C 1_-... —‘ fr.v-—1. -_ —o_~_o ‘4441044. — 4‘... . 1 . 1 1 .1 1. . 11. 1. 1.91.. 1.111. r;o».1v....o1.111,.1.41. 111.191.,1. 41 .r... .1... 1 1 . .. 1. . 1. 1. 1. 1 .1. . 1.1.49. o 1 1 a...1.1.1.111...0.‘.~—...11911. 0094.194_.9 1 . 9. o... ..1 1o . 1.1 441.11.11.14, 11.1‘14111 .14 o. 1.. 1 1 11.....149114 C1. 1 1"... . .4 1 . ..1. ......11.."ooo11 ..1, 1.. .4. 4 0.9.1 .1... . 31.1.4143,.0109?!Mara-4.9}Aouoouf43W1’{“‘ . 19.....1 .1 ... .11. .1... . 91.1.. . .111 1 . 1., 111v . .111. ..1 .1 1. 1.1 1.. .1: . . .1 o. 1 . . 1. 1.19.: 11 1 . v1 11.191.51.11. 9......4111411. ..111.11.1111.4.1.....0..4.v111. . ,.1.14... 1111.11.41 _ 1.1.1.1... .... 1. . .1 1. 1 11. ,1: 1 .. .1. 1 0.1.1.1 11.1..1oo1 1w....111.ro1111.1u4.oo o1... 1991:1f '14-:90 Jo. . 1 o.1 1.1.. .11...1..1.1.....l...11 ....vo11 .1411. 1.1.1.. I 1 11. 11 1 .1 _. 1 1. 11 . .1 . 1 . 1 .11.. . ._ .1 V.oo.,o.r.11 141.479.1444... 11.1...\ 1 1. ,1o...o1o...1.1.... 4.... .1 1 .. 1. 1 1.4. .9... 11...... o ..1. 1 1.1. .1 1 1. 1,11 1 1 . 1 1 . 1 ..1 ....111 1. . .111». 11-11. 14“.:v.1i114.rV99/A4..olJ1.’-140t4 1 “I o ... .1... 1.1.. . .4. '1 9.111....0 ..1. ..1. . .1 _ 1. .1 . 1 1 .1 .11. .1141 .1.......14.. 4.1.9.. 0.1..N'1v1 A4... .1 ol¢.4'- 14.441 0 1 v.41... 9....1111011111 1..11._1 1...... .1. . 11 1 1. ..1 .1 .1 1 . . 1 1 .1... .1 1,1..1...1o_11 ..4pv11111....4.1.w.4o.o) J111.la1{~4/ 11.11....4414411119.,.4 1 .. 1 11v. 1.... .11.. 11.171 1 .o . 1.. , 1 .1 .1._ o1 1 .1. .11 . 1.... 1 11 1:..44411-4140114 o! . 1 19.5141. .. 1.9.44.1 . ..1 J . 1.1... 1 .1 1,9 1 4 1.1 1 1 . . ,.._.1. 1, .1. , 4... 11 . . 11“. 1114.....a-1'044 . ..1 s 41.... 1111.0 . .1 .... ..1 1 . . . . . . _ 1 11 1.1 .. . .1 11. 1 .. o1 ...1....111.. .1 11011u1—11floowra. . 1 4.. ..o. 11.11.......11.4 .1 1.....4. 1 o. .....u...~ .44 .19.. a. . 1 1 1. . . _ .1 . 11 .111». . 1 . 1 . . 11.11.1111 ..1 4 1.14.14.11.43“: 1‘1 . 14141940444411o’014' . 1 . 1.. .1 . . 1.. _ .o 1. 1 . . I 1 . . . . 1 _ 1 1. . 0.. 1. .,.. —. 14.4.9.1...1. 1.1.111. , 1 ......11.9.11 ...1 1. 1.1 ' 111.111...11 1 . . .11.. ..1 _ . ...._...1_..1111..11..,1 1113111114....0111.01§.a¢1”. In.» 11.11.11.\ .1 1 .. . 1....r,vo1 H 1 V.1.11.1... .1. .1, . _ .1 . . 1 .1 _ . . . 1.... 1211.104. .14”. W 1 1.... 1. .1..........1. 1 . . . o 11.. 1. ...... o 19 1! .1 _. o . , 1 . 1 1 ....111 .1... .1 . ..pp.4_H44..1o“14.1. no.1/F11141-al94 ”4 . 41'. ..1 1 11.. 1 .1 . .. _ .4 1 . 1 1 1 .1 .o 1 . . 1. 1 1. 1 . . 1 11 .4 1 ’1 _ ,. 1o_l1_1;~_-.1._.O.... 1.4.44 1 1_.v._. 1.. .....11 1 11... 1. _ . . 11 . .1,1.. 1 14» .14...1.1.11-1oo1:oc_4o~.vwo914.41v011-f 1O} . ..1 .414011.4 , . .11 0.111.111,141 1.1 .1 1... a . 1. 1. 1 1 . . .1 1 11 ,_ .1 . 1 o. .1014. 1\1_o0. ’J11lflop...lvo 1. 14.: 1.91.1.19141_411911.411.411.11....114_111.~1...11...11. 1.1.1914...1... 3. . 1 .1 1 .. _, 1 .1 .. 1. ,1 . v . ..1 11- 1.1.4.4....14». 1 oft. , ....1 .9¢.‘.1o‘.. 41. 1 ... r ~v 11 1. . 1. . . . . . 1. 1 1 o» . . . .- 1......14 14.~..4411 , 1‘1 1... 1 .1 . 1. 4.. 1 1 . 1 . . .. _ .111 .. 1-11,...111...’l1_.'.b1;u. 4 1 9.1.1.... .o .1 ..11 1 1. _ . 4 . 11.1. 1.. . 1 . 4 .4... 1 . 4. ......1 .1 . . . 1. .1... _ . . 1 1 1. . 1... 41.... 1.410. 1. .. ...... .111 v... .1. n .1 1 . . 1 1 1 1 .. . 1.. 1 to. V . . 4.1... ..1. .... 1.. .... .10 1 1 1 11 . 1 o. .1. Fr1d. .1 1 o141..99.. .. 1.1:. 1. .4 .. _. . . ._ 1 ..1 F10 1 . .11 1 .14 ....1 . 1 . ... .11 1. .1 114114.. v 11.1 19.1 1 1. ..V .1 o 1 . . 1 1. 1 1. 1 1 ..tol.lo.a.4 >. 1 1. .1 ,v1111 1. 1 .1. 1 . 1 ... 1 1 1. 1. lordoo4 it 1 .1... 1 o .9 1... 1 . 1 .1 O . .1 1 . . ..1 441‘ 1. . 11.. 1‘11. 1.. 1.. . . 1 1 1 . . .1 ..1. an 11' kw. 101 11.4.4. 1 o. .v. . ..1. 1. . . 1. . 1 1 p 1’ 1 :s411. .1 14.. 1 11... 1 1 1 ..1 1 . . I 1 . 1 . 11 . . o . 1 1 . _ «v. 1. 4s . .1 '10141 - 1,11 '1' q .u 1 . . _ _. 1 o 1 11.94/4-‘Hl11flwwnof’1wur~ WhI-1d.fi1”u.._ .11. .1. 1. 1-0.1. ... 1 .1 1 1 (9". 4......019 ' 0 . . .1 1 .1 . . . r. 11 . . o. v 4 o... .119 11.11.11 .1..~.. 4:. .1. s . . 1. . f1 4 ..Iur01o1rlv afaJv... ”an .1 1 1141.1 .. ..1 1 . 1 . 1 1 1 . .1 1 11.nv140P1on11.0'.14¢.J.l In 4J1vl‘u11 1 . ..1 11.11.1.,1 11 11 1 .1 . _ 1.10. «(1.1, . . 11.... .1. 411.1 .. 1 . 1 1 .1 . 1 1.. 1.1.14.1... d1. 1— .1 9 1. 111o1o 1.\. 1. 1.1. 1 ..1 . . _ . 1 1. 1. 1. ..1 4.. ..1 . , . . . 1 ‘o 1.4 _ 1111 .... 1 1.. . 1 1 1.. . . 1 1 1 .. . 1 1 . . 1 #4 \11.1. 1.. 1 1 1 o 1 1.~. .411 ..1 _ 1 .1 1 . . \1 _ 1 . 1 _ 1 1. .1. .1. o1 . 1 1 o 1 1 . 1 1 . . 1 1 .v.. V1111. 1 1. o. 111, . _ _ _ . . _ 1 11.. 1 1 . 1 1. 1. 1 . . . .. ..1 - C L fi 1-4 9 ' Do I ‘4 . 11 1 1 o. . 1 g - 1 . 1 1 1 V . 1 1.. . 1 .11. . 1 . . 1 1 1 1 1 1 . . . 1 1 1.. 1 . _ 1 1 . 1 . 1 .11 . 1 1 4 o 1 _ . 1 n .1 o. 1 w _ 1 11. .\.1 .1 1 1 _ 11.1 1 1. . . 4 _ 1 1 . . 1 v 1. 1. ... 1 1 1. . _. ‘\I — ~ . v 0 1 _ o C ‘. 4 '1 v 1 1 1. 1 . o .1 l r¥l.11 0' 1. W1 1 . 1. .. ..1 1 . 1.1"? 19‘ 1o . .fi 1 11 1 1o . 1 1, .141 . u o o . _ 1 1 1 1 _ 1 . . . 1 1. _ 1 1 1 . 1 1 o . 1 1 1 .1. . . 1 1 1 . . 1 . .1 1 .1 .1 1 1 A . I o . 1 1 1 . w . . 1 I 1 1 . .Y o 1 _ o A . 1 _ u 1 n. I 1.. , 1 t . — 1 . . 1 . , 1 1 1 1 1 .. . 1 . . . 1 o. 1 . . 1 .o . 1 1 a o o \ v ‘ 0 n 0 O I . t 11 q I g 1 1 f . c. O! 1 1 1 _ . 1. 1 4 u 9.. 1. a. E11...— 6 .1. 1 . 1 1 . . 1 1 _ 1 1 .1 . e . 1 . .1 ., 1 . v. . . 1 . 11 1. . . 1. , 1 1 1. _ w .1 i G 6.. 8 v 1 0 1 , 1 . 1 1 . . . . .1 1 , . w , T1E1Ofi1h 1 1 . Ara V1. ,8 1.1.1 . f . . I _ _ . 1 . O , .. Eu . 1‘ 1 14).. 1.. 41.11. 11 1/1IN1144’JOIo1-4 91/1251. ~ . 1 1 71.1.. .1011, u 1144.91o0—141. .1 d 1 . 11 ‘B. 1 . . 1. 1 1. 1” .1 1 (.1. . 1 1 . 1 1 1 .. . r... 1 N . . . . n . 1 1 . 1 0 1 o o . . . , ‘ i 1. n 0 _ 1 1 .1 . 1 1 4 I 04' 1 . _ . o 1 . . — .1 V . . .11....1‘ . .1 M hr . 1- . — i 111 m _ . . .11 1 R . 1 0 4 . 1 m _ o ‘1 1 11 .1 DWI 1 1 * . \I IV 1.11... o 1 111w 11". . 4. .-. 1.!Ior’1uroo4i I. 1 1 I .90; H I {11.09.71v 11/ I [1111...) 1 . 1 1!)...0 1 I o . .. . .. 1 1 1 . I.» p- 1 I. ‘u 1 1". w 1 111! '1 1. I1. 1 I . r f’ 1r 1: u ...- u ~o'1ov P‘ ~ 0": 0'1 IA 9 r 1’01 1.. 1 m ’ol .0 O I 1 . 1—1110 1 1 0 4 on 1 a 0 Ilol 11'. U I I 11 a I . 1- . 1 1o 0 '1 1 4 . 1. ..1 1 . u U 1. 1.11.71101-1 1 1 10-1 1.. ‘1. . . 1 o 0-1 .. . 1 v.11..11.19r.....?.1_.ly11.. ‘1 111......11.211....111..1111.r.1 1.1.1. 1.991.111.1111... 1 . 1.. 1 1 . .... . . 1 .....1111w.!o1.411_ for... 1 11.44.11. .0111 1 1 22.. 1 .... 1. 1 ,1 11.11 12% 11.1.1. 1.11.1.3. . 11(1...1. $3111.11. 11.1.1 1. . 1... .1 1. ..1 2111.11 1—... 1 1 1.11 341., C1. 001Lfll1 .. 1114. 01.11.... T“ .1 41V1N14nl9,.:.oo1ooau121 ..1.ro1.1.Y“o.du1'1.14411111 2.11 .41.. 7.11... 1.11. .1 9.141.114 4 114.31.. .191951149J456411:11 “1.44.4.1“? 1c... 1.11.161»)..~10u1~..11 _ 5117.94.41...me 4.. a. . 9. 1 109. 1 .... .11....1..1.1I1111u\11.... . .1. .9“. 41:14.51. 14111: . «4.1.11. 11-111 4111.. . F1111? .1...1111......1311191u2o 1...... n. .4911- N 1. . 1 ..1. «1.31.3...111111113141500114 7.21.1.1 1:121... 04h“.fo1~ouh “a . 1-. 0 1 1r: I III. .II- LIBRAR; 'l I f IHlllllllllllllHllllUlllNllllllllllllllllllllllllllllllllll . p..-a;3;;::a:;ffan 31293 10390 4953 ”h V ‘w can’t-mi. l ABSTRACT AN EXAMINATION OF THE POWER INVERSION EFFECT IN THREE-PERSON MIXED MOTIVE GAMES by Steven G. Cole One of the phenomenon predicted by the theories which have attempted to describe behavior in a three person mixed—motive situa- tion is the power inversion effect. The present paper examined the power inversion effect as it relates to both the truel, i.e., a three person duel, and the triadic coalition formation situation. Due to the paucity of studies dealing with the truel, there is a lack of data to support the power inversion effect in that situa— tion. On the other hand, the coalition formation situation has been studied extensively over the last decade and the power inversion effect has been reported by virtually every study. The focus of the coalition research has been on the determini- stic coalition formation situation —- those situations in which some unit or combination of units will win with probability one. The probabilistic coalition situation —- those situations in which no unit or combination of units will win with probability one -— has been virtually ignored. This study extended coalition formation theory to explain individual behavior in the probabilistic situation. To test the proposed theory, the present study utilized a truel game paradigm. Four conditions were examined: (1) A deterministic truel which allowed coalitions, (2) a deterministic truel which did not allow coalitions, (3) a probabilistic truel which allowed coali- tions, and (4) a probabilistic truel which did not allow coalitions. The main hypothesis was that the power inversion effect would be replaced by the strength is strength effect in the probabilistic coalition formation situation. It was also predicted that power in- version effect would be observed in the truel situations in which no coalition was allowed to form and in the deterministic coalition formation situation. The results supported the main hypothesis. However, the support was weakened by the fact that the power inver- sion effect was not observed in the deterministic coalition formation situation. The power inversion effect was observed in the propensity to attack the stronger of the other two players in the truel situations and in the coalition