or an irrelevant medi- ator was presented. The six additional lists were iden- tical to the three listc in Table 1 except for the Table 1. Lists for the three conditions having a relevant, irrelevant, or no mediator and trigrams as stimuli Relevant Mediator WEV CLOTH UflEVE) TAL POWDER (IALGED Ema SEASON (YEAR): RUJ LIPSTICK (ROUGE) BAC EGGS (BACON) KAF SYRUP (COUCH) DEP SHERIFF (DEPUTY) IEOD (HOOD) SAD ITIC IiASKET (PICNIC) snrv (sures) MONEY ILLUV IEOOD ( LUNCH) TEAS} (IIAEUD) ESOFEI -HAR Irrelevant Mediator WEV CLOTH (WEATHER) TAL POWDER (TALENT) YAR (YARN) SEASON RUJ (RUG) LIPSTICK BAC (BACK) EGGS KAF SYRUP (CALF) DEP SHERIFF (DEPTH) MOD (MODE) SAD PIC BASKET (PICTURE) SAY MONEY (SAVAGE) LUN FOOD (LUNATIC) SOFT (HARM) Control I wEv CLOTH 5‘ TAL POWDER . YAR SEASON i RUJ LIPSTICK BAC EGGS KAF SYRUP DEP SHERIFF MOD SAD PIC BASKET SAv MONEY LUN FOOD HAR SOFT stimuli. That is, the bigram lists were identical to the trigram lists except for the deletion of the last letter from each trigram. The three single letter lists were identical to the three trigram lists except for the deletion of the last two letters from each trigram. The nine lists were typed on memory drum tapes, Similarly to. the way they are presented in Table I, wanna.“ mafiafifa. " . J with each list appearing in three different orders. That is, for the six lists containing mediators, the stimulus was on the left, the mediator in parentheses immediately below the stimulus, and the reSponse imme- diately to the right of the stimulus. For the three lists containing no mediators, the stimulus was to the left and the response was to the right. Procedure. The. study—test method of presentation was employed with. the stimuli, mediators for mediation groups, and reSponses being presented on the study trials. Only the stimuli appeared on the test trials. Throughout the experiment the memory drum was operated at a 2 sec. rate with a. 4 sec. intertrial interval. Criterion was set. at one errorless test trial. or a maximum of six. alternating study and test trials. The nine groups were given the same instructions for‘ P—A learning. However, in addition to the instruc- tions given the controls, the mediation groups were Ufld'flmt there would be a word in parentheses below sum mumulus which should help them associate or con- nectthe stimulus on the left with the response.on the right. After the last trial the drum was stopped and s, wasgflyen a list of the 12.pairs that had just been. Inmmented. At the top of the list were instructions to write, next to each pair, any mediators, mnemonic ‘wumfiuymurfinfiw-Inr I devices, or memory tricks that he had used in learning the pairs. Subjects. enrolled in introductory psychology courses at Michigan They were randomly assigned to one The SS were 135 undergraduate students State University. (1f the nine groups, each group consisting of fifteen subjects. RESULTS The mean number of correct reSponses for the six test trials is presented.in Table 2. Subjects reaching criterion prior to the sixth test trial were given a score of 12 correct on each of the remaining test trials. For any given stimulus.condition the group provided a relevant mediator was consistently superior both to the group not given a mediator and td the group provided an irrelevant mediator. In addition, each group not pre- sented a mediator was consistently superior to the corresponding group provided an irrelevant mediator. Each group presented trigrams as stimuli did better than the corresponding group presented bigrams, and in turn each group presented bigrams was superior to the corre- sponding group presented single.letters. An.analysis of variance:was conducted on the number correct for the first trial, the first two trials, the first three trials, and all six trials. Since all of these analyses yielded essentially the Same results, the analysis of the number correct for the first trial will be reported so that comparisons can be made between the present study and the study by Dallett. This analysis indicated, as expected, that the nature of the stimulus 9 lO hmle2h Mean number of correct responses on each trial for each condition Tunas 1 2 3 H 5 6 M 3.20 7.47 8.87 9.60 10.80 10.86 R SD 2.69 4.86 4.85 4.11 3.11 3.46 Single 14 2.33 4.80 6.73 7.86 8.27 9.00 i c , Letter SD 2.56 4.29 5.58 8.06 5.61. 