=7“;- ~ POLICY MAKING REGARDING THE DRUG PROBLEM: _ AN. EXPERIMENTAL STUDY OF COGNITIVE COMPLEXITY AND SMALL GROUP DECISION-MAKING - Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY. RICHARD IOHANN. OREND, 1973 I This is to certify that the thesis entitled Policy Making Regarding The Drug Problem: An Experimental Study of Cognitive Complexity And Small Group Decision-Making presented by Richard Johann Orend has been accepted towards fulfillment of the requirements for Ph.D. degree in Political Science Major professor Datew October 1,1973 0-7639 \ .5 .‘3“ ‘x 1‘ ‘ IIIIAII G SGHS' anex BINDERY INC LIBRARY BINDERs- “Ila-g! —'m" alums av ‘3 ‘5". A va‘: ~~+ V45 «3 n : ABSTRACT POLICY MAKING REGARDING THE DRUG PROBLEM: AN EXPERIMENTAL STUDY OF COGNITIVE COMPLEXITY AND SMALL GROUP DECISION-MAKING By Richard Johann Orend Despite a large political science literature on the behavior of decision—making groups little effort has been made to apply general social psychological theories to the making of policy decisions. This dissertation proceeds from the assumption that a social psychological approach, specifically the theory of cognitive complexity, may increase the predictability of the outcomes of small group decision-making processes. We argue that in groups composed of both "high" and "low" complexity individuals, the most common "real world" situation, the "low" complexity individuals would have a bargaining advantage and would dominate the policy output of the group regardless of the participants' attitudes. The theory of cognitive complexity describes various structures of information processing in individuals. This interest in the cognitive processes underlying individual and group decisions is not foreign to political scientists - witness, Herbert Simon's work on rational behavior in the face of numerous and largely unknown alternatives; witness, also, organization theory in general as well as various theories of .3” “v S. I #h in. b. v‘ file. I to»... «u "I "a” w 5..-- \r +r~p 2*"- A. “DO-‘C‘ :“3 Sc 5 3 ‘-~ ra‘mra but. ‘1. ‘- .— ..-v- ‘ . “-v-~ a...” .q‘... . . I u-)- g- n Nu-s . ... . . .. f. , L. 1. . e. .f . . _. . . f. . .r.“ . c“ ‘ ‘ aw... «. L. s u. . 4 -\x K; rt. .t .4 .. v n. *- .,,. w. s. 3 .. T. a. 3L .i I, L. . 3.. .. .r. ... .1 a. 3 .r .t P. i. :. .1 L... 7‘ w. r I. . ... r Mb .. v. I. .m at. ~m .ru Co v. .3 .. «o rt. n. sm 2.. u. . . J . .5 C +t e A?“ L y Q. “We h Mn. 9 2 C. .3 S o. C a AM s . . . . I :. .a J r. E T 3 a i c; at s . no a. an». N. A: An a v. .vW «L C. a ‘1‘ .m« C .W‘ Ca ~ru (\ a u“ :i .1 C :l C a we a t . a .. i . n. S I. L-.. e .4 Au .1 . 4.. F. m. 3 o“ &. Q: NH PG $e Au “L Gt. Amy G. 3: Q» C. :. a. .1 .l +. .h . it ”a .1 e S 5. 2 mm. C 1. a. an 1.. ac .(..\ A,» s i .1. a. a. at n: u. u. a . A. J u .. ~ I... a: A» 3.. A v .‘ ~ 2‘ A4. u v. z q «44 :u u“ 3.. u . 0‘ v :4 n .. ~: 3 g C. 5% A e .K. G. A. w 3y . a. .3 I. 0.1.. .\~ . . C. L. -.. h. A.» v. .. 1.. vs vs w... . . . . I... a. . . L . .1 .1 I a .5: A. ~ ». «5 G _ a». 2‘ Richard Johann Orend foreign ralations. However, it is in the social psychological literature that cognitive processes have received the most careful treatment. The idea of cognitive complexity has been fully explicated —hin a variety of ways — by Zajonc, Harvey, Hunt and Schroder, Tuckman, and Schroder, Driver and Streufert, among others. Common to all formulations of cognitive complexity are the dimensions of differentiation (the number of attributes used by an individual to identify an object or event) and inte- gration (the organization of the descriptive attributes). Our conceptualization of cognitive complexity and the instrument used to measure it are designed to measure these dimensions. An individual's level of complexity is the degree to which he exhibits a simple or complex structure toward a particular object, irrespective of content. A low complexity individual is characterized by categorical black-white thinking, lack of insight into new of different aspects of a situation, minimization of internal conflict, and absolute and rigid rules of integration. High complexity individuals exhibit a greater ability to see nuances, and a greater willingness to modify their positions. In a group composed of both high and low complexity subjects these behavioral propensities provide a bargaining advantage to the low complexity subject. The high complexity individual is more likely to see the need for compromise and thus modify his attitude position to produce agreement within the group. Our principal hypotheses are thus: IA: There will be differences in the recommendations made by experimental groups based upon the level of cognitive complexity of group members; i.e., in groups of mixed . ‘ 5““‘D\” *1' 0‘ I - . 4;- ~ ' ‘. .5..Ay \I' ‘ ’ I . ‘Q'V -- ;f‘vv fi“" ‘ H x “ ya - . ‘5‘,“ y».... .- . . . :duo-w 6" ‘a a» e‘ - JHV ‘ Dy, IAIY AA."“" 2 . _ - l n -Cm-- ..--‘ . C ' P ‘ ' q‘ J--- Ost "" v- Lose 5‘...“ C‘; I nqns*.av-~ I" | yv~---v..- O o .— '— ' A to ~..\—. g. ‘. \ “a V‘H \ ,_. .o. \..-.A 6--~A-~ ~~ . o . ~‘\‘OA-~ oh ‘ 5 ~ ~ a bang, -.- o AA“‘\ 9—.- . \C"'. ‘ "x ‘2: Vyo-vv- ‘~-_. A -u .0 -- ‘v. ' A. Ofi-Y 5n" 6- 1 a-~-¢ x.-- -\ . n. . ~ 1‘" :“ Ryaw ‘ID."“ " "“'\v 0v.....4-u~.: ~_.~AA ' “A .._,:\C ‘~\D. A” l 0 "" ”‘bv‘ y.- . o 1. _ . Al‘s-n cu : A ““VvsLx-‘T ~»-._.‘ \ ~-—' g... o _ - ‘ - u , . ~ -.~. . _" fi,"\‘—‘ ‘ H- “\‘t- 9-. \4. V H . - v 'upg-..‘ 1 ..‘~ A ‘ \ ~—‘..g, ( float AN .. ‘ " K.-. . . vo. - “In“ § -. .. Q s A -a-.. s -n- a“ u...\ H '4 i ‘ ~0'v'. .. 1‘1. CAWR‘-‘ . u __ Q O v ""‘t/: t., ‘ ‘ V .. ‘t. I .. A “.g “A.“ . E‘IVI‘D A‘ :A‘. ‘ “‘~» 7"“ H Ov :V."‘A“ ':V~. ESQ“ ”s s,“ . ‘y‘ :“-‘_.'S 'v:‘\ s 6r 5 _‘ ~ ‘.A‘. Q. ‘.‘ u ‘4. ._‘ -‘-.‘ ‘~<_ 2‘ :‘ ."}:- ‘J \ ‘ o s Q .5 ' ~‘~' .3“ l \- ‘x g Q-.. C“ -. H“. ‘H 5" »§ J: j w, . -‘MQS ' ( ‘ A ~! “ ., Q ~.',’ 2*». :¥\ NQ sag.-+Q) G -\_ . t3 :_, ~ . ~ .‘Y‘K ‘v .~ ‘~ ‘v V‘ § ‘ \' ~:(_ “a... e. 4 q \ .A AA . I" ‘van "- g; x Q ‘ ‘IV :h..~ H s .‘I. .‘ _‘ 5‘ - ‘-. ‘hc 5‘1 ‘ C V Vac? N 5.. v“ ‘5 . .. - a 't ‘\‘J:,‘ ‘ y‘a' ', ~S H ‘ A, ‘§ ‘ ' j-.. t a., ““ A. 5“!“ “v . I 5“ ‘ . ’L I s U ‘ n ‘5 ‘ I“" C ‘.M .~ R‘ . . ~ dag. “~S n c \o‘ . « (\A :‘ k‘J5- ‘ \~;‘ 7 .;_ ‘ x 3" \ PI , . JV oh 0 "D" V_‘ Cr.» ‘ ‘. ‘ Richard Johann Orend complexity levels (two subjects high and two subjects low complexity) the recOmmendations will favor the attitude position of the low complexity subjects. The low complexity subjects will "win" a greater proportion of the time when there is direct conflict in attitude positions. IB: When groups are composed of all high or all low complexity subjects, the mean proportion of "wins" for liberals or conservatives in both group types should not be signifi— cantly different from the overall mean. We also formulate a number of other hypotheses about group process based on our conceptualization and on preVious research in social psychology. In summary form these hypotheses state: 1) high complexity groups should have more complex recommenda- tions, 2) low complexity individuals should use more value arguments.and fewer "logical" arguments, 3) high complexity individuals should engage in more information seeking, and 5) high complexity individuals should introduce more new ideas into group discussions. Hypotheses were tested in a small group experiment using groups with all high, all low and equally mixed high and low complexity subjects. Each group had two "liberals" and two "conservatives" (on the drug issue used as the topic of the experiments) to establish a balance of attitude positions. Groups were asked to produce a common set of recommendations to "solve" the drug "problem." "Wins" and "loSses" were calculated on the basis of the proportion of suggestions introduced by "liberals" or "conservatives" which were accepted or rejected by each group. Subjects were paid student volunteers selected on the basis of complexity level and attitudes from a sample of about 1300 screening test subjects. 54.. -v >9 Y‘D Ania 0" R ”‘1- “a u” _ --- .‘gyu— a A HA‘n‘aw‘ \ .- v»..-.‘44~0 'VA pl R o . _. - u. . .. .; .i .. T T f .1 T T. t 3 r. E e a . e .5 r. a. at ..n by w. :. a. .1 F T E a S C ”a e 6 ti n. a. T "I C. sky .“ u.‘ 0“ D» a S C; e S .3. S C; .l a .. T. a. u E C .C T. C W nu . . 9. C a» ... A. n. v! s. 1.. i. ”A w. A: ‘1. :‘ u~¢ ~93 u :- R~ . . n. w, .u I: m . . A c . :2 I. a. G. ..¢ :m v. .. ... x: y . v. n.. a e u .. .. .. a L.» h‘ Richard Johann Orend The results of the central hypotheses were tested in a two—way analysis of variance model. The independent variables were group type (proportion of high and low complexity sub— jects) and sex (a H X.2 incomplete design). Attitude was held constant. Group process hypotheses were tested with analysis of variance designs and correlational analysis depending on the level of analysis. The findings supported the predictions of our central hypotheses. Low complexity subjects in mixed groups were able to "win" a significantly greater proportion of disputed issues regardless of their attitudes. In homogeneous groups the mean number of "wins" for liberals or conservatives was not signifi— cantly different from the overall mean. We are thus able to provide a better prediction of the kinds of decisions groups will make if we know the group is composed of individuals differ- ing in their complexity level and attitude toward the subject being considered. However, group process variables failed to illuminate the question of exactly how this result is attained in theintra— group discussion. At this point, then, we have a Skinnerian S-R model with no hard evidence about what, if anything, goes on inside the "black box" which would produce our results. Unsystematic observation of group discussions leads to the suggestion that hypotheses about the introduction and discussion of specific issues might provide fruitful results in future experiments. The chief problem in this area is that much of the process tends to be tacit. o quv‘N *FO "V .wv..5 Loos— . p- - :--,..:..-,.,... C" a“ 4.5...v‘uhyvvovai . _ ‘ O . u . ”V'H‘A.'\:RAS :“A "‘ -.v,v-.ovac “a.“ —.A.. .o - ... ”'55 w...“¢- .. - :vXtI o-ov a...» ‘.A§ v . c g 0 u: Annq:-+° F‘v—Qn‘ --.v va—‘Uvfiuv u-. v» .0 :r: .~"A“":F‘ LSA- g- guauv‘o but..- hurvv; . . - I A: :Ovugh‘uy‘na ‘6 ~ ‘Hv ... u»&‘. w KC- A on .‘ In “ ... ...t “""""""a cs VH-VVI-t ~ ...- ~ ... —~~n§ K —. 4 , - 9- E"‘ a.‘:L 5.: ~A .. . w - ~ Richard Johann Orend Among the group process variables we measured only the introduction of new ideas showed some positive support for our hypotheses and this finding was tempered by an interaction with sex. The amount of conflict in group types was actually in the Opposite direction of our predictions. Thus, our findings are important because they provide the first evidence that the structure of information processing has significant influence on the outcome of group decision-making processes. Our next goal must be to find out how this process operates. hu"77~sv'- .4 b V313 Loui‘s‘ H- :1 "U [U ’1 r4 ‘4. POLICY MAKING REGARDING THE DRUG PROBLEM: AN EXPERIMENTAL STUDY OF COGNITIVE COMPLEXITY AND SMALL GROUP DECISION-MAKING By“- Richard Johann Orend A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1973 I i’. if I? 1. .3 “% Copyright by Richard Johann Orend 1973 Iwish to that Prefessors Frank A P Q +c“‘ :cr tne though ti u- at this dissertati :iatisn to the cc: use initial impetu much needed er. exert. Such dedi If a COlmittee cha Several ACKNOWLEDGEMENTS I wish to thank the members of my dissertation committee, Professors Frank A. Pinner, Cleo Cherryholmes and Ada Finifter, for the thoughtful criticism they provided in the preparation of this dissertation. In particular, I wish to express appre- ciation to the committee chairman, Frank Pinner, who provided the initial impetus for the study and offered valuable advise and much needed encouragement through all stages of the research effort. Such dedication is more than anyone could have asked of a committee chairman. Several others have also made contributions to the completion of this research. Professor William D. Crano, of the Psychology Department, offered guidance in the area of the theory of cognitive complexity and encouragement in the early stages of the study. Mrs. Janet Eyster, then of the Statistics Department, provided vital suggestions on the analysis of the data. Without her incisive proposals for data analysis I might still be struggling with the large amount of collected data. My coders, Steve Fleming, Barbara Reynolds, John Grabber, Cynthia Carlson, and Michael Alfano provided astute and dedi- cated service. Steve, in particular, kept me alert to possible shortcomings in my measurement techniques. For financial assistance I am grateful to Dr. James Phillips, sumo made funds from Air ForCe Office of Scientific Research Grant iii 3.1+520-69-C-01114 a: test and experimenta tharjt Professor Cle. assistants to code I provided $830 for r‘. \4 “:1 ‘II ...itional coders . y. fiescann a. .h also prcvi ethese last two .. K. {5" ‘ ““a :flq‘. c was how 1ndi vii: 31 decision-ma}x fiv- \*A.€ intro 1! d.ces t 130 , late: ~- 5 q. ' EV Orig I ~.I\ c . ‘aQTCr~ 35.. ‘en,_ 7" 3 . ’ “hi“h .zjiv. . 5’3 6 image.26 Differentiation is the number of different elements or aspects in the image or essentially the amount of infor— mation we possess about another (nation).27 As our information (complexity) increases we are less likely to generalize and, therefore, more likely to be flexible in our dealings with other nations. Again we find the prOposition that the amount of information processed by the individual or group has a bearing on the nature of policy decisions. Pruitt has added several important dimensions to our -discussion, particularly in beginning to discuss the nature of the psychological variables to which Simon alluded. His concept of complexity (differentiation) brings into clearer focus how individual differences in information play a role in decision-making. Scott elaborates on this process.28 He introduces the idea of cognitive structure into the Political Science literature. Cognitive structure is the number and relationship of the attributes used by an individual to des- 29 The content of these attributes cribe a particular object. is not relevant to the nature of the structure, but it is important in describing the individual's overall image of the 'object. ,Scott has thus reduced the notion of individual information processing to its rudiments, free of content and affect, and seemingly capable of maintaining an impact on individual behavior. He does not treat cognitive structure as an isolated and independent variable. Rather it is related to other factors; e.g., affective, action and cultural com- ponents, which go together to form the total image held by the individual. It is the total image which bears on decision- u ' k C” sung and t..e 1-- “m“ve Structur 955“--- b the other dmens 1c structure of an 1:. Fpg‘“ .