formation situations in which coalitions did not form. However, it was not observed in the relative chance for survival. A theory based on the relative disparity of strengths was proposed which predicts that the power inversion effect, with reSpect to distri- bution of attacks, will occur in the truel situation as long as the strengths of the participants are not equal. When the relative chance of survival is considered, the power inversion effect will be a function of the relative strength of the strongest player. A . , - I: . ~-/ )1/ 1‘" . The31s Committee: Approved:_L ,t;?L¢7 g3 J//&;’:f7‘7 ,/T ,/ James L. Philli 8 Chairman 1; 1 A ‘w -- p ’ Datez‘thguta;Qi 175Lg//6 Q) I 7' 7* Lawrence A. Messe v Donald M. Johnson AN EXAMINATION OF THE POWER INVERSION EFFECT IN THREE—PERSON MIXED MOTIVE GAMES by Steven G. Cole A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1968 ACKNOWLEDGEMENTS The author wishes to express his sincere appreciation to Dr. James L. Phillips for his encouragement and criticism during the preparation of this thesis. Appreciation is also expressed to the members of my committee Dr. Lawrence Messe and Dr. Donald Johnson. The author also wishes to thank his wife Betty, for her assistance and love. ii ACKNOWLEDGEMENTS . LIST OF TABLES . . LIST OF FIGURES LIST OF APPENDICES INTRODUCTION . . . PROBLEM . . . . . METHOD . . . . . . RESULTS . . . . . DISCUSSION . . SUMMARY AND CONCLUSION REFERENCES . . . APPENDICES . . TABLE OF CONTENTS iii Page‘ ii iv vi 20 25 29 45 6O 64 67 TABLE LIST OF TABLES Caplow's eight triadic types and the coalitions which are predicted by the four major theories. . . . . Who to whom matrix for offers in the Pc and Dc conditions 0 O O O O O O O O O O O O C O O O O O O O O 0 Frequency with which attacks are directed toward the stronger and weaker players for each condition on eaCh trial. 0 O O O O O O O O O O O O O O O O O O O O 0 Games won by players A, B, and C when not in a coalition O O C O O O I O O O O O O O O O O O O O O 0 Frequency with which the desired share of the payoff (desired payoff), the share of the payoff the other player is expected to desire (expected payoff), and the final division of the payoff (final payoff) approximated the parity split or was an equal split in the Pc and Dc conditions. . . . . Frequency with which A was the preferred player or one of the other players was the preferred player in each condition before and after the game was played . . Summary of the analysis of variance on the preferred playing position before and after the game was played . Means and Standard deviations for the pre and post game preferences of player position . . . . . . . . . . Correlations between the dependent variables and the individual difference measurements in the Dc, D, Pc, and P conditions. . . . . . . . . . . . . . . . . . . . iv PAGE 31 33 36 38 40 42 42 44 FIGURE LIST OF FIGURES Page The predicted probability that the stronger of the other two members of a truel will be attacked based on the disparity of relative strengths . . . . . . . . . . . . . . . . . . 53 The predicted probability that the strongest member of a truel will win based on the relative disparity of strengths . . . . . . . 53 APPENDIX LIST OF APPENDICES Page Experimental Materials. . . . . . . . . . . 67 Factor Analysis of Experimenter Evaluation Data . . . . . . . . . . . . . . 88 A Review of Theories and Research Relevant to Coalition Formation in the Triad . . . . 91 vi Introduction One of the prominent phenomena predicted by the theories which have attempted to describe behavior in a three person mixed—motive situation1 is that the person who is the ”strongest" with respect to initial resources is often functionally the weakest. This phenomenon, referred to as the "strength is weakness" effect, has been predicted by theories which deal with coalition formation (Caplow, 1956; Gamson, 1961a) as well as theories which deal with the truel, i.e., a three person duel (Shubik, 1954). However, the literature has virtually ignored the fact that in many cases the person who is the ”weakest" with respect to initial resources is functionally the strong- est, i.e., the ”weakness is strength” effect. It is sensible to think of these two effects as sub-phenomena of a more general ”power inversion" effect. The present paper examined the power inversion effect as it relates to both the truel and the triadic coalition formation situation. Due to the paucity of studies dealing with the truel, there is a lack of data to support the power inversion effect in that situation. On the other hand, coalition formation has been studied extensively over the last decade and the power inversion effect has been reported by almost every study. Therefore the major portion of the present paper concentrated on coalition research. The mixed-motive situation is defined as a situation in WhiCh (1) there is no outcome which will maximize the payoff to all of the participants, and (2) at least two of the participants can increase their payoff by cooperative activity. For this study a coalition is the joint use of resources to determine the outcome of a decision in a mixed-motive situation involving more than two units." (Gamson, 1964, p. 85) Coalition theory and research, which is included within the boundary condition of this definition, has focused on the deterministic coalition formation situation —- those situations in which some unit or combination of units will win with probability one. The probabilistic coalition situation —- those situations in which no unit or combination of units will win with probability one -- has been virtually ignored. The present paper extends coalition formation theory to explain individual behavior in the probabilistic situation. Since the present study proposed to examine the power inver— sion effect in the truel as well as in the coalition situation, the experimental game paradigm was based on the truel. Although it was kept as similar as possible, the basic truel was manipulated so that the following four conditions were examined: (1) a deterministic truel which allowed coalitions, (2) a deterministic truel which did not allow coalitions, (3) a probabilistic truel which allowed coalitions, and (4) a probabilistic truel which did not allow coalitions. Iruel Theory and Research The truel was first discussed by Shubik (1954) and is based on the concept of a three person duel. It is well known that in a duel, two individuals attempt to eliminate each other and the winner is that person who survives. The truel is a similar situation which differs only in that three individuals participate. In the pure truel no eXplicit coalitions may form. This does not, however, exclude the formation of implicit coalitions. In fact, in the truel sit- uation it is often the case that two of the participants may increase their chance of survival by eliminating the third participant first and then attempting to eliminate each other. Shubik prOposed a theory to account for the outcome of the truel situation which was based on the presupposition that participants in a truel will act rationally. Although the possibility of implicit coalitions was noted by Shubik, he restricted his theory so that it did not allow explicit coalitions. One of the major predictions of his theory was that in some truel situations the power inversion effect would be observed. As an illustration, Shubik offered the following example. Person A, person B, and person C were each allowed to fire one shot at one of the other two. Person A had an 80% chance of hitting the person that he chose to shoot, person B had a 70% chance of hitting the person that he chose to shoot, and person C had a 60% chance of hitting the person that he chose to shoot. The rationalistic point of view assumes that each individual wishes to survive and will therefore shoot at that individual who poses the greatest threat to his survival. Shubik adopted this point of view and computed the probabilities of survival for each individual in each of the six possible firing orders. The following mean chances of survival for each individual were obtained: A=.260, B=.488, and C=.820. It is apparent from these results that the strongest person has the least chance to survive and that the weakest person has the best chance to survive. This illustration of the power inversion effect supports the hypothesis that power inversion is predicted in some truel situations, and that those truel situations in which power inversion is predicted are a function of the relative strength of the individuals involved. Therefore, "in a noncooperative environment it apparently does not pay to be slightly stronger than the others for this invites action against oneself." (Shubik, 1954, p.45) Only two empirical studies (e.g., Willis & Long, 1967; Cole & Phillips, 1967b) have examined the truel situation. Although the Willis and Long study utilized the truel situation, the subjects Each had a 100% chance of hitting whichever player they chose to attack. Since the players were all of equal strength, an examination of the power inversion effect in the Willis and Long study was im- possible. Therefore, the Cole and Phillips study furnished the only empirical data