8.66 E M 1.20 3.20 5.27 6.07 7.00 7.47 IR SD 2.29 2.42 7.49 7.89 10.33 8.86 R M 3.93 7.80 9.13 10.27 10.27 10.80 SD 2.53 4.56 5.11. 2.40 4.80 2.29 M 2.93. 6.53 8.47 9.60. 10.27 10.40: Bigram C SD 3.99 4.76 3.79 3.30 2.27 3.84 IR. M 1.27‘ 4.73 6.33 7.93 8.60 9.73 SD 1.53 3.03 4.26 5.98 5.17 4.53 11 11 8.13 10.80 11-40 12.00 12.00 12.00 SD 4.03 1.36‘ 1.17 0.00 0.00 0.00 . . ' M 6.07‘ 9.93 11.13 11.87 11-87 11.93 Trigram C SD 1.63 1.33 1.25 0.18 0.18 0.15 IR. M 3.47 6.80 9.00 10.47‘ 10.60 11.33 so 5.96 8.43 6.27 3.65 5.71 4.90 11 auiproviding mediators were both significant sources cfi’variance yielding Es (2,126) of 34.13 and 55.02, p<.OOl, respectively. The interaction was not sig- nificant, I: (4,126) = 2.26, p>.05. The failure to Obtain the interaction was not as expected since it was predicted that the trigram control group would have an I easier time discovering mediators than the single letter V' .' o' i‘ A . Is‘i‘ control group, but little difference was expected be- “GEL; ‘ tween the single letter and trigram groups provided mediators. Although Dallett failed to find a significant difference between the control and relevant groups, a significant difference. between the trigram relevant and trigram control groups was obtained in the present study, 3 (1,28) = 10.53. 2<.01. The fact that the present study did demonstrate that. P—A performance can be facilitated by providing subjects with relevant mediators may be attributable to the methodological differences. between. the two studies. Each S in the trigram and bigram relevant groups reported using all twelve mediators provided by E. The mean number of mediators that the S_s in the single letter relevant group reported was. 11.67 with all of the mediators reported being those provided by E. The mean number of mediators reported by the trigram, bigram, and single letter irrelevant groups was 7.1+O‘, 8.93, and 8. 60 respectively, with 863, 92%, and 90, of the me- diators reported by (D } .c.; of these groups respectively, also being the ones provided durir ng the study tria The mean number of mediators reported by the trigram, bigram, and single letter control groups was 8.80, 6.13, and 5.00. An analysis of variance indicated that there was a significant difference in the number of mediators reported by the three control groups, 5; (2,L+2) : L+.33, p(.05. This finding would seem to NW," ‘ unmannwmflxfm ' ) I u} .5.. r, . indicate that the nature of the stimulus does affect the number of mediators reported. {owever, whether all §s utilized the mediators they reported is ques tiona ble. For example, it is quite possible that the groups provided with irrelevant mediators reported using these mediators merely because they had been told to do so. Yet, if one assumes that gs actually utilized the mediators they reported, then the superior perfor- mance, relative to the controls, of the grOUps provided rel- evant mediators and the inferior performance of groups provided irrelevant mediators would have to be accounted for on the basis of the efficiency and not necessarily the number of mediators reported. That is, the control groups reported using the fewest mediators; yet their performance was consistently superior to that of the irrelevant groups. This finding indicates that either some qualitative distinction between mediators should I3 be made or that the number of mediators reported is not necessarily related to performance. Possibly the me- diators discovered by the control groups were more effective than the irrelevant mediators in that the former more readily elicited the correct response. DISCUSSION The results clearly indicate