- ‘ v:-o5.a¢, more Cr ‘ .- . 3 3-3ect1ve, etc. ". Scott's co 0‘) of Pmitt's comple :a:et of cognitive ‘vo ' . m: defer 1!: ‘h * bit g ’ _ ‘ fiSL Pengly associ the Individual . C . :\‘,'v ‘~v‘ t this argume“ In) +‘ . ovkag cnshl P in “r. _ 099+. :Cr‘p:~ w x 5‘ r . 5* 1391le : . ate 21‘? the 232 7 making and the isolated effect of any of its parts is unclear.30 Cognitive structure, however, is the organizing framework for the other dimensions. Attributes which make up the cognitive structure of an individual may be affective, more or less central, more or less open, more or-less rigid, more or less objective, etc.31 Scott's cognitive structure is a more generalized version of Pruitt's complexity concept. Pruitt's complexity is one facet of cognitive structure, the information component. The two differ in that for Pruitt, complexity is considered to be strongly asSociated with the amount of information held by the indiVidual. Scott raises some theoretical reservations about this argument, despite citing an empirically positive relationship in one of his own studies.32 He argues that new information may be integrated into the cognitive structure in such a way as not to inorease the number of dimensions of characteristics attributed to the object. In other words, the new attribute reinforces the old without adding to the complexity of the cognitive structure. The narrowing of perceptions by decision-makers, discussed earlier, and refinement of the concept of cognitive structure are brought together in literature on the role of stress in foreign policy. Both Charles Hermann and Thomas Milburn investigate the problem of how increasing stress (such as a crisis situation) restricts the number of alternative policies considered.33 -In most research it has been found that increased stress limits the number of policies considered and restricts the degree to which various attributes associated 8 with the object of the policy are interrelated (policy makers tend to channelize thinking). Milburn and Hermann also note that the degree of restriction varies from individual to individual. They cite studies by social psychologists Schroder, Driver and Streufert to demonstrate that the effects of stress vary with the individual's ability to process information.3u As the complexity of an individual's cognitive structure increases, so does his ability to tolerate stress without restricting his alternatives. The function is not strictly linear but curvilinear. After a certain point (degree of stress) even the highly complex individual loses the capacity to Operate at a maximum level. The curve is more like an inverted U. We are not interested in stress, but in the fact that the ability to c0pe with stress is at least partially a function of an individual's level of cognitive structure. In any decision— making situation certain individuals will be better able to cope with the existing level of stress. «This increased ability .to adapt is related to perception, but there are almost no specific hypotheses about the outcomes of the decisions processes. For instance: What kind of policy alternatives areveliminated (Zinnes, North and Koch found threat was amplified at the expense of other considerations)? Are the types of alternatives expanded and reduced the same for individuals with a complex cognitive structure as for indi— viduals with a simple structure? When individuals of complex and simple structures interact, do the broader, more open 'e x x.“ Q ,. me C h’“ i to now we .ich suc .. '9 a." V Q ‘ocks Of t N“ Ud- .1. :cal facto e EOSt Pece 7A... sv‘ T; n. ‘al ». 9 outlooks of the complex prevail, or are they overriden by the narrow rigid views of the simply structured individual? Up to now we have taken pains to point out the situations in which such questions are appropriate in the political arena and some of the political science literature that has been focused on these questions. Simon has pointed out the possible role of limited information and probable effects of psycho- logical factors in helping to create this situation. Several scholars have pointed to the role of individual perceptions in decision-making situations and particularly to the effect of amount of information possessed by the individual. Scott theoretically separated the simple information from cognitive structure,.which deals with how a person organized his infor- mation regardless of the amount or content. It is only in the most recent literature (with the exception of Scott) that the question of level of cognitive structure-—we shall call it cognitive complexity--has been directly brought to bear on questions of decision-making. In the social psychological literature there has been a continuing, if not overwhelming, concern with this question since the middle 1950's. Although this literature has been concentrated on the effects of cognitive complexity on indi- vidual and group processes rather than on decision outcomes, it will form the basis of our theoretical discussion. Experi- mental research has been aimed primarily at isolating the effects of complexity level on intra—group behavior and is much less concerned with the output of groups, particularly groups of mixed, high and low, complexity subjects. 10 It is our intent to inquire into the questions of the effect of mixed group make~up on policy outcomes because we feel they are of greater significance to political science: first, because we are not interested in factors which do not affect political behavior or policies; second, because much research into the area of group made policy has not satis— factoraly answered the questions of the researchers; and third, because most policy making groups probably are made up of individuals of differing complexity levels.36 More specifi- cally, we will ask, what effects do differing levels of cognitive complexity among members of the same group have on who dominates solutions to a particular task presented by the group? We shall use a small group experiment to test various propositions about expected behavior, both at the level of group outputs and in terms of individual differences in behavior. The experimental method is used because of our rudimentary stage of theoretical and measurement development there is no other reliable and practical way to isolate the effects of cognitive complexity. In the next chapter we shall examine previous formulations of the theory of cognitive complexity, discuss relevant literature and formulate a set of hypotheses about group out— puts. We will also formulate hypotheses about group processes, particularly those related to previous research. In Chapter III we describe the measurement instruments and the experimental controls applied to isolate the effects of complexity. Chapter IV describes the findings of the experiment. Studies cf 1 Fred L StrC-i‘_ ”Eccial Status in Luz-1:: Cece Sidney 8. C1: (Sew Yerk: Genera David Rhcde , ”q: 3.? n I tawny Opinicn :r- ‘ s A ‘ Footnotes 1Studies of the Judicial System include: Fred L Strodtbeck, Rita M. James, and Charles Hawkins, "Social Status in Jury Deliberations," American Sociological Review, XXII (December, 1957), pp. 713- 719. idney S. Ulmer, Courts as Small and Not so Small Groups, (New York: General Learning Press, 1971). David Rhode, "Policy Goals, Strategic Choices and Majority Opinion Assignments in the U.S. Supreme Court, "Midwest Journal of Political ScienceJ'XVI, M (November, 1972), ‘pp. 652-683. Congressional studies include: Richard F. Fenno, "The House ApprOpriations Committee as a Political System," American Political Science Review, LIV (June, 1962), p. 310. Ralph K. Huitt, "The Congressional Committee: A Case Study," American Political Science Review, XLVIII (June, 195”), p. 3u0. Charles 0. Jones, "The Role of the Congressional Sub— committee," Midwest Journal of Political Science, VI (November, 1962), p. 327. James A. Robinson, "Decision-Making in the Committee on Rules," Administrative Science Quarterly, III (June, 1958), p. 73. Foreign and defense policy making studies include: Richard C. Snyder, H.W. Bruck and Burton Sapin, "Decision— Making as an Approach to the Study of International Politics," in Richard C. Snyder, H.W. Bruck and Burton Sapin, eds., Foreign Poligy Decision-Making} An Approach to the Study of International Politics (Glencoe, ILL.: The Free Press, 1962), pp. 194185. Ross Stagner, Psychological Aspects of International Conflict, (Belmont, California: BrockleoIe Publishing Co., Charles F. Hermann, "Alternative Theories of International Crisis Behavior," Paper delivered at the Conference on "Politi- cal Theory and Social Education," Michigan State University, February 5— 6, 1971. Thomas W. Milburn, "The Management of Crises," Unpublished paper, The Ohio State University, 1971. Herbert C. Kelman, ed., International Behavior (New York: Holt, Rinehart and Winston, 1965). Organization theorists: James March and Herbert Simon, Organizations (New York: Wiley, 1958), Chapter 6. 11 12 Herbert Simon, Models of Man (New York: Wiley, 1957), Chapters 19 and 15. Richard M. Cyert and James G. March, A Behavioral Theory of the Firm (Bnglewood Cliffs, New Jersey: VPrentice-Hall, , apter 5. 2James D. Barber, Power in Committees: An Experiment in the Governmental Process (Chicago: Rand McNally, 1966), p. 1. 3Prominent exceptions include: James Barber, Ibid. Robert T. Golemb1ewski, The Small Group: An Analysis of Research Concepts and Operations (Chicago: University of ChicagoPress, 1962). Harold Guetzkow, C.F. Alger, R.A. Brody, R.C. Noel, and R.C. Snyder, Simulation in International Relations (Bnglewood Cliffs, N.J.: Prentice-Hall, 1963. Sidney Verba, Small Groups and Political Behavior (Prince- ton, N.J.: Princeton UniVersity Press, 1961). ”Paul A. Hare, Edgar F. Borgatta and Robert F. Bales, eds., Small Grou s (New York: Alfred A. KnOpf, 1966). DSrW1n Cartwright and Alvin Zander, Group Dynamics (New York: Harper 8 Row, Publishers, 1968). 5 . Harold M. Schroder, Michael J. Driver and Siegfried Streufert, Human InfOrmation Processing (New York: Holt, Rinehart andWinston, Inc., 1967). BIbid. S1mon, 22, cit., pp. 151 and 159. 7See footnote 2. 8Simon, 92, cit., p. 2H2. 9Ibid., p. 29” and Robert B. Zajonc, "The Process of Cognit1ve Tuning in Communication," Journal of Abnormal Social Psychology, LXI 2 (1960), pp. 159-167. . Zajonc, a social psychologist, develops a theory based on d1fferences in perception which follows essentially the same prlnciples presented by Simon. We shall discuss his formula- tion in more detail in the next Chapter. 10Simon, op. cit., p. 2H3. 11Ibid., Chapter 6. lzIbid., p. 262. 13Schroder, Driver and Streufert, o . cit. O.J. Harvey, David E. Hunt and Haragd MT_Schroder, Conceptual Systems and Personality Organization (New York: Wiley, 1961). 13 James Bieri, A.L. Atkins, S. Briar, R.L. Leaman, H. Miller anle. Tripodi, Clinical and Social Judgement: The Discrimina- tion of BehavioraICInfOrmation (New York: John Wiley, 1966). (luSee footnotes 1, foreign and defense policy making studies. 15Snyder, Bruck and Sapin, Ibid. 16Richard Snyder and Glenn Paige, "The United States Decision to Resist Aggression in Korea," in Snyder, Bruck and Sapin, 92. cit., p. 229. 171pm. 8Stagner, loc. cit. 19James A. Robinson and.Richard C. Snyder, "Decision- Making in International Politics," in Kelman, 92, cit., pp. 391-”32. 20 21Dina A. Zinnes, Robert C. North and Howard Koch, "Capability, Threat, and the Outbreak of War," in James N. Rosenau, ed., International Politics and Foreign Policy (New York: The Free Press, 1961), pp. “69-982. 22 Ibid., p. nun. Ibid., p. M76. 2301e R. Holst, Richard A. Brody, and Robert C. North, "Measuring Affect and Action in International Reaction Models," Peace Research Society (International), Papers, Vol. II (1965), pp.>170-190. 2”Dean G. Pruitt, "Definition of the Situation as a Determinant of International Action," in Kelman,'gp, cit., pp. 391-”32. 261bid. 27Ibid., p. u13. 28William A. Scott, "Psychological and Social Correlates of International Images," in Kelman, 22, cit., pp. 70-103. 29Ibido, Pp. 78-79. 30At least in this article. Some later work examines the cognitive structure aspect of images directly and Scott comes to conclusions similar to those of other Social Psychologists interested in this phenomenon. We will discuss this research more fully in Chapter II. 1” 31Ibid., pp. 80-81. 32Ibid., p. 85. Problems in measuring'cognitive struct- ure could explain the descrepency between the theory and the empirical finding. 