dealing with power inversion in the truel. Cole and Phillips used Caplow's (1956) Type 5 triadic situation, i.e., A>B>C, A<(B+C), which means that player A had greater power than player B, player B had greater power than player C, and player A had less power than players B and C combined. An examination of the number of times each player was attacked and who attacked him on all moves of each game, particularly on the initial move of each game, revealed evidence for the power inversion effect. That is, it was found that (1) player A was attacked a significantly greater number of times by both players B and C than either players B or C were attacked by each other; and (2) player C was attacked signifi- cantly less by players A and B than they were attacked by each other. However, in the truel situation which was examined every player could successfully attack the player of his choice with probability one, while in the situation which was discussed by Shubik the probability of successfully attacking the player of his choice was different for each member of the triad. Furthermore, Cole and Phillips allowed coalitions to form which may have added some aspects to the situation which differentiated it from the pure truel. Coalition Formation: Relevant Theories and an Evaluation2 The study of coalition formation has taken two forms characterized primarily by the nature of the task which the subjects are required to perform. One form, represented by the studies by Mills (1954), Strodtbeck (1954), and Torrance (1955), presents a group discussion task in which a coalition is said to have formed if two subjects come to an agreement. The procedure utilized by Mills (1954) offers an excellent illustration of the type of group discussion task that has been used. Mills asked the members of a triad to pretend that they were on a "military review court sitting on the case of Billy Budd." (p. 658) Mills used two confederates to manipulate the situation and noted with whom the subjects agreed as an index of coalition formation behavior. The other form of coalition research, characterized by Caplow (1956), Vinacke and Arkoff (1957), and Gamson (1961a), deals with situations in which the subjects play a mixed-motive game. The data is examined to determine if a manipu- lation of the initial resources effects the outcome of the game, how the game is played, or both. The first type of coalition studies 2 O O O O 0 For a more exten31ve rev1ew of the coalition theories and research see Appendix C. (coalitions in a group discussion task) has been thoroughly re- viewed by Marie L. Borgatta (1961) and will not be discussed in this paper. The present paper will focus on the second form of coalition Studies, i.e., those dealing with mixed—motive games. Caplow's (1956, 1959) theory of coalition formation provided the impetus for the theories which have subsequently been pr0posed. Table 1 presents Caplow's eight triadic types and the coalitions which are predicted by the four major theories of coalition formation, i.e., Control Theory (CT), Minimum Resource Theory (MRT), Minimum Power Theory (MPT), and Anticompetitive Theory (AT). CT was developed by Caplow (1956, 1959). MRT, which was deve10ped by Gamson (1961a) is a refinement of CT. MPT is based on game theory and was proposed by Gamson (1964). AT was first prOposed by Gamson (1964) and was later revised by Phillips and Nitz (in press). The eight triadic types presented in Table 1 were based on the relative power of the members of the triad. For example, in the Type 5 triad, player A has more power than player B, and player B has more power than player C; but player A has less power than player B and player C combined. Subsequent research has defined power to be equivalent to the amount of resources controlled, and resources have been defined as weights assigned to each member of the triad such that some critical quantity of those weights is necessary to control the distribution of the payoff. Table l Caplow's eight triadic types and the coalitions which are predicted by the four major theories. Coalitions predicted by CT MRT MPT AT* Type of Triad A B 1 A=B=C ...... ........ any any any any any 2 A>B, B=C, A<(B+C)... BC BC any BC AB or AC 3 A(B+C), B=C ........ none --- none none none 5 A>B>C, A<(B+C) ...... AC or BC BC any AB or BC AC 6 A>B>C, A>(B+C)...... none —-- none none none 7 A>B>C, A=(B+C). ..... AB or AC —-— AB or AC AB AC 8 A=(B+C), B=C' ........ AB or AC -—— AB or AC AB or AC AB or AC * = the predicted coalitions when the payoff is equally divisible. B = the predicted coalitions when the payoff is not equally divisible. Control Theory Control theory defines the term control as the ability to domi— nate the other members of the triad. For example, if one member of a triad has some part in determining the outcome of the situation and another member does not, then the first member is said to have control3 over the second. Moreover, if two members of a triad have some part 3This type of control is similar to "fate control” (Thibaut & Kelley, 1959) in that once a coalition has formed, the fate of the isolate is determined. in determining the outcome of the situation, then the stronger member, i.e., the member with the larger resource weight, is said to have control over the weaker member, i.e., the member with the smaller resource weight. The major assumption of CT is that each member of the triad will attempt to gain control over both of the other members if possible. If he cannot control both of the other members, he will attempt to control only one. Therefore, in the triadic types which do not contain an all powerful member, i.e., a dictator, coalitions will form. Which coalition will form is determined by the relative power of the members of the triad. For example, consider the Type 5 triad. Since A>B>C, player A is indifferent with reSpect to B and C as coalition partners because he can gain control of both of the other members by forming a coalition with either player B or player C. However, player B must form a coalition with player C if he is to gain control over both of the other members of the triad. An appraisal of player C's power reveals that he must form a coalition to gain control over one of the other members of the triad. Therefore, player C has no preference between player A and player B. As a result CT predicts that either AC or BC coalitions will form in the Type 5 triad. Minimum Resource Theory MRT is based on the assumption that the ”parity norm" will be salient, i.e., "any participant will expect others to demand from a coalition a share of the payoff proportional to the amount of resources which they contribute to a coalition." (Gamson, 1961a, p. 376) As a result, each member of the triad will attempt to form the winning coalition to which he contributes the greatest share of the resources, 9 i.e., the "cheapest winning coalition." MRT also assumes that reciprocal partner preferences are necessary before a coalition will form. The combination of the parity norm and the requirement of reciprocal partner preferences results in predictions which are more precise than the predictions of CT. For example, the fact that in the Type 5 triad player A wishes to form a coalition with player C does not mean that AC coalitions are predicted in the Type 5 triad. Similarly, the fact that player B wished to form a coalition with player C does not mean that BC coalitions will form. The critical factor is which player does player C prefer as a coalition partner. Since player C would contribute a larger share of the resources to a BC coalition than he would to an AC coalition, and since both coalitions are winning coalitions, it is predicted that player C would prefer player B as a coalition partner. As a result, MRT predicts that BC coalitions will form in the Type 5 triad. A comparison of CT and MRT reveals that there are situations in which only MRT predicts the power inversion effect. For example, consider the Type 5 triad. Although CT predicts that the "weakest” player (C) will be included in all coalitions, it does not predict that the "strongest" player (A) will be excluded from all coalitions. In fact, CT predicts that player A will be included in 50% of the coalitions. Therefore, the "weakness is strength" effect is predicted, but the "strength is weakness" effect is not. Thus by definition CT does not predict the power inversion effect in the Type 5 triad. On the other hand, MRT predicts that both the "strength is weakness" effect and the ”weakness is strength" effect will be observed in the 10 Type 5 triad. Thias is, player A will not be the preferred coalition partner of either of the other two players, and player C will be the preferred coalition partner of both of the other two players. As a result, player A will not be included in any predicted coalitions, i.e., ”strength is weakness", and player C will be included in every coalition, i.e., "weakness is strength.” Therefore, the power inversion effect is predicted by MRT. Minimum Power Theory MPT is based on the assumption that the participants in a coalition situation utilize a rational strategy to maximize their control over the outcome. It is also assumed that the real power of