that providing sub-- jects with a relevant mediator facilitates their learn- ing a F—A list, whereas providing an irrelevant mediator has a deleterious effect. The fact that each control group was superior to its correSponding irrelevant group indicates that providing a "mediator" not having'a pre-established connection with the response interferes with learning. It seems necessary that there be a strong connection between the mediator and the reaponse in order for the mediator to facilitate P~A learning. Manipulation of the stimulus also led to a sig- nificant difference in performance. In varying the stimulus it was assumed that the likelihood that con- trol §s would discover a "relevant" mediator was also being manipulated. That is, with an additional letter in the stimulus the likelihood of a control § discovering a relevant mediator should be increased. It was pre- dicted that the increased likelihood of discovering- relevant mediators should be reflected in a significant interaction between the main effects. The fact that control.§s given trigrams as stimuli reported more Inediators than the control 8 given single letters as 11+ 15 sthnfli.does suggest that they were able to discover Imue mediators. Yet, since.havihg a trigram as a stimulus rather than a single letter or bigram did not produce a differential facilitation for control §§.as compared to-gs provided with relevant mediators, the prediction was not supported. Perhaps having a. i trigram for a stimulus enabled control §s to discover i more mediators and to utilize them effectively. Yet, i if §s provided with mediators were able to utilize them more effectively when they were given trigrams rather than bigrams or single letters as stimuli then there should be little reason to expect an interaction between providing mediators and the nature of the stimulus. Although Dallett failed to find a significant differ— ence.between.the control and relevant groups, a signifi- cant difference was obtained between the comparable groups (i.e., trigram relevant and trigram control) of 'the.present study. The difference in the results ob- 'tained.by the two studies may be attributable to the (ltflferent methods or to the presentation rate. Dallett 1ised the anticipation method and a 2:1 rate of presenta- titni while the study-test method and a 2 sec. rate of piwasentation was employed in the present study. Particu- lxxrly'in;regard to the presentation of mediators, the new; of the study-test method may be advantageous. hat I6 is, the results of the present study indicate that providing a mediator will have the greatest facilita- tive effect on P-A learning when a strong association exists not only between the stimulus and the mediator but also between the mediator and the response. The strengthening of the. connection between the mediatOr and the reSponse was likely to occur in the present ‘ “OH“..‘H Aunt-KC“ &« 5‘”, ... . . .1)- ‘f . 1'. .‘t-L'r. I- Ln u..- study in that the stimulus, mediator, and response were presented together for 2 sec. during the study' trial. In the Dallett study the stimulus and mediator were presented for-2 sec. prior to the presentation of the stimulus, mediator, and response for a 1 sec. period. Obviously, the present study does not afford evidence regarding why providing mediators facilitated perfor- Inance in.the present study and priming mediators did rurt produce significant facilitation in the Dallett stautr. It is highly likely, however, that presentation rate or presentation method was responsible for the differences obtained because identical. materials were sted,i11 the two studies. SUMMARY One hypothesis for the present study was that providing §s with mediators would facilitate their uh learning a P-A list. A.second hypothesis was that the extent of the facilitation would depend on.the nature of the stimulus. 3" . .. x2, - will. E’T' J'J‘Z ‘2." 35.81.31". - . . .- - ’ " 3:. . - ., h .