33Herman, loc. cit. Milburn, loc. cit. 3”Schroder, Driver and Streufert, loc. cit. 35We shall discuss the relevant literature in greater detail in Chapter II. 36We are taking this somewhat exploratory approach in full knowledge of the fact that the theory and research has not provided the conceptual clarity and empirical validity which might be desireable. n i. . We adthcrs HL' k mic.sna11 gp-. “n r: . CHAPTER II THEORY AND LITERATURE In the first chapter we examined a body of Political Science literature in which the possible effects of cognitive complexity on small group decision-making were considered. Several authors reached the conclusion that the process by which small groups make decisions and presumably the nature. of the decisions will be affected by the complexity level of group members. In this chapter we shall be concerned with clarifying relationshipsbetween complexity level of group members and the nature of the decisions made by groups whose members differ in both complexity level and attitude. We shall examine the conceptualization of cognitive complexity, formulate some general propositions about behavior associated with individuals differing in complexity level, discuss the relevant literature concerned with providing evidence about the varioustheoretical formulations of complexity, and formulate sOme specific hypotheses on the functioning of groups composed of individuals of different complexity levels. Conceptualization "The way an individual receives, stores, processes, and transmits information" may be referred to as his "level of "1 conceptual structure. "Thus, beliefs, hypotheses, attitudes, 15 16 needs, concepts, and so forth may be viewed as structures for processing information."2 Cognitive structures are, therefore, the hypothetical link between the stimulus infor- mation, on the one hand, and the ensuing judgment about their meaning and appropriate reactions on the other.3 In studying struCture we are deliberately eliminating consideration of content and considering only the degree of articulation and the organization of cognitive systems.” Whether an individual holds a particular set of beliefs, is liberal or conservative oris manifesting a particular stage of personality develop- ment is not relevant to this formulation. What is of signifi- cance is the particular way a person organizes the content of this judgment. It is the nature of the organization which will be called "cognitive structure."5 Several dimensions of cognitive structure may be distinguished. Zajonc, for example, preposes four: degree of differentiation (the number of different attributes pro- jected upon an object); degree of complexity (the extent of the organization of those attributes); degree of unity (the interdependence of attributes); and degree of organization (the extent to which one or a few parts dominate the strucé ture).6 Schroder, Driver, and Streufert add discrimination, which is "the capacity of the conceptual structure to dis— tinguish among stimuli", to differentiation and integration (Zajonc's complexity) to form their model of the components of cognitive structure.7 Scott discusses several other dimensions of cognitive structure, including attribute centrality, attribute articulation (clarity), affective salience, and a: openness-closed: standing the di'. presented by of: all major conce; are differentiat We shall I“ ‘1) (fl cognitive CC c. attributes a in his environ: limited to onl‘ (‘31.: ...ch are certs: l7 salience, and ambivalence.8 Harvey, Hunt and Schroder propose '0penness-closedness as still another dimension.9 Notwith— standing the diversity of these and additional dimensions presented by other writers there are two common elements to all major conceptualizations of cognitive structure. They are differentiation and integration.10 We shall refer to the combination of these two elements as cognitive complexity; that is, the number and relationship ,of attributes an individual uses to define some phenomenon in his environment. Cognitive complexity is thus being limited to only those two dimensions of cognitive structure which are common to most other conceptualizations, differenti- 11 This formulation does not preclude ation and integration. the possible significance of any other structural dimension, but merely attempts to isolate the minimum neCessary elements of cognitive structure as it has been described in the literature. Differentiation is simply the number of attributes (characteriStics) used by an individual to identify an object or event.12 These attributes may include descriptive, affective, belief, value or any other characteristics. Differentiation includes as many or as few of these characteristics as are required by the person to orient himself to the object or situation which he perceives as being relevant. 2 Integration is the organization of the descriptive attributes in a given cognitive structure. The attributes may be conceived as coming "from a single class or category of discriminanda, or they may represent many categories."13 18 Zajonc uses the example of a painting which may be perceived in terms of its objective qualities, size, subject matter, type of frame, etc. Or, it may be perceived partly in terms of objective qualities, partly in terms of formal qualities (period, style, symmetry, etc.) and partly in terms of its impact on the viewer.lu The number and intricacy of the discriminations among attributes constitutes the degree of integration of the individual.c In Zajonc's example, per- ceiving the painting only in terms of its physical characteristics would constitute a rather low integrative structure, no matter how many attributes (differentiation) there were. A subject's use of all of the mentioned dimensions would indicate a more complex integrative structure. An individual's level of cognitive complexity is the degree to which he exhibits a simple or complex structure with regard to a particular object. We shall refer to simple structures as having low complexity (or as being cognitively simple) and to complex structures as exhibiting high complexity (or as being cognitively complex). The designation is, of course, relative to the group of individuals being described. A low complexity individual perceives relatively few des- criptive attributes in a particularly uncomplicated way. 15 At the extreme end of the scale attributes See Figure l. are also isolated from one another. The cognitively "complex" person tends to perceive a large number of attributes and these attributes are interrelated in some 19 Figure 1 Low Complexity Representation Lines represent connections or "rules" for perception of the object. "Rules" are in a fixed relationship. - II II II Attributes 16 manner. See Figure 2. The particular subset of possible attributes used to identify the object for the individual is Figure 2 High Complexity Representation "Rules" are in an interdependent relationship; making new rules possible. Attributes not relevant in determining integration level. We are interested in the degree of interrelationship, not its content. Once we know the number and degree of inter- relationship of an individual's perceptions of a particular object; i.e., an indication of how he thinks, we should be able to predict, at least partially, how he will behave in certain situations. 20 Our approach thus far has limited cognitive structure and complexity to a single object or event; i.e., it is "issue specific." This is consistant with Zajonc.l7 Others have developed the concept of complexity as a general char- acteristic of cognitive functioning;18 i.e., individuals exhibit the characteristics in all situations. The "issue Specific" test we use does not preclude the possibility that -comp1exity is a general characteristic. Since our research concerns drugs throughout testing and experimental conditions we are unable to provide direct evidence on this issue. But we may assume that if cognitive complexity is a general characteristic, subjects who score as cognitively simple or complex with regard to drugs would exhibit the same characteristics in other issue areas. The fact that our formulation of complexity is tied to specific issues does not alter the content free element in the conception. Within each issue area the conceptualization and measurement of complexity is free of content (attitudes, values, beliefs, etc., about the issue). The fact that com- plexity level may vary across issues does not tie complexity to the content of the subject's perception of the issue. I may be very complex on the drug issue regardless of which specific attitudes or beliefs I hold about drugs. On the Other hand I may have a.rather simple approach to income tax, again regardless of my attitude, etc., toward it. In neither case is the determination of my level of complexity determined by my attitudes on the specific characteristics I use to describe it or the way I interrelate those characteristics. v.) Cozplexify 1,. r? the degree or v I In usmg c. we also must tre: differentiation Bieri, for exam: determinant of " differentiated ‘ ".1. ...e degree to w EEIEI‘C'gETZEOUS CC tie integrative ...e way in which environment is " approaches, to”: A "'i\‘ .~.econdition to 0-7 . “alone incl‘ 21 My complexity level is simply the number of attributes and the degree of their interrelations. In using our conceptualization of cognitive complexity we also must treat the question of the relative role of differentiation and integration. Previous theorizing has 19 usually stressed one or the other. Witkin, Crockett, and Bieri, for example, emphasize differentiation as a major determinant of behavior.20 The interrelationship of the differentiated aspects of cognition is of less concern than "the degree to which information processing occurs through a heterogeneous conceptual state."21 Harvey, Hunt and Schroder; Scott; Tuckman; and Schroder, Driver and Streufert emphasize the integrative dimension of complexity.22 They argue that the Way in which individuals relate their perceptions of the environment is more than the number of perceptions. In all approaches, however, the role of differentiation as a partial precondition to integration is recognized.23 Zajonc includes both dimensions in theorizing and measurement without attempting to emphasize either.2u Despite separating them conceptually he does not offer individual predictions about their independent contributions to behavior. He also measures them together. We shall also conceive of cognitive complexity in this balanced way for reasons already alluded to above. Differentiation and integration seem to be the fundamental dimensions of cognitive structure. There is little theoretical or empirical evidence that effectively separates them into independently operating factors.25 In fact, they are almost always conceived of as interdependent, “”erent type s Y‘-‘ A i"tt'xw'duals wit' t." : '0 c .neir level C- difference doe s each is unequal ;:,,L : “‘5‘; integrat: 6+ :~‘ .1 .egration difc _ serentlat i: h ;e . lRCrease F: Qhaviol‘a‘u or our ‘f‘un .K'\ I .: VJ? ‘E \ 22 although the relative level of each may play a role in different types of behaviors. Individuals with high differ- entiation and low integration may behave differently from individuals with the Opposite characteristics, even though their level of cognitive complexity may be the same. This differencedoes not indicate that the relative importance of each is unequal.26 Our empirical interest in this research is with individuals of high and low complexity levels. We shall not be concerned with persons in the middle ground where this question is most relevant. At a low level of complexity both differentiation and integration will be low. Only in the middle range, where differentiation increases enough to create the capacity for high integration will the question raised above be of concern. At high levels of complexity it will generally be necessary to have relatively high levels of both differentiation and integration. There is no reason to assume that high levels of differentiation will automatically bring about high cognitive Complexity, although a slight correlation could exist due to the increased capacity. Behavioral Consequences of Differing Complexity Levels For our purpose cognitive complexity may be conceived of as a continuous variable with individuals normally distributed along a single dimension. Differences in the level of cognitive complexity lead to different behaviors, but given the primitive state of theorizing, we can deal only with the degree of dissimilarity in certain kinds of behavior rather than quali Earvey, HJ Streufert, plexity 1e I ‘ I 33.21 DOLE. thn‘ «\‘Jiver’ u b) E E ‘v C idpthc \- r‘~' . i‘ch. 23 than qualitative differences predicted by some scholars. Harvey, Hunt and Schroder, and later Schroder, Driver and Streufert, for example, have developed a typology of coma plexity level andiibehavior.27 This typology includes four nodal points or general levels of complexity; the authors feel that this classification scheme could serve as a basis for describing qualitatively different modes of behavior. Since we are using a somewhat different approach our concern will be with the first and fourth "systems" (the author's term for lowest and highest complexity levels) rather than with all four.28 It is in these systems that the individual and group behavior patterns of interest to us are most prominent and, therefore, most suseptible to testing. Low Complexity: The lowest level of cognitive complexity is characterized by compartmentalization of a small number of attributes. Each attribute is isolated from the others, as illustrated in Figure 1, and the attributes tend to be hier- archically organized.2g This means that the object being perceived would tend to be seen unidimensionally. The fewer the numberof attributes perceived, the lower is the potential for generating internal conflict; i.e., dissonance, and the greater is the likelihood that potential conflict will be resolved by excluding the potentially dissonant attribute from further consideration. Attributes will also tend to be dichotomous. Schroder, Driver and Streufert argue that "a single hierarchy of rules for stimulus placement in a given category, which is compartmentalized. . . , has little "30 potential for developing scaled dimensions. Stimuli tend 24 to be seen in yes-no categories, either they do or do not fit. Less categorical discriminations will emerge, when it is possible for the individual to apply more than a single inter— pretation to the stimuli. Schroder, Driver and Streufert summarize this level Of cognitive complexity in the following manner: "In information processing, a concrete structure (low complexity) has comparative certainty and deter— minate character. Stimuli are evaluated more or less unidimensionally, and, from the subject's point of view, the problems of choice and error arise less fre— quently. Rules can be explicated more definitely and there is a minimum of ambiguity."31 In this pattern of cognition there are three characteristics present: (1) Stimuli tend to remain in the original categori- zation and are "minimally affected" by placements on other dimensions (compartmentalization). (2) "New stimuli are either distorted to fit existing dimensions or excluded." (3) Environmental conditions may affect categorization, but have little effect on the level of complexity. The behavioral characteristics associated with this level of complexity include: (1) Categorical, black-white thinking. There is a reduced ability to see nuances Or gray areas. If Blacks are perceived as being "good" or "bad", all Blacks will be seen the same way.32 Corresponding to this categorical way of thinking will be a lack of interest in new and different aspects of the relevant situation. The amount of new infor— mation sought will be low because the issue has already been decided and any new information that does arise will tend to be forced into an existing category or excluded. (2) There w Stimuli either f bility for alter is quickly mini: (3) "If a there is a cor: processes, and to arise.