each participant is accurately perceived by all of the participants in a given situation. The rational strategy which the subjects are assumed to utilize is based on the real power of the members of the triad. Real power as defined by MPT is different from power as de- fined by CT. That is, MPT utilizes the Shapely (1953) notion of pivotol power. In triadic coalition situations the pivotol power of an individual is equivalent to the number of winning coalitions which he may enter. If the notion of pivotol power is utilized, only three power relationships can occur in the triad: (1) all three members are of equal power, i.e., any two members can coalace and control the outcome, (2) one member is all powerful, or (3) one member has veto power. MPT predicts that if all members of a triad are of equal power, then all possible coalitions will be equally likely; if one member of a y-l bub fJ (2) FL) 1 .1311 tha as: ll triad is all powerful, no coalition will form; and if one member of a triad has veto power, he will be included in every coalition. An examination of the Type 5 triad reveals that all three members of the triad are of equal power. Therefore, MPT predicts that all coalitions are equally likely in the Type 5 triad. Anticompetitive Theory AT was first suggested by Gamson (1964); however, it was based on the results of a series of studies by Vinacke and his students. (Vinacke, 1959, 1962; Bond & Vinacke, 1961; Uesugi & Vinacke, 1963; Vinacke, Crowell, Dien, & Young, 1966). Vinacke (1959) hypothesized that ''female subjects might be much less concerned with winning and more oriented toward social and ethical considerations —— i.e., try to avoid an aggressive display of power and attempt rather to be 'fair' to everyone.” (p. 344) This hypothesis and subsequent research lead Gamson (1964) to postulate that an ”anticompetitive norm" exists and that as a result "coalitions will form along the lines of least resistance in bargaining." (p. 90) Therefore, Gamson's AT predicts that coalitions will form between players who are equal in resources. An extension of AT (Phillips & Nitz, in press) suggests that the predictions of which coalitions will form should be based on the divisibility of the payoff as well as the initial distribution of resources. That is, if the payoff can be divided equally, the coalition which requires the least bargaining is between the members of the triad who have the smallest resource differences. On the other hand, if the payoff cannot be divided equally, the least bargaining 12 will be required for the coalition which includes those members of the triad who have the greatest resource difference. Therefore, if the payoff can be divided equally, coalitions will form between the members who are the most similar with respect to resources. If the payoff cannot be divided equally, coalitions will form between those members who are the most dissimilar with respect to resources. It is apparent that AT predicts that when the payoff is equally divisible AB and BC coalitions are equally likely in the Type 5 triad. On the other hand, if the payoff is not equally divisible, AC coalitions are predicted. Evaluation of Theories Reference to Table 1 indicates that the Type 5 triad (A>B>C, A<(B+C)) is the only triadic situation which differentiates between the four theories. That is, for the Type 5 triad CT predicts that AC and BC coalitions are equally likely, MRT predicts that BC coalitions will form, MPT predicts that all coalitions are equally likely, and AT predicts that if the payoff is equally divisible, AB and BC coalitions are equally likely and if the payoff is not equally divisible, AC coalitions will form. Most of the research on coalition formation has examined the Type 5 triad. As a result, the following comparison of the theories will concentrate on how well the data obtained from the research on Type 5 triads is pre- dicted by each of the four theories. However, where it is deemed important, the data from other triadic types will be reported. The early research on coalition formation in the triad l3 (Vinacke & Arkoff, 1957; Vinacke, 1959; Chaney & Vinacke, 1960; Bond & Vinacke, 1961) seemed to support CT. At the same time, there was a general refutation of MPT. However, the fact that there was a tendency for more BC coalitions than AC coalitions to form in the Type 5 triad could not be accounted for by CT. MRT, on the other hand, does account for this effect. Furthermore research subsequent to Gamson's (1961a) develOpment of MRT (Vinacke, 1962; Uesugi & Vinacke, 1963; Chertkoff, 1966; Vinacke, Crowell, Dien, & Young, 1966; Cole & Phillips, 1967a, in preparation; Phillips & Nitz, in press; Cole, Nitz, & Phillips, in preparation) has shown support for MRT. That is, most of the research has shown some support for the hypothesis that the parity norm is salient in the coalition situation. Moreover, a power inversion effect has appeared in almost every study. Thus MRT has proven to be the most reliable theory which has been proposed to account for behavior in the coalition situation. AT has been supported to some extent by most of the research which has examined feminine behavior in the coalition formation situation. (Vinacke, 1959, 1962; Bond & Vinacke, 1961; Uesugi & Vinacke, 1963; Cole & Phillips, in preparation; Cole, et. al., in preparation). The support for AT has been manifest in the strategy which seems to be employed by females. There is a consistant tendency for females to employ a strategy which results in a weaker power inversion effect than is obtained in the situations which use all male triads. Research which has attempted to test AT as it relates to masculine behavior has concentrated on the Type 2 and Type 3 triads 14 (Phillips & Nitz, in press). The results of the Phillips and Nitz study supported MRT over AT; however, an indication that an ”anti- competitive norm” may exist was noted. Since AT has received sufficient support to indicate that an anticompetitive norm is Operating in the coalition situation, it is necessary to account for the resulting effect. However, it is not felt that the effect of the anticompetitive norm is strong enough to necessitate a theory based on it. It is proposed that a better method of acknowledging the existence of an anticompetitive norm would be to extend MRT so that it accounts for the possibility that there are individuals who do not wish to strictly maximize their share of the reward. Instead, they desire to divide the reward as "fairly" as possible. Therefore, in those triads which include one or more individuals who play a strategy based on the anticompeti- tive norm, the coalitions which form may not be a function of the parity norm. They may be a function of the anticompetitive norm and in some cases, both norms combined. For example, consider the Type 5 triad in which the strategies of player A and player B are dictated by the parity norm and in which player C's strategy is a function of the anticompetitive norm. MRT predicts that players A and B will both prefer player C as a coalition partner. If the payoff can be divided equally, player C would prefer player B and a BC coalition would be predicted. However, if the payoff can not be divided equally, player C would prefer player A and an AC coalition would be predicted. In many situations this could account for those coalitions 15 which occur and are not predicted by MRT. There are two major criticisms of MRT which are evident to the present writer. One criticism of MRT which can be made of most psychological theories, is that the predictions are too strong. This results from the fact that if MRT is strictly interpreted, only BC coalitions are predicted to form in the Type 5 triad. However, there have been no studies which have reported only BC coalitions. Two attempts to rectify this criticism, i.e., Chertkoff (1967) and Shelly and Phillips (1966) have pr0posed to do so by developing a mathematical model. Both the Shelly and Phillips model and the Chertkoff model are based on the assumption that the probability that a simultaneous reciprocal contact will occur is an important variable to consider when determining which coalitions will form. Both models have had some success but neither the Chertkoff model nor the Shelly and Phillips model was able to satisfactorily account for the coalitions which form. The second major criticism of MRT is that the boundary condi- tions are too narrow. This is indicated by the fact that there are variables that weaken the power inversion effect which have not been taken into account by MRT. In order to extend MRT so that it will cover a wider range of situations, it is necessary to examine those variables. One procedural variable which has weakened the power inversion effect is the use of cumulative score. (Vinacke, 1959, 1962; Kelley & Arrowood, 1960; Cole & Phillips, in preparation). The studies which have examined cumulative score had the subjects play a series 16 of games in which the score for each subject at any point in the series of games was the amount of the payoff that he had accumulated. The winner was the player who accumulated the most points over all games. It is suggested that the amount of the payoff