- _ e w A 3 x 3 factorial design was utilized with the first variable being the nature of the stimulus (tri- gram, bigram, or single letter) and the second variable the appropriateness of the mediator (relevant.mediator, irrelevant mediator, or no mediator). The assumption was made that varying the nature of the stimulus would influence the likelihood of control §s discovering a "relevant" mediator. The irrelevant mediation condition was included to control for the possibility that a medi- ator may simply serve as a better functional stimulus. The study-test method of presentation was employed, with the stimuli, mediators, and reSponse being pre- sented on.the study trials and only the stimuli appears ing on the test trials. Both the nature of the stimulus and appropriate- ness of the mediator produced reliable effects. The results clearly indicated that providing gs with 17 18 relevant mediators facilitated their learning a P-A list while providing irrelevant mediators had a dele- terious effect. The expected interaction between the nature of tae stimulus and the ap.roariateness of the mediator was not obtained. The reason;for the failure to obtain the interaction was not clear. Since.the 3 nature of the stimulus influenced the number of media-— E tors reported for the control §s but not for he gs 5 provided relevant mediators, it seems apparent that at t least with the present materials this measure is not necessarily directly related to P-A learning. REFERENCES Bugelsni, B. B. Presentation time, total time, and mediation in paired-associate learning. g, exp. Psychol., 1962, 6 , 409-Hl2. Dallett, K. M. Implicit mediators in paired-associate learning. 1. verb. Learn. verb. Behav., 1964, 3, 209-214. Runquist, w. N., & Farley, F. H. The use of mediators in the learning of verbal.paired associates. g; verb. Learn. verb. Behav., 196%, 3, 280-285. Schulz, R. w., & Lovelace, E. A. Mediation in verbal paired—associate learning: the role of temporal factors. Psychon. Sci., 1964, ;, 95—96. Underwood, B. J., & Schulz, R. w. Meaningfulness and Verbal Learning. Chicago: J. B. Lippincott Company, 1960. ., 19 'I'x‘ AWE-i 1:539.” mmuzfif-nqu “if- APPEIIDDC A I‘TUIEER CORRECT OI.T EACH TRIAL FOR EACH SUBJECT A-l TTER RELEVAHT GROUP INGLE.LE G u Trials F'dflivflllhg .ld‘iiu 1:1. . iiivl' 11 12 12 10 ll 8 9 10 Subjects 1 2 3 1+ 12 10 12 12 10 12 10 10 10 12 12 IO 10 11 12 12 10 12 12 10 I3. 11 12 IO 11. 11 m 15 12 SINGLE LETTER CONTROL_GROTP Trials Subjects 1 l 10 10 10 10 00 10 12 11. 10 10 11 13 10 10 13+ 15 11-3 “r d SINGLE LETTER IRRELEVANT GRO‘ Trials '. II . with. llrwv Subjects 11 11 11 1 10 11 IO 10 10 10 11 10 I3 14 15 10 10 10 10 Ape BIGRAM RELEVANT GROUP Trials. 1.1 “Hathaway? . 2 5.2.15 In...“ w - Subjects 10 10 ..Q. .. 4 12 a, 12 10 12 11 12 12 12 12 12 11. 12 11. 11 11. 11 12 10 11 11 11 10 11- 12 13 1L» 15 11 10 11 10 12 lo; BIGRAM COHTROL.GROUP Trials Subjects 11 10 O I 10 12 10 10 11 10 11. 12 11. 12 12 11 12 12 12 12 10 12 12 10 10 10 11 12 13 m 13 12 12 12 IO 12 12 10 A96 BIGR M IRRELEVAHT GROUP Trials Subjects 12 10 IO 10 12 IO 10 10 ll 10'. 12 13‘ 14 15 12 12 12 10 Subjects 1 \oooumme’wm :4 e4 r4 i4 mswmpg 11 10 10 TRIGRAM RELEVAET GROUP 12 11 11 10 10 11 12 12 11 12 10 12 11 A-7 11 12 12 12 10 12 12 12 12 12 12 12 12 12 12 12 I2 12 12 12 12 ‘v .173}... . - ' ‘s ‘ 0' a . 1210313 correct caosp' Trials 1 2 3 4 5 6 Subjects 1 7 12 12 12 12 12 2 6 9 12 12 12 12 3 6 9 10 12 12 12 4 9 1o 11 12 12 12 5 6 12 12 12 12 12 6 6 10 9 10 10 11 7 6 9 12 12 12 :12 8 5 12 12 12 12 9 4 11. 12 12 12 10 5 1o 11 12 12 12 11 7 11 12 12 12 12 12 u 9 1o 12 12 12 13 6 10 9 12. 12 12 14 6 11 12 12 12 12 15 8 10 12 12 12 12 Subjects 1 \0 03 *J O~ U1 -F t» h) e4 14 F4 14 e4 14 \n .F'Lu M H O H C) cm s“ Ed to O\ O) -F (A n: 14 km ox co :4 TRIGRAH I 11 11 10 \O-F'UI\00\\IQ\ 10 RRELEYA. Trials 3 4 3 5 12 12 11. 12 9 ll 9 12 12 10 ll 10 12 8 ll 10 12 10 6 9 10 ll 12 12 6 8 UT GROUP 12 12 10 12 11 12 10 12 12 12 12 NIVERSITY LIBRARIES :1 11’ llll nu III I 3 1293 03 96 0374 MICHIGAN STATE