“J3 increase cont: 1": j a Ccmpi N1. ‘8 \ ’HEabS :(.rihutqs QiEh L_ . a "it n \."e 25 (2) There will be a tendency to minimize conflict. Stimuli either fit or are excluded. There is little possié bility for alternative ways of viewing the subject. Dissonance is quickly minimized or resolved. (3) "If a stimulus is categorized in an absolute way, there is a corresponding restriction of interval integrative processes, and alternative resolutions or interpretations fail to arise."33 Therfore, if low integration exists there is an increase control of "external stimulus conditions." With _greater integration the same information can generate more alternative interpretations, thus increasing the role of "self" as an agent, "going beyond any single or externally given interpretation, and (increasing) the conception of internal causation. ' (4) The more absolute and rigid rules of integration will, when they are changed, produce a greater more abrupt change. Conflicting interpretations tend to be "warded off" because of the lack of ability to sense shades of difference. If, however, the changesin a situation reach a certain threshhold, the categorication of the individual will change rather abruptly.3u High Complexity: The highest levels of cognitive complexity are characterized by a large number of attributes and a complicated interrelationship among those attributes. The hierarchical organization present in low complexity levels disappears and is replaced by a more flexible organization of 3” The larger attributes and rules for relating attributes. number of attributes generate a greater likelihood of conflict, but the flexible organization minimizes the effects of that conflict. The it can also more ea: of more aspects 1 those characteri ences is less 1: to be less signj C ‘3! LI: “*gh COT?" . a. L hirgs i n an a? 3:}i H ‘ C, . ls t} ‘zle: 1:1 26 conflict. The individual is able to see differences, but he can also more easily account for them through the utilization of more aspects of the situation and/or the interrelating of those characteristics available. The larger number of differ— ences is less likely to be disturbing because they will tend to be less significant. That is, the complex individual's ability to form more intricate scales for judging produces the ability to observe differences not observable in less complex subjects.36 But he is also better able to assimilate theSe and larger dissimilarities. He is able to apply a greater number of interpretations to what he observes which means that he can make greater use of more information and fewer observations are excluded. High complexity peOple develOp the ability to deal with things in an abstract manner; what Schroder, Driver and Streufert call a "theoretical" rather than an "empirical" 37 This faculty results from the ability to manipu- orientation. late large amounts of information in such a way as to be independent of the current empirical situation. The individual can generate laws or principles which transcend the immediate empirical relationships. High complexity facilitates the development of alternative ways of perceiving the same pheno- menon without "the imposition of new external conditions."3 AbstraCt functioning also provides the individual with a more effective means of adapting to a complex changing situation. A high complexity level leads to cognitive functioning which is the opposite of that associated with low complexity individuals: (1) The laPE extensive interac open thinking. force diverse at‘ fiances are pero They are tied to (2) The dif and managed. Th these difference tolerance for cc be threatened 13., Site IO terms w: P5? of his com change as the 1 - < SltuatiOn . ”e 27 (l) The large amount of information processed and the extensive interaction between attributes contributes to more open thinking. There is less tendency to categorize and. force diverse attributes into inapprOpriate categories. Nuances are perceived and dealt with as individual cases. They are tied together at a higher (more abstract) level. (2) The differences between stimuli will be recognized and managed. The ability to perceive nuances and to integrate these differences in more than one way will create a greater tolerance for conflict. A high complexity individual will not be threatened by thinkind Of conflict and will more easily come to terms with it because Of a greater ability to modify part Of his conceptual system. He need not make as great a change as the low complexity individual in the same conflict situation. Hewill also be more flexible in handling dissonance.39 (3) The role Of the "self" in generating alternative solutions to problems is also increased with increasing complexity.”0 In a complex situation the complex individual would be able to generate a greater number Of problem solution alternatives independent Of external impetus. The same information can generate more alternative interpretations. It will also allow the complex individual to recognize the validity Of alternative solutions to problems. (u) The high complexity individual will be more likely to change his own Opinions on a subject because that change is likely to be much smaller and less abrupt than would be true Of the low complexity individual. A "shade" Of difference prcdu. a change is much vi him the frame:- villingness to C? in decisions. A1 does the perceiv not entirely ccr The abilit‘ +41 . ““9 PPOITLOtin E a 1 ' ' enabling the 1?. ”ham”, etc .) wun-terrart. - .0. h..at8vep inf: uuc< consideraticp nc+ mean 111 Eh A}, h yuanses 3 but '9- OVEI‘whel m: “Ahg v- 28 difference produces a "shade" Of change in behavior. Such a change is much less threatening and much easier tO handle within the framework Of a complex cognitive structure. This Willingness tO change will also be accompanied by uncertainty in decisions. As the number Of alternatives increases so does the perceived likelihood that a particular alternative is not entirely correct. The ability tO perceive conflicting and subtle alternatives, while promoting minor change, works against radical change by enabling the individual to modify his position (attitudes, behavior, etc.) to a lesser degree than his low complexity counterpart. Any change will be a product Of arguments (or whatever inducements) that are subject to the same cognitive consideration that produced the original position. This does not mean high complexity individuals will not make radical changes, but that they are much less likely to encounter the overwhelming pressure (Of whatever type) to make such changes ‘than low complexity subjects. One final set Of possible relationships requires discussion. We have stressed the structural nature Of the concept Of cognitive complexity as Opposed to content and affective elements Of cognition. The exclusion should also include othercommonly used individual cognitive attributes, namely I.Q. and information. We would not expect the indi- vidual's level Of cognitive complexity tO be significantly related to either his intelligence or the amount Of infor- mation he had about a particular subject. In both cases the procedures for measuring these attributes are important tO ISL-’5 [iii I-".‘u :‘ \“I‘E our discussion. measured in can some conceptual our ideas Of di concerned with information. I With both COEI‘:1 notable being J a general agI‘e researchers, a‘ is .. ‘ el‘eaper th "Mb idle Pelat Shul ar. Inf; 5 .e related tO St UdEnt W‘FC abQLt a sub. 1.‘ 3%.“ "W0 a (20‘ 31:“: Herer \‘ er‘t‘ s la*. . . “if“ a‘bt _ large .ld hot n eCc: 29 our discussion. Intelligence has been conceived Of and measured in many different ways. It would be unlikely if some conceptualization did not include concepts similar to our ideas of differentiation and integration, since both are concerned with cognitive processes and the processing Of information. In addition, some scholars have been concerned with both cognitive sturcture and intelligence, the most ”1 TO our knowledge, however, our notable being Jean Piaget. conceptulization Of cognitive complexity is quite different from any general approach tO intelligence.”2 We would there- fore expect an empirical relationship only insofar as there is overlapping in measurement procedures. This position is in general agreement with that taken by most cognitive complexity researchers, although there is some evidence that the overlap is greater for some measures Of complexity than others.I43 The relationship Of information to complexity level is similar. Information about a particular subject area could be related to differentiation in that area depending on the kind of test used. In our own research, for example, the use of an information measure dependent upon the number Of different aspects of an Object perceived by the individual could raise his total complexity score. For example, one can conceive Of a student who possesses a great deal of memorized information about a subject, but who cannot integrate the various details into a coherent pattern. Such a person might do well on the differentiation part Of our complexity test because he could list a large number Of descriptive details. Such a result would not necessarily mean that the individual would be 30 classified as being complex since that score is also dependent u” The integration on the amount Of integration exhibited. element would necessitate that the individual be able tO relate the various aspects Of the Object he perceived. Possessing a large amount Of information about a subject does not mean that it is organized in a complicated way. Using our previous example, we may know a large number Of physical characteristics Of a picture and find them all important in describing the picture, but we may not be able to relate these characteristics tO other aspects. Previous Experimental Results There has been a steady flow of research results on cognitive structure since the middle 1950's, but results are Often difficult to compare. The theory and methodology (particularly measurement) vary a great deal from one researcher to the next. The differences in theoretical orientation (e.g., differentiation vs. integration, issue specific vs. generalized, interpersonal vs. all perception and the number of different dimensions involved) are only exceeded by the number Of different measures Of cognitive structure.“5 We have attempted to skirt many Of these problems by using a narrow definition Of cognitive complexity and limiting our interestto those areas which are common to most formulations. In choosing Zajonc's measurement technique we have selected that measure which we feel best manifests the aspects Of cognitive structure we describe.”6 When atte.‘ problem Of mul‘ in the light 03 that most of t is curious tha hese studies for the conver basic elements all theoretica which accurate the kinds of c t E Cannon to meet each of these least some of SEC 31 When attempting tO compare experimental findings the problem Of multiple approaches cannot be avoided, particularly in the light of findings such as that Of Vannoy, Cox, and others that most Of the measures are not empirically related.“7 It is curious that in spite Of this confusion, many findings Of these studies are remarkably similar. One likely explanation for the convergence Of findings is the presence Of the two basic elements Of differentiation and integration in virtually all theoretical models, and the use Of measuring instruments which accurately assess at least one of these dimensions. The kinds Of general behavior predictions made above are common to most formulations Of cognitive structure and in each of these formulations there is evidence to support at least some Of the hypotheses, despite the measurement differ— ences. These results make it useful to briefly discuss earlier research findings. We shall limit our discussion tO the four behavioral areas mentioned in the previous section.“8 Otherfactors, such as, the effects Of stress and failure, types Of group initiated organization, leadership patterns, and different levels Of environmental complexity, will not be considered here because they are not part Of our research interest. (1) A finding supported by most experimental research is that complex subjects tend to engage in more information search. Three studies using the Sentence Completion Test (SCT) in widely varying situations exhibit this result.”9 (See Appendix I for a brief description and comparison Of the major cognitive complexity tests used in the literature). In addition, Tucicnar " 4 23‘? and the rep . 50 - same result . , ga£,but th 385 However, the light of some cor hcncgeneous dyad: Streufert and Ca: infcrnation seek C. levels." Envip present in the e C Pitter‘n for tie N as PEVeI-xse’: a+ L 38a rch Op the a with th' ”13 Capa: (2) The; a“, g 6~ ~ . bltlltition i S We 5 a steep \9 1 Inter‘rla“ Sm- \ H 1964» l Se “ a digging 32 addition, Tuckman, using the IFT and Lunderg, using both the IFT and the Rep test as indicators Of complexity achieve the sameresult.50 Driver has the same finding in a Stock Market game, but the measure of complexity is not reported.51 However, these findings must be carefully weighed in the light of some complicating factors. In an experiment using homogeneous dyads playing the Inter—nation Simulation Game, Streufert and Castore found no difference in self-initiated information seeking among subjects Of differing complexity levels.52 Environmental complexity and the amount Of stress present in the experimental task produced a modification to the expected pattern. Streufert, Suedfeld and Driver found that self-initiated information search was higher for sample subjects in a low stress situation, approximately equal at moderate levels of stress, and lower at high levels Of stress. However, while delegated information search showed a similar pattern for the lOw and moderate stress levels, the pattern was reversed at high stress levels.53 There is no consistent evidence that high complexity leads to greater information search or the ability tO see the nuances supposedly associated with this capability.5u (2) The ability Of high complexity individuals and groups tO recognize and deal with a larger volume Of aspects Of a situation is well documented. In two studies by Driver, one using a Stock Market game and the other using a version Of the Inter-nation Simulation, it was found that high complexity subjects use a greater number Of dimensions when making judgments and are more attentive to complex information.55 In the former 'ndividuals at The significar "best" strateg ness of a par“ on the situat. Several findins- PLO CC’r-Plexity it“. see nope aspe in the respec COEP'leXity St; "good—bad" A: Mi hlgh compleXi 33 In the former experiment it was also found that low complexity individuals attend to less complex and more salient information. The significance Of this finding is the implication that the "best" strategy is not always the most complex. The useful— ness Of a particular approach to problem solving is dependent on the situation not just the Subjects' level Of complexity. Several studies using the Rep test duplicate this finding. Plotnich, Price and Campbell all find that higher complexity individuals are more likely tO differentiate, or see more aspects Of, the particular subject being considered 56 Campbell finds that low in the reSpective experiments. complexity subjects tend to make judgments only along a "good-bad" dimension. Plotnich and Campbell also find that high complexity is an aid in making "correct" perceptions in the experimental situation. This finding is again subject tO the limitation Of the task being performed. The ability to differentiate and absorb a wider variety of information, among high complexity subjects, carries with it the capacity to tolerate conflicting information. High complexity individuals are less threatened by the conflict generated in a wider variety Of alternatives perceived.