that each subject had accumulated at any point in the series of games must be considered when the relative strengths of the members of the triad are determined. In all games after the first game the strength of each member of the triad is a combination of his assigned resources plus his accumulated rewards. Therefore, in the first game the power inversion effect is manifest by the propensity to form coalitions between the two members of the triad who are weakest with reSpect to assigned resources. However, as the series of the games progresses, the tendency is for coalitions to form between the two players who are behind in accumulated score (Vinacke, 1959, 1962). This indicates a continuance of the power inversion effect with the strength of each member of the triad being determined by his accumulated rewards rather than his assigned resource weight. Therefore, MRT can account for the effects of cumulative score by redefining the concept of resources. Previously a resource has been defined as a weight assigned to each player in a game such that the distribution of the payoff is de- termined by some critical quantity of that weight. It is suggested that a given weight which is assigned to an individual player is a value on some resource dimension. The total resources of the players in.a given situation consist of values on a combination of resource dimensions, i.e., multiple resource dimensions. Therefore, one of 17 the key variables which MRT must consider is the concept of multiple resource dimensions. One manipulation which has removed the power inversion effect from the coalition situation is found in a study by Chertkoff (1966). Chertkoff used the "political convention paradigm"4 for the study of coalitions in the triad. However, he manipulated the probability of future success by varying the probability that a coalition would win the election after it had won the nomination. In addition he made the payoff contingent upon winning the election. The Type 5 triad was examined in all conditions, i.e., A>B>C, A<(B+C). In all three experimental conditions, if player B or C won the nomination, he had a .5 probability of success in the national election. If player A won the nomination, his probability of winning the national election was .5, .7, or .9, depending upon the condition. A control condition was used in which the probability of future success was not a feature. In the control condition the power inversion effect was reported. However, in the conditions in which player A had a .5 or a .7 probability of future success the power inversion effect was weakened, 4In the political convention paradigm each subject plays the part of a delegation chairman at a political convention. Each subject is assigned a given number of votes at the beginning of the convention. For each subject the object is to acquire as many "jobs" for his faction as he can. To acquire "jobs" the subjects must form a coalition which can win the nomination. The number of "jobs" that each subject acquires if he has entered into a winning coalition depends on the division of the "jobs" that the two subjects who formed the coalition agree upon. If a subject is not in a winning coalition, he receives no "jobs" for his faction. (Gamson, 1961b) l8 i.e., the distribution of preferred coalition partners was not different from what would be expected by chance. In the condition in which player A had a .9 probability of future success the power inversion effect was nonexistant. That is, a "strength is strength” effect was reported. Moreover, as the probability of future success increased the parity norm became more salient. This was indicated by the tendency for the player with the highest probability of future success to demand and receive a larger share of the payoff. An attempt to explain why the results vary when the probability of future success is manipulated leads to a differentiation of probabilistic and deterministic situations. Probabilistic situations are those in which no unit or combination of units will win with probability one. Deterministic situations are those situations in which some unit or combination of units will win with probability one. Prior to Chertkoff's study the theories and research had all focused on the deterministic situation. However, Chertkoff's study extended the research on coalition formation to the probabilistic situation. Subsequent to Chertkoff's study only one study (Vinacke, Lichtman, & Cherulnik, 1967) has examined the differential effects resulting from the probabilistic and deterministic nature of the coalition for— mation situation. The major finding of the Vinacke, et. al., study ‘was that there is a propensity to "gamble" in the probabilistic situa- tion, i.e., there is a tendency to play without forming a coalition. Dioreover, the propensity for a subject to play the game alone increases as the probability that he will win alone increases. As a result of the research which has examined the probabilistic l9 situation, it is apparent that behavior in the probabilistic situation differs from behavior in the deterministic situation. Therefore, any theory which offers a comprehensive explanation of behavior in the coalition formation situation must consider the probabilistic situation as well as the deterministic situation. Problem Research on coalition formation has consistently reported the power inversion effect in those situations for which it is predicted. However, as has been noted, Chertkoff (1966) replaced the power in- version effect with the "strength is strength" effect by making the payoff contingent on chance. That is, rather than designing the situation so that any coalition would win, Chertkoff designed the situation so that each coalition had a different probability of winning. A consideration of the possible effects of the probabilistic and deterministic aspects of the coalition situation indicated that two types of coalition situations must be accounted for. The first type was the deterministic situation and was defined as those situations in which some unit or combination of units will win with probability one. The second type was the probabilistic situation and was defined as those situations in which no unit or combination of units will win with probability one. The present study extended the theories on coalition formation from the deterministic situation which has been the focus of previous coalition theories to the probabilistic situation which has been virtually ignored. Extension of MRT As a base for a theory that is applicable to the probabilistic situation the assumptions and definitions of MRT were adepted. In essence this means that the parity norm and the requirement of reciprocal partner preferences are presumed to be in effect in the probabilistic situation as well as in the deterministic situation. 20 21 However, the "cheapest winning coalition" is not a useful concept in the probabilistic situation, since, in the probabilistic situation there is no "cheapest winning coalition." This follows from the definition of the probabilistic situation which rules out any a_priori "winning coalitions" as defined by Gamson (1961a), i.e., a coalition with sufficient strength to dictate the terms for the distribution of the payoff. Since there can be no "winning coalition", there can be no "cheapest winning coalition." The major assumption which was added to MRT so that it would predict behavior in the probabilistic situation was that in all situations the members of a triad will form the coalition which maximizes their chance of winning. Therefore, in the probabilistic situation the parity norm, although salient, has little effect in determining which coalition will form. Rather, the uncertainty of the situation fosters a need for security which will be designated the "security norm". Since any coalition will win with probability one in the determi- nistic situation, the members of the coalition will be secure, i.e., will win, regardless of which coalition is formed. In the deterministic situation the only manifestation of the security norm is that the members of the triad will form coalitions. The coalition that is preferred will be a function of the parity norm. As a result, if for any two possible coalitions, the probability of their winning is one, each player will prefer the coalition to which he contributes the largest share of the resources. The probability of winning is what determines the strength of 22 the players in a probabilistic situation and no player or coalition is guaranteed to win. Therefore, the security norm will be salient in the probabilistic situation and will influence preferred coalitions. If coalitions have different probabilities of winning, each player will prefer that coalition which maximizes his chance of winning. In order to maximize his chance of winning each player must form a coalition with the stronger of the other two players. Thus, it is apparent that the strong member of the triad will be included in all preferred coalitions in the probabilistic situation while the weak man will be the least preferred coalition partner. This means that in the probabilistic situation the "strength is strength" effect is predicted as Opposed to the power inversion effect which is predicted in the deterministic situation. Even though the security norm will be salient in the probabi- listic situation with respect to the preferred coalition, the parity norm will remain salient for the division of the payoff. This hypothesis is made because of the reasoning that the greater preference for the strong member of the triad as a coalition partner will give him a better bargaining position. Therefore, he will demand and receive the largest share of the payoff. _flypothesis As a result of the data which have been collected in the past few years, it was predicted that in the deterministic situation which allowed coalitions the power inversion effect would be observed. However, it was predicted that in the probabilistic situation which allowed coalitions the "strength is strength" effect would be observed. 23 That is, in the probabilistic situation each member of the triad will wish to form that coalition which maximizes his chance of winning. Therefore, he will prefer that coalition which maximizes the strength of the coalition relative to the isolate. As a result the stronger of the other two members of the triad will be the preferred coalition partner and the " strength is strength" effect will appear. A further prediction was that in both the deterministic and probabilistic pure truel situations, and the coalition situations in which no coalitions formed, the power inversion effect would be observed. This followed from the assumption that the players would attempt to eliminate those players who had the best chance to win, since by this strategy each player would maximize his chance of winning. As a result the weaker attack choice5 would be attacked the least and the stronger attack choice6 would be attacked the most, i.e., power inversion. The final prediction was that the parity norm would be salient in both situations which allowed coalitions. Specifically, the hypotheses tested in the present research were: (1) In the probabilistic coalition formation situation each player prefers the stronger of the other two players as a coalition partner, resulting in AB coalitions being the most frequent. 5The weaker attack choice for player C is player B, and the ‘weaker attack choice for players A and B is player C. 6The stronger attack choice for player A is player B, and the stronger attack choice for players B and C is player A. 24 (2) In the deterministic coalition formation situation each player prefers the weaker of the other two players as a coali- tion partner, resulting in BC coalitions being the most frequent. (3) In the pure truel situations and those games in which coalitions were permitted but not formed the power inversion effect would appear in the following manner. First, there would be a prOpensity for each player to attack the stronger of the other two players, and second, player A would win significantly fewer games than player B, and player B would win significantly fewer games than player C. (4) The parity norm would be evident in both the probabi- listic and deterministic situations which allowed coalitions. That is, the division of the payoff would tend to approximate the ratio of the resources contributed to the coalition by its members. Method Subjects. The subjects were 255 male undergraduate volunteers who received course credit for their participation. In addition, they were informed that they would have a chance to win $10 if they parti- cipated in the experiment. Apparatus. The apparatus for the game was 45 white poker chips; three wooden tokens, one inscribed with an A, one inscribed with a B, and one inscribed with a C; and a die. Plain white 3" X 5" scratch pads were used for all communication, and clipboards were used to allow the communication to remain secret. Game Design. The structure of the basic game was as follows. Each of the players began the game with 15 chips. The position that each subject played was determined by chance, i.e., each subject drew a token (marked either A, B, or C) from a hopper and played the game in the position indicated on the token. The rules of the game required that each player take away a given number of chips from one of the other two players on each move. Player A took away 4 chips, player B took away 3 chips, and player C took away 2 chips. All of the chips that were removed were taken out of the game and did not belong to any of the players. Each player wrote down the letter of the player he wished to attack and gave the message to the ex— perimenter on each move. At no time was a player's choice known 'before every player had made his choice, i.e., the moves in the game Ivere simultaneous. The players were told who had attacked whom after the experimenter had received the choices from all of the players. Procedure. Four conditions were examined: (Dc) a deterministic 25 26 game which allowed coalitions, (D) a pure truel deterministic game, (Pc) a probabilistic game which allowed coalitions, and (P) a pure truel probabilistic game. The same basic truel game was used to keep the conditions as similar as possible while varying enough to allow the game to be defined as either deterministic or probabilistic. Eighty-five triads participated in the experiment. Five triads (one Dc, one D, two Pc, and one P) were discarded either because the subjects had previous knowledge about the experiment or because two of the subjects in a triad knew each other. Therefore, there were twenty triads in each condition. Each triad played one game. Communication was limited, i.e., all messages were written and directed toward the experimenter. A pre-game questionnaire was administered to provide a measure of the subjects acceptance of the implied power structure. (See Appendix A for the details of the pre-game questionnaire) In the Dc and D conditions, each player took chips away on every move. In the Pc and P conditions, whether a player took chips away or not was determined by chance, i.e., each player rolled a die after he had decided which player to attack, if the die came up even he took the chips away, if the die came up odd he did not take the chips away. This meant that each player had a 50% chance (If taking chips away on each move in the Pc and P conditions. In the Dc and Pc conditions each subject was given a chance to forum a coalition prior to the actual play of the game. They were askxed to indicate on a form provided (see Appendix A), whether they 27 wished to form a partnership or not, and if so with whom. If two of the subjects indicated a reciprocal preference, they were given two minutes in another room to reach an agreement on how to divide the payoff. The payoff was a $10 prize which the winner or winning coalition in each game had a chance to win in a raffle. If no agreement could be reached in that two minutes, the game was played as a truel. If an agreement was reached, the partnership started the game with 15 chips and the power to take away the number of chips which was equal to their combined power, i.e., an AB partnership could take away 7 chips, an AC partnership could take away 6 chips, and a BC partnership could take away 5 chips. This meant that in the Dc condition any partnership was guaranteed to win. In the Pc condition, although a partnership did have a better chance to win than a lone player, a partnership was not guaranteed to win. For example, if an AB partnership formed in the Pc condition it had a 50% chance of removing 7 chips while C, the isolate, had a 50% chance of removing 2 chips. The expected number of trials for an AB partnership to eliminate player C was 3 while the expected number of trials for C to eliminate an AB partnership was 8. In all four conditions the game was over when only one player (Ir partnership had chips remaining. The winner or winning partner- stiip received a chance to win $10 in a raffle. After the game was completed all of the subjects were re- qiuasted to fill out a post-game questionnaire. (See Appendix A) 28 In addition, to ascertain whether there was any effect due to the subjects perception of the experimenter, the subjects were asked to evaluate the experimenter on a 19 scale semantic differential. (See Appendix A) The subjects were also asked to complete the "Survey of Interpersonal Values” (Gordon, 1960). This was utilized to determine whether there were any effects that were due to personality differences. Results Results Relevant to Hypothesis 1 Hypothesis 1 predicted that in the Pc condition each player would prefer the stronger of the other two players as a coalition partner resulting in AB coalitions being the most frequent. Table 2 presents a who-to-whom matrix for offers in the Pc condition. An analysis by chi-square supported the hypothesis that the stronger of the other two players would be the preferred coalition partner in the probabilistic situation (X2 = 8.43, p<.05). Further support for the ”strength is strength" hypothesis was observed in the significant difference in number of offers received by each player. (X2 = 11.29, p<.Ol). Player A received more offers than player B and player B received more offers than player C. Since only three coalitions were formed (2 AB coalitions and 