~ This tendency has been noted in a number of different studies. Mayo and Crockett, Tripodi and Bieri, and Nidorf all find the tendency to discriminate and integrate conflicting information higher in cognitively complex subjects than in cognitively I 5 7 simple subjects under certain conditions. ‘All these experiments were done using measurement instruments biased 3n toward the differentiation dimension Of complexity and should be interpreted in that light. Other authors have beeninterested in the tolerance Of ambiguity from the theoretical perspective Of balance theories. Scott and Crano and Schroder, using more integration oriented (instruments (the Object Sorting Test and the SCT), provide further evidence that a high complexity subject has a higher 58 This finding could tolerance for incongruent situations. help to explain previous inconsistencies in the research Of balance theOrists, particularly in the important area Of what 'kind Of decision processes an individual uses to resolve conflict.59 Crano and Schroder found that complex individuals were not "internally consistent" in the use Of resolution pro- ceSses as has been predicted in Festinger's model. Internal consistency is the use Of all dissonance reduction processes (compliance, distortion, and dissociation) to achieve a reduc— tion of conflict in'a dissonant situation. Cognitively simple subjects, on the other hand, tended to use only one Of the conflict resolutiOn processes rather than some combination as had been predicted earlier. The implication is that one must consider complexity level when attempting to assess individual reactions to dissonant situations. McCarrey, 33, al., came tO the same conclusion in their experiment. As differentiation 60 This increases the strain toward congruity decreases. result may be at Odds with others concerning the attitude change prOpensities Of high and low complexity subjects. This problem will be discussed under number u below. (3) In a 3t icre indirect a subjects are ex ally developed acre alternativ round that t " o‘ (7' J strategy, whic‘: anility to gene; -. the of} ab‘. e to genera: PCSitio . E) than related fll‘dir ‘ l S Opposed t %e 35 (3) In a third behavioral area the evidence is much more indirect and consequently less clear. High complexity subjects are expected tO produce a greater number Of intern- ally develOped solutions to experimental tasks and Offer more alternative interpretations of the situation. Tuckman found that high complexity groups have a more "integrated" strategy, which may be interpreted as a direct link to the ability tO generate a complex solution employing constructs and prOpositions not Obviously contained in the available discreet pieces Of information.61 Driver calls a quite psimilar phenomenon the ability tO develOp "higher level "62 Harvey found that with subjects who are strategies. "taking the other's position" those Of high complexity are able to generate more Opposing arguments (to their own position) than those Of low complexity.63 In a somewhat related finding Terhune and Kennedy report that as integrative complexity increases so does a group's reliance on conceptual, as Opposed to Objective, information.6u All Of these studies seem to indicate an increased ability, on the part Of high complexity individuals, tO formulate more complex solutions to certain kinds Of experimental tasks. The tasks in these studies, the Inter-nation Simulation game, two different stock market games, and a role playing situation in which subjects argued against their own viewpoint, were varied enough tO provide evidence Of the general applicability Of the ability Of high complexity subjects to abstract information. The varied tests Of complexity level used in these studies further this interpretation. Driver used a Multidimensional Scaling 1 ‘ ‘_ A. tecnnigue to re Y. L hcl‘VE’ used t.. (D is oriented tow. PI ! .ernune and Ken Iert and the st is that none of ' 5 .cv complexity he, 0d of attit..,- complexity ind: subject‘s s+a \ «61 Dr I + ESQUPE . Chan‘s uan‘l‘hk 5» ‘¥ . 1 found t0 ‘ t C} 36 technique to measure abstractness, Tuckman used the SCT, and Harvey used the "This I believe. . ." test, the scoring Of which -is oriented toward the content of responses. (See Appendix I) Terhune and Kennedy were cited by Schroder, Driver and Streu— fert and the study is not generally available. The problem is that none Of the studies provide direct Objective measures Of the increased ability to generate alternative solutions from within (i.e., without supplied outside information). Mayo and Crockett provide some negative evidence with their finding that in some situations high complexity subjects dO not gO beyond given information to any greater degree than low complexity subjects.65 (4) The final behavioral manifestation we discussed was the greater uncertainty about positions and increased likeli- hood Of attitude or Opinion flexibility exhibited by high complexity individuals. Scott found that when attacking a subject's stand on the distribution Of nations across des— criptive areas (the Object Sorting Test) high complexity sub; jects exhibited greater flexibility or more willingness to change than low complexity subjects.§6 (See Appendix I) Higgins, using a modified Rep Test, reported that high complexity subjects were less confident Of their judgments in all conditions except where the information was highly incongruent (i.e., where Opposite).67 In an experiment using social pressures, similar to S. E. Asch's classic experiment, to test change in judgment Of the distance between two lights, Janicki found that high complexity subjects were more likely to change their Opinion.68 Janicki used the SCT, as did 37 Streufert, when he found that the attitudes Of concrete (low complexity) subjects were less affected by incongruent information under all tested conditions.69 Stager, using groups varying in their proportion Of high complexity sub- jects from 25% to 100%, found that as the proportion Of high complexity subjects went up so did the amount Of uncertainty within groups.70 Suedfeld and Vernon present evidence par— tially supporting the above results when they find that abstract (high complexity) subjects exhibit greater compliance to experimenter pressure for behavioral change (in a sensory deprivation situation), but less attitude change in a post- test than concrete (low complexity) subjects.71 Lundy and Berkowitz reinforce this finding with their own similar results using a Rep Test instead Of the SCT used by Suedfeld and Vernon.72 The apparent contradiction in these findings is the difference between opinion change (i.e., what is eXpressed tO the experimenter) and attitude change (the underlying position Of the individual). The cognitively complex individual is will- ing to change his Opinions, but not his more fundamental atti- tudes. In the three experiments just mentioned attitude change was measured with paper and pencil tests before and after the experiments. Suedfeld and Vernon measured Opinion change by the degree Of compliance tO experimenter pressure during the experiment. In other words, simple subjects did not cooperate during the experiment, but exhibited greater change on the post experiment questionnaire, while complex subjects behaved in the Opposite manner. The high complexity subject 38 is more likely tO see the subtleties Of the situation and present compromise solutions without modifying underlying attitudes. Attitude changes exhibited by low complexity subjects are no more fundamental depending on the source Of the pressure to change. Tuckman found that homogeneous high complexity groups are more flexible and Open in their organization.73 Sieber and Lanzetta found that high complexity subjects are more likely tO qualify decisions and express uncertainty about them.“L These findings add credibility to the interpretation presented above by reinforcing the general willingness to change in high com- plexity groups. SO do the results reported by Higgins and Fielder who show that high complexity subjects tend to select more moderate probability choices and to have lower belief intensity respectively.75 Some Related Concepts Several more popularly used Social Psychological formulations are related, either theoretically or empirically, to the concept Of cognitive complexity. We shall briefly dis- cuss these ideas and the nature Of the relationship before developing our final set Of testable hypotheses. The most widely known Of these concepts is authoritarianism.76 The authoritarian individual is described in terms similar to our model of the low complexity individual. He is characterized by categoric black-white thinking with limited acceptance Of out- side ideas. The authoritarian concept leaves the structural area, however, and is concerned with the content on the attitudes and values Of the individual. This makes it a considerably different centent, i :Mnn‘3rt ‘Mlyv. it‘ll » b L.’ v Megs"; on .5 cOZ‘l‘el a: 39 idea than cognitive complexity. The difference carries over into the measurement Of authoritarianism where not only the content, but the specific direction of the content are important factors in identifying the authoritarian indiVidual.77 8 Despite pronounced theoretical and measurement differences-some authors have found significant relationships between complexity and authoritarianism. Tuckman, using his own ITI test, found that high authoritarians tended to score low on the complexity test, while low authoritarians scored high on the complexity test.78 Streufert and Driver found a correlation Of -.l8.7g Schroder, Driver and Streufert report that using several different tests Of complexity, various scholars reported correlations Of from -.25 to -.55 with authoritarianism.80 They remark, however, that low complexity scorers are Often among the high and low authoritarianism scorers. The conservative bias Of the authoritarianism scales could help to explain this finding. If authoritarianism was measured so as tO include both liberal and conservative indi- viduals, there might be a considerably stronger relationship. Itigh complexity subjects fall mostly, though not always, into ‘the middle Of the F scale.81 Vannoy's study provides mixed evidence.32 Authoritarianism and the SCT, as scored by ‘Vannoy, shOw no significant correlation (r = .01).83 The same Ilack.of relationship is shown for Scott's test Of complexity .and.two other lesser used tests. However, Bieri's version 111the state or even the local county area, but students w . ' . . ' i -. ' - are readily available in large numbers and inexpenSIve. D- C L”‘E—‘Eerences between students and a more general population we ' it“s: _expected in drug attitudes and intelligence. We felt st L1Gents would be more liberal on the drug question than the 79 population as a whole and that they would be somewhat less likely to maintain a hard "conservative" position in our experimental phase. The former prediction seemed to be true. The mean attitude score Obtained, 96.5, was significantly different from that we might have expected, 90, based on a middle or neutral answer selection on each of the thirty questions (t = 12.7, p < .001, N = 1278). The second difference was an assumed higher intelligence level for students. The difference was probably an advantage in one area because of a probable narrower range of intelli- gence scores. This would decrease the likelihood of intelli- gence being a major factor in experimental results. It could be detrimental in the comparison of intelligence to complexity scores because the reduced variance would be less likely to show a significant relationship. Therefore, we would get a somewhat distorted picture Of the total population relationship. A total Of 1291 students from Michigan State University and Lansing Community College took our battery of tests. Of these, 1278 completed all tests and served as the base Of our experimental pool of subjects. The students were volunteers, solicited by newspaper and poster advertising, and were paid $2.00 for taking the battery of screening tests and an additional $9.00 if they were selected for the experimental Phase Of the research.25 Subjects were selected on the basis of their scores on the: cognitive test and the attitude test. We used students scoring in the extreme upper and lower regions of both tests M 75 for our subjects. Since our major hypothesis (IA and B) explores an effect Of complexity not previously examined, we felt it was more important to establish the possible relationship than to be concerned about the nuances Of how middle range complexity subjects might interact. A wide separation Of attitude positions was also expected to help prevent the inclusion of subjects unsure of their feelings. This procedure would also help overcome any measurement error in both the complexity and attitude tests. Our initial aim was to use one standard deviation above and below the mean as the shut-off point for inclusion in the experiment.‘ The fact that our subjects had to be in the, extreme 1/3 in both complexity and attitude made our task more difficult. For this reason we were not able to adhere strictly to this criterion for selecting subjects. Instead of including 16% at each end of the distribution we have included about 20% for cognitive complexity and 18.5% for attitude.26 Sexually homogeneous groups added further problems, but only among males, where some subjects were marginally outside the above standards. The basic results were as follows: Complexity N = 1278 Mean = 67.09 Standard Deviation = 31.71 High Score = 291 Low Score = 5 Range = 236 For those classified as high complexity the lowest score used was 92. ‘ 76 For those classified as low complexity the highest score was 39. Thus, there was a difference of at least 53 points between any low and any high complexity subject used in the experiment. Attitude N = 1278 Mean = 96.50 Standard Deviation = 18.11 Limits Low = 30 High = 150 High Score = 1M2 (Liberal) Low Score #6 (Conservative) For those classified as "liberal" the lowest score used was 110. For those classified as "conservative" the highest score used was 79. There was a difference of at least 31 points between the most liberal "conservative" and the most conservative "liberal." A total of 35 four person groups were finally filled from the qualified students. Twenty groups were female, five of each group type, and 15 were male, with four repli— . cations in group types I, II, and IV, and three replications of Type III. The total number Of experimental subjects was lu0. Within different types of individuals, high complexity-- "liberal", high complexity--"conservative", low complexity-- "liberal", and low complexity--"conservative", individuals were assigned randomly to groups.27 The Experiment Design Our main intests in this experiment were to isolate the etifects of cognitive complexity and to test for the effects 77 of interaction between individuals of high and low complexity. For this reason we needed two general group types, those homogeneous on complexity and those heterogeneous on com- plexity. The former type includes groups of all high complexity and all low complexity subjects. The latter groups are Of mixed high and low complexity subjects. 1 Given the objectives of the experiment it was also necessary to insure that experimental groups would consist Of individuals differing in attitude toward durgs; thus subjects would find themselves in a conflict situation which might motivate changes in position on the drug issue. More- over, Crockett notes that low attitude and issue salience may have led to the failure of some earlier experiments on cognitive complexity.28 For these reasons the experimental groups also contained equal numbers Of "liberals" and . 9 "conservatives."2 We did not include homogeneous attitude groups, which would completethe design for attitude, because of financial limitations and the lack Of sufficient numbers of qualified subjects. Group size was set at four to allow for balance in attitudes and to lessen possible problems Of individual dominance that might exist in a two-person group. The use Of four subjects would increase the number Of new ideas and the amount of discussion without unduly complicating the experiment. _The inclusion of attitude as a control in all groups PFKDduced the necessity for four group types: two types lunnogeneous on complexity (high and low) and two types hetrerogeneous on complexity (those with high complexity 78 liberals and low complexity conservatives and those with low complexity leberals and high complexity conservatives). Because there were two subjects homogeneous on both complexity and attitude within each group and because these individuals were assigned to the group conditions randomly, it became apprOpriate to consider the dyads, homogeneous pairs in each group, as fundamental units of analysis. When considering group activity and final group solutions to the experimental task this will be our procedure. There are four dyad types: 1. High complexity "liberal" (HL) 2. High complexity "conservative" (HC) 3. Low complexity "liberal" (LL) H. Low complexity "conservative" (LC) In a factoral design this means there are 16 possible group types. However, some of these groups are duplications, such as (LC, LL) and (LL, LC), and need not be included in the design twice. Others do not meet the criterion Of Opposing attitudes that we have included in our design; therefore, group types like HL, HL and LC, LC will not be included. This leaves four basic group types to be used in our test: Type I - HL, HC Type II — HL, LC 'Type III - LL, HC Type IV - LL, LC Finally, because of possible dominance—submission roles P133Wed.by men and women respectively in group situations, 79 groups will be homogeneous on sex. This creates the third independent variable, group type (or level of complexity) and attitude being the first two. The final configuration is therefore a n X 2 X 2 nonfactorial design. We had intended to use five replications Of each group type, but the lack of high and low complexity conservative males prohibited this. Instead we have four replications of group Types I, II and IV and three Of group Type III. We were able to fill all female groups. This means that in addition to being incomplete our design has unequal cell frequencies. We shall take up these problems again in the analysis section of this chapter. Procedures The experiment took approximately one hour and forty five minutes and consisted of three phases. Phase I: Subjects were instructed to provide individually originated solutions to what they saw as the "drug problem."30 They were given paper and pens and allowed about 30 minutes to complete this part of the experiment. Phase II: Upon completion Of their individual recommendations, the subjects were taken, as a group, to a discussion room where they were given a second set of instructions.31 These instructions asked them to discuss the rurture of the drug problem and to make a set of recommenda— ticuis concerning its solution. The important aspect Of this Profizess was that they should all agree on the final recommen- datcions. A "pressure" toward compromise was thereby created. 80 They were also told that there would probably be disagreement since their attitudes on the drug issue were different, but they were not informed who held which attitudes.32 If irreconcilable differences did occur individuals were allowed to dissent, formally, from the final group recommendations or Isubmit their own "minority" recommendations. They kept their own individual solutions with them to use as references. Finally, they were told they had one hour to agree on the final recommendations. The setting was a large carpeted seminar—type room in which the subjects sat in a configuration around a table. At the end of the room was a one-way mirror and a second room from which the experimenter and coders Observed the discussion. The subjects knew about the experimenter's Observation but were not told Of the coders unless they asked. Nobody did. A "secretary" was also in the room with the subjects. Her job was to record the recommendations (and dissents) made by the group. She wrote nothing until told to do so by the group members, and then only what they dictated to her. She took no part in the discussion and was seated inconspicuously away'from the subjects. This procedure was used to speed up the process Of recording the recommendations and to make sure that all group members were fully aware of what was being said in recommendations to which they were agreeing. It avoided the problem of editing by a "volutary" secretary chosen from the group. There was no noticeable reaction to the secretary from 81 any group with the exception Of a few of the male subjects who could see her without noticeably turning from the rest of the group. The subjects were also informed that the group discussion would be video taped. There were microphones on the table and two cameras, one in the room with them and one behind the one-way mirror. They were given assurances Of their anonymity and that the video tapes would only be seen by the experimenter and his coders, and then would be destroyed. It is difficultv to assess how much the cameras inhibited the discussion. Certainly some subjects acted as if they were very aware Of its presence, by frequently glancing at it, but there was no way of knowing if it affected their participation in any way. Phase III: After completing their discussion and making recommendations subjects were given a post-experiment question- naire dealing with two general questions; their impressions Of the experiment and its purposes and their evaluation of 33 We are concerned with only the latter other participants. questions at this time. Our instructions to the subjects included a statement that we were interested in how groups I wOrk together "to produce a common set of recommendations" when they have "different opinions" on the subject. When asked What they thought the experimenter was trying to accomplish, 39 Of 1M0 (27.9%) repeated essentially that description. Another 16 (11.9%) added an interest in the substantive drug recommendations as a central aim Of the experiment. Fifty (35.7%) thought the experiment concerned only the content Of 82 the group recommendations; i.e., drugs. None Of the subjects realized that cognitive complexity was Of interest, although One suggested that one of the screening tests taken several weeks earlier was the basis for selection to various groups. Of the remaining 35 subjects (25%), 13 saw compromise in a conflict situation as the primary focus Of the study. Most Of these said they realized this during the experiment itself. This cOuld have had an influence on how they behaved, making them more or less likely to compromise. Three Of these indivi- duals were in the same group. The group, however, was not noticably different in behavior from other groups Of that type. It seems safe to conclude, on the basisof the small number of individuals who were more aware of what was going on and the fact that none recognized the principal independent variable, that the results of the experiment were not compromised. Post-Experiment Measurement In this section we are concerned with the measurement of change in individual positions on solutions to the drug problem as manifested in positions expressed in individual and group recommendations. Our task is to determine who changes the most as a result of the group discussions. For this purpose we shall use a measure that describes the prOportion of "wins"- achieved by each dyad (with the same attitude position) in the gPOup.35 A "win" is scored if either: (A) a member of a dyad secures acceptance of a recommendation that does not occur in the original recommendations of either Of his two Opponents or (B) if a dyad succeeds in keeping out of the final set of 83 recommendations a point suggested by one or both of the Opponents but not supported by the dyad. We have predicted that low complexity subjects should have a higher prOportion of wins; i.e., they should dominate the final recommendations of mixed groups. The relationship between recommendations Of Opposing dyads can take one of seven forms. These relatiOnships are summarized in Figure III. A. PrOposals which both dyads make which are included in the group recommendations. B. Proposals which neither dyad suggests, but which are included in the group recommendations. C. Proposals which both dyads make, but which are not included in the group recommendations. D. PrOposals suggested by the "liberal" dyad, but not by the "conservative" dyad which appear in the group recommendations. E.‘ Proposals suggested by the "conservative" dyad, which are not included in the group recommendations. F. Proposals suggested by the "conservative" dyad, but not by the "liberal" dyad which are included in the group recommendations. G. 'PrOposals suggested by the "conservative" dyad, but not by the "liberal" dyad which are not included in i the group recommendations. 814 Figure III Possible Combinations of Dyad Proposals and Group Recommendations Proposal Relationships Proposals by A B C D E F G Liberal Dyad X x x X Conservative Dyad X X x x L Final Group Recommendations X X X X A = Neither side wins B = Neither side wins C = Neither side wins D = Liberals win E = Conservatives win F = Conservatives win G = Liberals win Types A—eC are of no interest to us because there is either agreement or we are not able to determine if any disagreement existed, as in B. It is unclear exactly how situation C might arise. One factor might be time, another relevance of the proposals. Situations D--G are of interest because they constitute instances where differences exist. The resolution Of these differences indicates a compromise on the part Of one Of the dyads; i.e., they are or are not included in the group's final recommendations despite initial opposition by one Of the dyads. 85 To calculate the prOportion of wins for either dyad (we shall use the "liberal" dyad) one need only compute the total number of wins, in this case D + G and divide by the total number of proposals in conflict, Dr+ E + F + G. The score for the "conservative" dyad is complimentary, substituting 36 E + F in the numberator. Thus: . . . D+G PrOportion of "liberal" wins = D + E + F + G It is this prOportion that will be used as the indicant of domination Of a group discussion in our final statistical analysis Of the effects Of complexity. A score close to .5 indicates that neither dyad dominated the discussion by controlling recommendatiOns. A score significantly different from .5 would indicate that one or the other dyad did dominate the substance of the group proposals. Measurement of Group Processes Three independent coders were used to measure behavior taking place during the experiment. These coders Observed and scored the discussions from behind one-way glass. Coders were trained for approximately 39 hOurs by the use Of video tapes and live practice groups before beginning actual coding. Each coder scored all four subjects in six different areas of 37 Final scoring of these behavior during the experiment. variables was an average of the Observations of all coders. Reliability will be reported below. 