1 BC coalition), a test of the second part of hypothesis 1, i.e., AB coalitions will be the most frequent, was not feasible. Results Relevant to Hypothesis 2 Hypothesis 2 predicted that in the Dc condition, each of the players would prefer the weaker of the other two players as a coali- tion partner, resulting in BC coalitions being the most frequent. Table 2 presents a who—to-whom matrix for offers in the Dc «condition. An analysis by chi—square indicated that there was no Siignigicant difference from chance expectancy with respect to who prxeferred whom as a coalition partner (X2 = 3.25, .50>p>.30). Thiriflner refutation of the power inversion hypothesis was observed in 29 30 the distribution of offers received, i.e., the distribution of offers received did not differ from chance expectancy. An examination of type of coalitions formed revealed 6 AB, 2 AC, and 2 BC coalitions. Al- though an accurate statistical evaluation of the situation is not feasible because of the small number of coalitions that were formed, it seems clear that the power inversion hypothesis was not supported. The data in Table 2 suggested that there was a tendency for the preferred coalition partner to be the same in the Pc and Dc conditions. Since the Pc and Dc conditions were independent, the coalition partner preference in the Dc condition was compared with the coalition partner preference in the Pc condition by computing an F-ratio of the chi— squares (Lindquist, 1956, p. 39). The F-ratio implied that there was no significant difference in preferred coalition partner in the Pc and Dc conditions (F=2.59, NS). Chi—squares to test the hypothesis that each player makes the same number of offers in the Pc and Dc conditions revealed a significant difference between players in the Pc condition (X2 = 7.23, p<.05), but no significant difference between players in the Dc condition (X=.37, .90>p>.80). An F-ratio of chi- squares indicated that there was a significant difference between the distribution of number of offers made in the Pc condition and the distribution of number of offers made in the Dc condition (F=9.62, p<.05). Although there was no significant difference between the total number of offers made in the Dc condition and the total number of offers made in the Pc condition, there was a significant difference between total number of coalitions formed in the Pc condition and total number of coalitions formed in the Dc condition (X2=3.88, p<.05). 31 Table 2 Who—to—whom matrix for offers in the Pc and Dc conditions Pc Dc Recipient Recipient 2 2 A B C Tl X A B C Tl X A 0 5 0 5 5.00** O 8 3 11 2.27 Offerer B 7 0 2 9 2.78 8 0 13 .69 10 7 0 17 .65 8 o 14 .29 8.43** 3.25 T2 17 12 2 31 16 14 8 38 *p<.01 ** xi = 7.23** xi = .37 p<°05 1 1 xi = 11.29* xi — 2.73 2 2 xi = 1.67 3 tests the hypothesis that each player makes the same number of offers. (df=2) 13‘... X tests the hypothesis that each player received the same number of offers. (df=2) HN tests the hypothesis that the total number of offers in A“... the Dc condition equals the total number of offers in the Pc condition. (df=l) 32 Results Relevant to Hypothesis 3 Hypothesis 3 predicted that in the pure truel situations and in those games in which coalitions were permitted but did not form, the power inversion effect would appear in the following manner. First there would be a prOpensity for each player to attack the stronger of the other two players; and second, player A would win significantly fewer games than player B, and player B would win significantly fewer games than player C. Table 3 presents the frequency with which attacks were directed toward the stronger and weaker player for conditions Dc, D, Pc, and P on each trial. To test hypothesis 3 adequately would have required that the relative strengths of each member of each triad be determined for each trial. To accurately determine the relative strengths of each member of each triad on a trial by trial basis would have required that the number of chips that the players had remaining on a given trial be taken into consideration. Since such an analysis would have been prohibitive, the analysis of the data on a trial by trial basis was conducted as if the number of chips that the players had re- maining on any given trial had no effect on the relative strengths of the players. Therefore, the analysis of the attack data for initial trial was the only analysis which was based on an accurate assess— ment of the diSparity of relative strengths. For all other trials the initial strength of each player was taken as his strength on that trial. An examination of the data in Table 3 revealed a significant power inversion effect on the initial trial in each condition, i.e., an... HCHNH. 33 .chHquaou om was on onu dH pocHmeo mums mGOHuHHmoo snow uoc vHuanoHns mUMHuu mmofiu %Hao H .mlm mHmHuu do mGOHqucoo m dam om ozu CH memo onu mH meMuum mo GOHuanHume onu umsu mHmonuomm: onu mumou UNN .m HMHHu co mGOHquaoo m pew .om .Q asp CH meow mnu mH mxomuum mo GOHuDQHuumHo onu umnu mHmmnuomhn onu mumou QNN .meHuu know umHHm man do chHUHwaoo HHm cH meow man mH mxomuum mo :oHunnHuume onu umnu mHmmnuom%£ mnu mumou me "muoz Hump «o. oq. Hw.H co. uoNM Numo 9V“ 0 " wwwvahi 3 Mm” 24 84. N. madman mo.va« w o .N II II mm. II 2 II *¢N.q we. «NN.¢ .xoq.m.. NN.m ,sz.q NN N e N NH HN mH oN NN HN mN NN Hoxmoz m a m S 2 AN mm 3 mm am 2 mm umwcobm II II aoq.o II nw.H. mm. , oo.H «mm.q oo.m om. xxwo.n...nm.N «amoqw,*««mN.qH ..... NM. H N H m m w mH MH wH NN 0H 0N mH NH umxmoz um N H m m 0H HH HN 0N om mN mm Hm mm mm Hmwcouum mm. «wo.q. 00.. .«oq.m. «a~o.w .NN 0 NH NN HN mH memos a m S mm mm 3 Hmwcohm 1|. .NHq. mH. ...mm.m1«««mH.0H NN N HH «H OH q HoMMo3 on e 2 3 ON 3 8983.0. «H mH NH HH 0H. m ..... w n. ..o m. q. n N H (HwHua .HHmHuu comm do GOHuHuaou sumo you ampthm uoxmoB m mHQMH paw Hmwcouum map oum3ou poHUOHHv mum mxumuum SUHSB nuHB kocmsvmum 34 the stronger player was attacked more often than the weaker player. A trial by trial analysis revealed that there was no significant power inversion effect on any trial after the initial trial in the Dc condition. In the D condition there was a significant power inversion effect on trials 2 and 4. In the Pc condition there was support for the power inversion hypothesis on trials 2, 4, 7, and 12. In the P condition the power inversion hypothesis was supported on trials 3, 4, and 6. To further test the power inversion hypothesis a sign test was used to examine all trials for each condition. If the strong man was attacked more than the weak man a plus was assigned to the trial. In the Dc and D conditions, there were five pluses in five trials. The probability that this would have occurred by chance was less than .05. In the Pc condition there were 12 pluses in 14 trials (p<.01), and in the P condition there were 9 pluses in 11 trials (p<.01) To compare the distribution of attacks between the four con- ditions, a chi-square was computed on all trials in which the eXpected attack frequencies were greater than 5. (see Table 3) This meant that all four conditions were compared on the first four trials, conditions D, Pc, and P were compared on trial 5, and condi- tions Pc and P were compared on trials 6—9. With the exception of the initial trial (X2= 6.28, .10>p>.05) the chi-squares revealed no indication that there was any difference between conditions with Irespect to the distribution of attacks. When the initial attack data for the combined Dc and Pc conditions were compared with the initial 35 attack for the combined P and D conditions, a significant difference was observed (X2= 4.95, p<.05, df=l). The second part of hypothesis 3, i.e., that player A will win fewer games than player B and that player B will win fewer games than player C, was partially supported by the data. Table 4 presents the number of games won by each player in each condition. An analysis by chi—square revealed support for the "weakness is strength" hypothesis in the Dc condition (X2= 8.33, p<.02), however, the power inversion effect was not reported in any condition. An examination of Table 4 indicated that there mignt be a difference between coalition and no-coalition situations in the distribution of games won. Therefore, F-ratios of the chi-squares were computed between the Dc and D conditions, the Dc and Pc conditions, the Pc and P conditions, and the D and P conditions. The resulting F-ratios revealed a significant difference between the Dc and D conditions (F= 69.42, p<.01) and no other significant differences. The significant difference between the Dc and D condition and the lack of a significant difference between the Po and Dc conditions or the D and P conditions was taken as support for the suggestion that a difference may exist between the conditions which allowed coalitions and those that did not allow coalitions. As a result the data from the Pc and Dc conditions were combined and the data from the P and D conditions were combined. A chi—square on the combined data indicated that there was a non-significant tendency for the power inversion effect tn) appear in the truel situations which allow coalitions (X2= 5.82, .(15
p>.05). There was an in—
sufficient amount of data to compute a chi-square for Pc-final payoff.