86 The six variables coded were: 1. New Aspects - We were interested in the number of original aspects of drugs and the drug issue brought into the discussion by each subject. Essentially this variable could serve as a validity check on our measure Of complexity. Those Of high complexity should introduce more new aspects into the discussion. A new aspect was considered any new suggested solution to the drug problem or the introduction Of a different consideration into the discussion. The former is straight forward. Examples of the latter might include the suggestion that drugs are a social problem, suggestions Of the limitations of other proposals, introduction of a new characteristic of a drug that affected other proposals, etc. A code was devised to help coders score the iscussion on this variable by its subject matter. This Code helped coders keep track Of what had been said and also allowed for a more careful reliability check since we were able to determine if the new aspects scored were in the same specific subjggt area. Reliability was .89 on this variable. Information Seeking - The coder scored a subject each time he requested information about drugs or some drug related area or activity. It did not include questions asked to get clarification of someone else's discussion. Difficulty in coding arose around just where this breaking point, between information and clarification, was, and the intercoder reliability was a very marginal .65. COOperation - COOperation is defined as any behavior that Offers support for the position of another subject. It includes such verbal manifestations as, "I agree with him," "I think he's right," or elabora— tion and explanation of another subject's argument. The difficulties in scoring this behavior are reflected in an unacceptable reliability of .99. Because of this low reliability further discussion Of COOperation will be drOpped. Conflict - Conflict is defined as behavior which shows disagreement among subjects. It may be expressed as simply as "I disagree," "you're wrong," or a "no" when asked if the subject agrees with what another subject has said. It may also involve a judgment by the coder on whether the response offered by a subject is in fact in disagreement with what someone else has said. Reliability on this variable was a passable .75. The reliability check on this variable includes a check of with whom the subject is in dis— agreement, so that entries must agree on number and adversary. ‘ 87 5. Value Arguments - These are expressions of opinion in a discussion that are unsupported by facts or logic. In coding this variable coders were instructed to include only the most blatant examples. Following this procedure the number of scored value arguments was reduced, but the reliability was an acceptable .78. 6. Logical Arguments - Arguments that are well reasoned and factually supported are considered logical. Again there was a training (reliability) problem for this variable. The problem was the identifica- tion of a "factually supported" argument. Since both coders and subjects had different levels of infor- mation about drugs, the facts behind an argument were often in dispute. Even by telling coders to consider an argument logical if the subject thought he was using factual support for his position did not clear up the difficulty. As a result we Obtained only a highly marginal reliability Of .60. In addition, coders tended to shy away from this foggy area and few logical arguments were coded, although part of the reason for this was the type of discussion that prevailed in may Of the groups. Data Analysis We are concerned with two areas Of analysis. First, we need to determine to what extent external factors, such as sex, intelligence, information and attitude, are related to our subjects' complexity level. Second, we want to analyze experimental results. The former refers to questions of discriminant validity and the latter to construct validity. The first problem was solved by correlating the level Of complexity with all of the possibly related variables for which we had data. This was done for both the total sample (N = 1278) and for those who were selected to participate in the experiment (N = 1H0). Nominal data were cross-tabulated and nonparametric measures Of association were used. 88 Experimental results were analyzed on three levels: first, the differences between group types were analyzed in two-way analysis Of variance models.1+0 Second, gross differ— ences between groups were analyzed a one-way analysis of variance model. Third, the behavior Of high and low complexity individuals was analyzed by means Of a correlation analysis. TO test for significant differences in the proportion of "wins" for high and low complexity dyads within group types we used a two-way analysis of variance design. The two inde— pendent variables are group type and sex (the model is n X 2). The theoretically interesting third "independent" variable, that of attitude, does not appear in the analysis, first because attitude composition is held constant in this incom- plete design, and second (had the design been complete) because of the scoring procedure in which Observations Of "liberals" and "conservatives" are not independent; i.e., the measure for "conservative wins" is simply the compliment of that for "liberal wins." For this reason we could not use a three-way model to analyze main and interaction effects for attitude. Interaction between attitude and complexity was determined by the use Of Scheffe'pp§p_ppp comparisons. With this prO- cedure we were able to Obtain much of the same information in the two-way model that we would have gotten in a three-way model. ’At the same time we were able to adOpt a more useful method fOr measuring group dominance than would have been available for a three-way design. Each of the dependent variables tested was based on the Proportion of that variable attributable to the "liberals." 89 In addition to recommendation dominance, we tested for the amount of discussion, the number of new ideas introduced, information seeking, and conflict. Cooperation was dropped from the analysis because Of low intercoder reliability. Logical and value arguments were dropped because of an insufficient number of observations. The procedure followed allowed us to test the hypotheses about behavior Of low and high complexity subjects in mixed and homogeneous group types. It was this comparison Of dyad scores which produced what we have called group type differences. Other group differences were tested in. a one-way design in which the sex variable was omitted. In these cases we were interested in gross difference between groups where scores were not dependent on the dyad differences. The hypOthesis about the increasing complexity of outcomes with an increasing PrOportion Of high complexity subjects was one of these. The final level of analysis, individual differences,was IIleasured by comparing group behavior Of high and low complexity Subjects regardless of group association. For this analysis Simple correlational procedures were used. It should be emphasized that measures at the individual and general group levels among the group process variables are not independent. The measures for the amount Of conflict, COOperation, the introduction of new ideas, etc., exhibited by groups are simply the sums on each variable for all individuals‘in the group. This is not true for the relationship betWeen output and process variables. In this case output 90 variables are measured completely independently of process variables at any level. The purpose in measuring process variables at different levels was to test for the effects Of different group organizations, i.e., prOportion of high and Llow complexity subjects, on individuals differing in complexity Ilevel. For example, is the relationship between complexity ZLevel and conflict behavior for individuals the same in liomogeneous and heterogeneous groups? Footnotes lOur statistical analysis, to be described in detail 1>ealow, is a two-way model because differences in attitude £1176 not independently measured in this design. We have aa1ttempted to overcome this problem by using post—hoc tech- riquues in the two—way analysis of variance design. 2Robert Zajonc, "The Process of Cognitive Tuning," Journal of Abnormal and Social Psychology, LXI, 2 (1960), pp . 159-167. Robert Zajonc, Cognitive Structure and Cognitive Tunin , (unpublished Ph.D. Dissertation, The University of filchigan, 1959). ~ 3Zajonc, "The Process of Cognitive Tuning," Ibid., p . 160. l4See Appendix A for a copy of the exact format of the text. Inclusion of a particular attribute into a specific group is left to the individual as is the selection of the attributes themselves. Scott finds this procedure a drawback because Of the possibility an individual will use a number of terms with essentially the same meaning and thereby artifici- ally raise his complexity score. Using the differentiation Part of the test alone this is probably true, but the second Part of the measure will tend to wash out this effect by ‘ f0r‘cing the individual to divide the attributes into groups. William A. Scott, "Cognitive Complexity and Cognitive Flexibility," Sociometry, xxv, n (1962), pp. 905-919. 5Zajonc, lOc. cit. 6Ibid. 71bid. 8This does not mean that they could not have thought Of aSiditional attributes if allowed to ponder throughout the Elght, but that the initial outpouring of descriptive phrases ‘vafi fairly well stOpped by that time and subsequent attributes tofilldfiave been labored. This may have hindered the slow tfilnking cognitively complex individual, but it is not felt art it seriously impared our results. In addition, the second Part of the test was probably more important in determining the lr1£~11 complexity score. 91 92 9Zajonc, loc. cit. 10William A. Scott, "Structure of Natural Cognitions," Journal Of Personality and Social Psychology, XII, n (1969), pp. 261-278. 11Harold M. Schroder, Michael J. Driver, and Siefried Streufert, Human Information Processing, (New York: Holt, Rinehart and Winston, Inc., 1967). 12Joseph Vannoy, "Generality of Cognitive Complexity- Simplicity as a Personality Construct," Journal Of Person— .ality and Social Psychology, II, 3 (1965), pp. 385-396. 13The validity question I wish to raise in the use Of the ssentence completion test and Tuckman's modification (ITT test) (:f it, is as follows: In scoring these tests the assumed laehavioral characteristics of different complexity types are tised to indicate the presence or absence Of that type. This jLs particularly true for Types II and III where the theoretical . jformulation and behavioral implications are also weakest. -VJhat is needed is an independent means to determine the level <>f complexity (of complexity Type). Except at the extreme levels this differentiating capability is not present in the 53C test or the ITT test. It is also somewhat lacking in our <>wn measure, if you accept the notion that Type II and III individuals are qualitatively different (not just more or less) from Types I and IV. See Schroder, Driver and Streufert, loc. cit. Bruce W. Tuckman, "Personality Structure, Group Composi- tion, and Group Functioning," Sociometry, XXVII, 1+ (196“, Pp. I#694487. , . 1“Vannoy,.gp, cit. A study of this type was conducted in Ccnnjunction with the present research, but the data have yet t<> be fully analyzed and unfortunately no results are available 15We tried, as much as possible, to adhere to a minimum of? five subjects per item as a rule of thumb. This is not exactly a "conservative"‘rule but it was as much as could be Obtained at the time. See , . Jum C. Nunnally, Psychometric Theory, (New York: 51:11 Book Co., 1967), for a discussion of this problem. McGraw- 16For a discussion Of this approach to scale development See Paul Horst, Personality: Measurement of Dimensions, IgS‘an Francisco: Josey-Bass,Tnc., 1968) and Anne lnastasi, CI'STLch‘Ol'O‘g‘ical Testing, (New York: The Macmillan CO. , 1968), apter 13. 93 17We realized the danger of misclassifying some subjects because Of heavy loading on one or two Of the factors, but this chance had to be taken in order to include some items as those concerned with legalization, which did not load highly on the first factor. 18See Appendix B for final Attitude Test. See Appendix H for the rotated factor matrix of the :final pre-test from which the 30 item test was selected. 19See Appendix C. 20John H. McNeill, "The Pharmacology Of Drugs Of Abuse," I’repared for The Governor's Office of Drug Abuse, State Of riichigan, no date. - - Additional information was obtained from DARTE Resource Guide put out by Drug Abuse Reduction Through Education, ‘Vlayne County Intermediate School District, Detroit, Michigan, June, 1971. ‘ 211 am indebted to Dr. A. E. Joula, Of the MSU Office of iEPvaluation Services, for giving me access to this information. 122Relevant Pearson Product Moment Correlation Coefficients are: MSU reading - Total Score on Aptitude Test (SAT, CQT or IK'CIZT).= .71. MSU reading - Verbal portion Of Aptitude Test = . 3. 23For a discussion of the validity and reliability Of tflie SAT test (which was the test taken by more than 60% Of our sample) see William H. Angoff, ed., The College Board Admissions Testin Pro ram, (New York: College Entrance Examination Board, IWI) . 2”In our later discussion of results the comparisons will be made with the MSU Reading Test, the total score of which- eVer aptitude test was taken (if any) and the verbal portion of the test. . 25Funds for paying subjects were provided from a United States Air Force Office of Scientific Research Grant, . #F"fu620-69-C-011H, through the COOperation Of Professoerames 1}lllips, Computer Institute for Social Science Research, 1Chigan State University. 26Procedures for calculating percentages taken from Wilfrid ‘J- iDixon and Frank J. Massey, Introduction to Statistical W (New York: McGraw-Hill Book CO.), p. 53. 9” 27One minor problem was scheduling meeting times. Several individuals were shifted because they could not meet at a time when the other members of the group were available. It is unlikely, however, that this problem introduced any systematic bias into the experiment. 28Walter H. Crockett, "Cognitive Complexity and Impression iFormation," in B.A. Maher, ed., Progress in Experimental IPersonality Research, (New York: Academic Press, 19657: pp. 57-90. 29We used an equal number to avoid situations where the rnajority attitude group could simply outvote the minority vvithout making compromises on the issue. Future research must ailso test for the effects Of complexity in this imbalanced ssituation. 30 See Appendix D for instructions given. 31See Appendix E for instructiOns given. 