Results Relevant to Perceived Relative Power
To ascertain whether the subjects perceived player A to be the
strongest player before the game began, each subject was asked to
fill out a pre-game questionnaire. The questionnaire asked two
questions: (1) "Which player do you think has the best chance to win?",
and (2) "Which player would you choose to be, if you had your choice?"
A significant positive correlation between the answers to the two
questions in all conditions was obtained (r=.79 in Dc and D conditions,
.59 in the Pc condition, and .80 in the P condition, p<.0005 for all
conditions). Therefore, only question number 1 was used to determine
which player was preferred prior to the game. A similar question
on the post-game questionnaire was analyzed to determine which
player was preferred after the game had been played. Table 6 presents
the frequency with which player A was the preferred position before
and after the game was played. Analysis by chi-square indicated that
player A was also the preferred position after the game had been played
in conditions Dc, Pc, and P.
Since the post-game data were not independent of the pre—game
data, a three factor analysis of variance design with repeated measures
Vfias employed (Weiner, 1962, p. 337) to test the pre-game post—game
(finanges in position preference. Table 7 presents the summary table for
the:analysis of variance. There was a significant effect due to the
40
Table 6
Frequency with which A was the preferred player or one
of the other players was the preferred player in each
condition before and after the game was played.
pre-game post—game
2 2
A Other X A Other X
Dc 50 10 67.50* 34 26 24.07*
D 42 18 36.30* 21 39 3.20
Pc 56 4 86.70* 36 24 39.20*
P 46 14 50.70* 31 29 22.76*
*p<.001
note: Expected cell frequencies are A=20, Other=40 for the pre-game
data and A=15, Other=45 for the post-game data.
The differential expected frequencies result from the fact that
a "no preference" choice was offered on the post-game questionnaire
and was not offered on the pre—game questionnaire.
41
deterministic vs. probabilistic manipulation (F= 3.59, p<.05), and
the coalition vs. no-coalition manipulation (F= 10.59, p<.01) as well
as a significant difference between the pre-game and post-game data
(F= 77.70, p<.001). There were no significant interactions between
these'variables.
Table 8 presents the means and standard deviations for the data
used in the analysis of variance. An examination of Table 8 reveals
that (1) there was a tendency to prefer player A more in the pro-
babilistic situation than in the deterministic situation both before
and after the game was played, (2) there was a tendency to prefer
player A more in the coalition situation than in the no-coalition
situation both before and after the game, and (3) there was a tendency
to prefer player A more before the game than after the game in all
conditions.
Results Relevant to Individual Differences
To ascertain the effects of the personality variables that were
measured, a product-moment correlation was computed between each of
the six scales on the personality inventory and the dependent
variables in each of the four eXperimental conditions. Three out of
156 correlations computed were significant, i.e., in the Pc condition
"recognition" was negatively correlated with winning (r= -.37, p<.003)
as well as with.amount won (r= -.34, p<.009) and "independence" was
positively correlated with the amount won (r= .27, p<.039). Table 9
presents the correlations between the dependent variables and the
individual difference measurements for all conditions. The small
number of significant correlations between the dependent variables and
42
Table 7
Summary of the analysis of variance on the preferred
playing position before and after the game is played.
Source SS df MS F
Between Subjects 63.97 239
A(Deter. vs. Prob.) 1.01 1 1.01 3.95*
B(Coal. vs. no-coal.) 2 7O 1 2.70 10.59**
AB .08 l .08 .29
Subject within groups 60.18 236 .26
Within Subjects 44.00 240
C(Pre. vs. Post) 10.80 1 10.80 77.70***
AC .01 1 .01 .06
BC .00 l .00 .00
ABC .21 l .21 1.50
C X Subj. Within groups 32.98 236 .14
*p<.05 **p<.01 ***p<.001
Table 8
Means and standard deviations for the pre and post
game preferences of player position
Pre—Game Post—Game
Deterministic Probabilistic Deterministic Probabilistic
E SD 3? SD x SD E SD
Coalition .833 .373 .933 .251 .567 .494 .600 .490
No Coalition .700 .458 .767 .423 .350 .476 .517 .500
a
N = 240 60 replications per
cell
43
the personality traits which were measured was accepted as an indication
that those individual differences which were measured had a very
slight effect on either the strategies which were employed in the
game or the outcome of the game. A further indication of the lack
of a meaningful effect due to individual differences was obtained
from the analysis of the 19 scale semantic differential with which
the subjects were asked to evaluate the experimenter. (See Appendix A)
A principal—axis factor analysis with a varimax rotation (Williams, 1967)
collapsed the 19 scales to three factors, i.e., Factor 1, sociability;
Factor 2, capability; and Factor 3, emotionality. (See Appendix B)
An examination of the correlation of each of the three factors with
the dependent variables (Table 9) revealed that 1 out of 78 corre-
lations was significant, i.e., Factor 3 correlated negatively with
who was attacked in the D condition (r= -.29, p<.03l). Because of
the virtual lack of observed effects due to individual differences,
there was no attempt to correct for those effects in the analyses
of the data.
44
Table 9
Correlations between the dependent variables and the
individual difference measurements in the Dc, D, Pc,
and P conditions
Personality Experimenter
""""" Inventoryfi""" "'Eva1uation**
.......... .1 S "‘C" R""I"“B"‘ L""1""2 __3
Player -.08 .20 .05 —.24 -.O6 .10 -.15 .08 .00
Preferred partner .01 .16 .00 -.03 -.O6 -.05 -.12 -.10 -.22
Desired share .02 .13 .01 -.02 -.07 -.05 -.11 -.ll -.21
Expected share .01 .18 .01 -.05 -.06 -.07 -.13 -.07 -.23
Dc Final share -.02 -.03 .00 .09 .02 -.06 .00 .00 -.12
Attacked -.03 —.O3 -.10' .19 .01 -.10 .10 .13 -.17
Coalition -.08 .09 .05 .02 -.15 .06 -.05 .07_-.l7
WonHLost -.Ol -.06 .14 -.02 -.06 .02 -.08 -.01 .00
... Amount won ........ e.02.e.01...15.-.O4.-.07 ..04 -.07,-.03,..09
Player .01 -.10 .08 -.04 .02 -.07 .15 .23 -.04
D Attacked -.1O -.10 -.ll -.11 .13 .00 -.29d-.24 .01
WonHLost -.13 -.13 -.14 .ll .14 .06 -.04 -.09 .07
..Amount won ..—.13.—.13.—.14...11 .14 ..06 —.04 -.09 .07
Player .18 -.15 -.O6 .02 .06 .06 -.O4 -.22 .11
Preferred partner .05 -.05 -.21 .04‘ .22 -.08 -.12 -.20 -.06
Desired share .03 -.03 -.21 .04 .22 -.O7 -.11 -.19 -.07
EXpected share .03 -.07 -.22 .06 .24 -.08 -.12 -.21 -.04
Pc Final share -.07 .21 -.21 .06 .12 -.ll -.08 .05 -.21
Attacked -.09 .03 .07 .03 -.17 .14 -.03 .00 -.01
Coalition -.13 .13 -.09 .04 .05 -.01 .01 .03 -.21
Won-Lost -.06 .20 -.37a .25 .21 -.25 .05 -.03 .17
Amount won. -.04 H.12.e.34b_.279..20-.25 .07 -.o4...22
Player .15 -.14 .07 .16 -.15‘-.03 -.07 -.10 -.07
Attacked .01 .08 .02 -.02 -.04 -.01 .13 .20 -.04
WOn-Lost .01 -.O7 -.07 .05 .08 .08 .05 .13 -.05
.....Amount.won ......... .01.r.07.r.07 ..05,..08...O8. .05 .13,e.05
51 P<-603 2-tailed *S=Support I=Independence **Factor l= Sociability
1’ P<:009 2-tailed C=Conformity B=Benevolance Factor 2= Capability
C3 P