32If asked the purpose of the experiment at the beginning c>f this Phase we referred to the instructions concerning the Clifferent attitudes about drugs and stressed the importance c>f the substantive recommendations to be made by the group atnd their possible use of individuals making real drug policy decisions. Whether or not they accepted the explanation is Ixroblematic, but it is clear from post-experiment question- ruaires that none of the 1ND subjects saw the real purpose Of tine experiment; i.e., the relation of group decisions to COgnitive complexity. Beyond that we were not really trying ‘tc> hide any aspects of the experiment. This question will be discussed in greater detail below. 33See Appendix F. 31*It was found, through trial and error, that this was as ‘Veuried a coding task as each could comfortably handle and still nliiintain an acceptable degree of reliability.. The fact that Several of these behaviors, as used here, did not occur very c>ften made possible the scoring Of what may seem to be a rather large number of variables. 35Again we assume that individuals with the same attitude IKDsition randomly assigned should reflect the same positions. We therefore assume the following in combining the recommenda- tions of these individuals into dyad scores: 1. That recommendations made by one member and only losely referred to by the other are the same or that they are in agreement. 2. If a recommendation is made by one member and there is nothing in the other' 3 recommendations to indicate disagreement, they agree. If there is an indication 95 Of disagreement on some recommendation it is treated as if the dyad were in agreement with its Opponent dyad and thereby not scored. This happened only three times in 70 dyads. 36The use Of procedure means that "liberal" and "conserva- tive" dyad scores are dependent. This will have an effect on statistical procedures used to evaluate the data. Our ostensibly 3—w design will be treated as a 2-w model in the statistical analysis to be described below. t~ 37Our procedure for determining intercoder reliability J was as follows: Each discussion was divided into 15 minute time periods. The numbers correlated were the total number Of behaviors, Of each type, Observed druing that time period for each subject. There were, therefore, 16 Observation periods per coder per session for each variable. The multiple R was calculated for all subjects in all.groups.. The limitations of this procedure should be clearly understood. We have not way Of determining if each item scored by the coders in a 15 minute period is the same one scored by another coder. We can only assume that if they have an equal number that they are recording approximately the same eVents. Breaking the sessions into smaller time blocks would have caused greater confusion and would have served no really useful prupose. PrOper training of coders on the meaning of each coded variable provides credibility for the procedure. In addition, no other method seemed more efficient and reliable. 38Appendix G contains the Code Book used to code individual behavior. 39This will be discussed in greater detail in the next Chapter. no . . The analySIS Of variance procedures and subsequent Scheffe method contrasts were drawn from William C. Guenther, Analysis of Variance (Englewood Cliffs, New Jersey: PrenticeéHall, Inc., 196%). CHAPTER Iv Findings and Discussions The results of our research provide supporting evidence for our hypotheses about the effect of complexity level in ' group outputs in heterogeneous and homogeneous group types. Essentially, we have found that low complexity dyads do have a bargaining advantage in that they tend to "win" a greater proportion Of the disputed recommendations in heterogeneous groups, while in homogeneous groups the split is not signifi— cantly different from the overall mean. Our measure of dis- criminant validity, also, supported the independence Of our measure of complexity. Hypotheses about group process, drawn from our conceptualization of complexity and previous research on complexity were not as clearly supported. Discriminant Validity and SampeDescription Campbell and Fiske discuss two procedures for determining the meaning of a psychological measurement instrument.2 The first is convergent validity, which is established by comparing the results of test presumed to measure similar concepts. Vannoy's comparison Of several different measures of cognitive complexity is an example Of an attempt to establish convergent Validity.3 Our initial screening test included several measures 96 97 of cognitive complexity other than our own, but unfortunately these data have not been analyzed as Of this writing.|+ Discriminant validity is established by showing that the concept under consideration is not empirically related to other concepts. It "discriminates" between this particular concept and others which might be seen as equivalent. In the case of cognitive complexity we discussed related concepts in Chapter II; intelligence and informatiOn or knowledge about a particular subject. Our screening tests and several aptitude tests provide the data for testing for the discriminatory power of our complexity test. Table II (p. 99) shows the-results for our sample of 1268 screening test subjects. Using the aptitude test results as equivalent to I.Q., we found that cognitive complexity is not significantly related to total I.Q. (Pearson's r = .0u6). In this case the three different tests used (ACT, CQT and SAT) 5 The verbal portion Of the I.Q. tests are not separated. were significantly related (Pearson's r = .072), as was the MSU Reading Test (Pearson's r = .069). The magnitude of these relationships, and thus the prOportion of explained variance, was so small that it does not seem to be'a factor likely to contaminate the experimental results. Tables III and IV (pp.100 8 101) shOw»slight.differences between the sexes. The relationship between I.Q. and com- plexity is somewhat higher for males than females. For males it is significant (p < .05) for all three measures used and IKIt significant for females. Again, however, the magnitude 98 of the correlation is very small (all less than .1) and would not seem to be a major factor in defining cognitive complexity as we have measured it. A slightly higher correlation exists between I.Q. and level of integration.6 In all cases, except total I.Q. for females, the relationship is significant at the .05 level, although it never exceeds r = .18. There is a somewhat closer tie between the ability to integrate and I.Q. than between the ability to differentiate and I.Q. This relation- ship might be expected among college students, since the I.Q. tests used are essentially college aptitude tests which are relatively good predictors of success in college where we expect some facility at integrating large amounts of infor- mation. It will be noticed, in fact, that differentiation correlates in a negative direction with our various I.Q. measures, although still at very low levels. The low range of these relationships hardly detract from the independence of the cognitive complexity measure, at least with regard to I.Q. Table V demonstrates the lack Of impact on experi- mental results with the finding that among the 190 experi- mental subjects, cognitive complexity and I.Q. (the MSU 'Reading Test) correlate at r = .017, which is considerably below the level needed for statistical significance. The second area Of possible overlap with cognitive complexity is the amount of information and individual has about a particular subject area. This is especially true for an issue specific measure such as the one used in this study. 99 Nmo. A COHPMHOQQOO cm£3 mo. «mama u z mom. mso. mmo. moo. mmo.| mso. :ma. sma. med. :mo.| mHH. oms. can. Hmo.| mmH. mmm. oso. HmH. mmH. :NH. mmo. Ame Ame Asv Ame Ame sz Ame Amendmedm HHv .o.H “Heroes .o.H xmm soapmpmmvcH COHpMfiHCOQOMMfiQ >vwxmameoo m>flpwcmoo meanmw9m> pmme mcficmmsom emnpo can wixmameoo O>wacmoo mom stvmz cowpwaoghoo HH mamv .o.H sea. meo.u mmo. fies “envoys .o.H meo.u new. Ame doeumnmmwcH 0mm. Ame :OfipmwpcmanMfio Adv zvflxmaano O>fivwcmoo Ame Ase Ame Ame are Ame Ame AHV Ammawxv , , meAMHQm> pmma mcficomsom smnpo can szxmaeeou m>fiuwcwou sow xwpvmz :Ofipmampsoo HHH mam<fi 101 mo. A cowvmamseoo ewes mo. v me «How u z meQQ mam. Hoe. Nee. meo.u wee. flea. Nee. Ame unmade endpape< ema.. ems. wee. wee. ewe. Hoe. Ase doeemenonH mean mew. mew. mad. meo.u owe. Aev.pnma mdeemmm 2m: ems. ewe. oao.n ems. Ame Aflnnne>o .o.H mes. mmo.- mas. “so “Heroes .o.Hl HHe.- see. Ame doapnnmepeH Has. Ame noapmepceanMea Adv szxmaasou m>flpwcmoo Ame Ase Ame Ame are Ame Ame Adv AmmHmEmmv . mOHnmw9m> pme mcwammsom swnpo cam hpwxmaeaoo m>wvwcmou pom xflapmz cowwwamspoo >H mgm<fi 102 For this reason a test was constructed to measure the amount of knowledge subjects had about drugs and administered to all persons who took the complexity test. Table II shows a low, but significant, positive correlation between level Of com- plexity and the amount of drug information (r = .129,.p < .05). Drug knowledge is also positively related to both differentia- tion and integration. A similar pattern exists for males and females considered individually (Tables III and IV). Given the nature of our complexity test it is not surprising that some correlation with knowledge exists. The ability to differentiate characteristics of an Object would be expected to show at least a slight relationship with the amount Of information a person has about the subject. The important point is that this relationship is small, accounting for less than 2.5% Of the variance in the highest correlatiOn (inte— gration and information for males, r = .199). We interpret these figures to indicate the virtual independence of cognitive complexity and information, at least for the subjects tested. A similar relationship characterizes the 190 experimental subjects (Table V). We may conclude, then, that our measure of cognitive complexity is not assessing the subjects' intelligence or the amount Of information he has about a particular subject. This discriminating power helps to establish the validity of the test used. We have also tested the attitude of the subjects toward drugs. Generally, there is no reason to expect a relationship 103 TABLE V Correlation Matrix for Cognitive Complexity and Other Screening Test Variables (Experimental Subjects) (1) (2) (3) (N) (5) Cognitive Complexity (1) sex (2) .079 MSU Reading Test (3) .107 -.102 Drug Information (H) .162 .015 .131 Attitude Toward (5) .083 —.010 .068 .328 Drugs N = 190* *p < .05 when correlation > .170 between complexity level and attitude, but in a given population such as relationship could exist. Essentially, we are deal— ing with the same kind of problem discussed by Rokeach and others with regard to authoritarianism.8 The measure of cognitive complexity should be free Of attitudinal overtones. We have tested a college student pOpulation, however, where independence does not seem to hold. The correlation between cognitive complexity and attitude is r = .097 (p < .05). (Table II)._ It is also significant for males and females separately (Tables III and IV), but it is again very weak. In addition, the possibility of a significant effect Of attitude on decisions of the experimental groups led tO the inclusion of attitude as an independent variable. For the 1H0 experimental subjects the correlation between attitude and complexity is not significant. There is no bias in the 10H experiment and any interaction between group type and attitude will show up in our post hoc comparison Of the analysis Of variance outcomes. The relationship between sex and cOmplexity is very low Women had a mean of 69.u0 = 30.75) (r = .061), but significant. (SD = 32.25), while men had a mean of 65.5 (SD (Table VI). Men, however, knew more about drugs and were more "liberal" on the drug issue than women. Again all of these differences were relatively small and do not seem to indicate that important differences should be expected between Then and women in the experimental results. Data for the 140 experimental subjects are very similar to the overall data (Table VI, p.105). The MSU Reading Test aand the Information test results are very close to the overall Ineamn The somewhat lower mean for attitude reflects the liarger number of women in the experiment (80 women in 20 groups The slightly higher mean for complexity to 60 men in 15 groups). In if; also indicative of the fact that there are more women. alty case, differences on I.Q. and information should be randomized throughout the groups and have no systematic effects on experi- im91Ttal results. The lack of truly low complexity scores could ibiias the results toward greater willingness to change, etc., :th‘the hypotheses are correct. 105 03H mm.mm mm.:m m:.m nom.mH oe.mm ow.ea m~.ma mo.ms .a.m and: mpommosm HmchEflamexm How Nw.mH mm.:m :w.: :m.:H Hm.am Ho.m: He.:m mm.m: mm.mm mh.mm mm. mm.H ms.~H N:.sm mm.mm o:.mw .Q.m com: mmadfimh see Na.sH aa.em s~.m em.mH mm.sm ~3.Hm He.em ~m.ma oe.sm ~o.em as. He.a as.NH Hm.em ms.om mm.me .c.m and: mean: ho.mH mo.m mH.mm mb.mm H~.mm Hm. mb.~H Hm.Hm .QCm mead um.mm mo.mH om.m: mH.:: mw.:: :m.H :m.mm mm.bm cmmz GHQEOm HMHOH mmpoom wmwa mcwcmmpom pom mcowvmw>ma Uhmocmum can mcmoz H> mam<fi z OO59HHH< mama. cow» IMEQOMGH mcwpmmm sz asaas> .o.H Heads .o.H cowvmanpaH GOMHMMH (consumed hpwxmaeaoo m>wuwcwoo 106 Experimental Results: Group Recommendations Group Type Differences in the Proportion Of "Wins" Hypotheses IA and IB predict that low complexity subjects. should dominate the group recommendations in heterogeneous groups and the dyads in groups homogeneous on complexity level.' The analysis shows that there is a direct effect Of group composition (GC), according to the cognitive complexity of group members, upon group decisions (p s .090) (Table VII).9 The effects Of Sex and interaction of sex by GC were not significant. Specific questions about the effects of GC will be elucidated below by means of post hoc comparisons of means. TABLE VII Analysis of Variance for PrOportion of Group Recommendations Won by Liberals Sum Degrees Mean Signifi- Of of Sum of 'cance Effect , , Squares Freedom Squares F Level Group Composition (Complexity) (A) .1680 3 .0560 3.0926 .099 Sex (B) . .0021 l .0021 .1183 .739 (A X B) .0206 3 .0069 .3791 .769 Error .u888 27 .0181 Total .6706 Cell means are presented in Table VIII. A Scheffe post hoc comparison of the two heterogeneous groups (XII - XIII) shOws that the difference between them is significant ).10 (p < .05 This difference is in the predicted direction with low complexity subjects "winning" a significantly greater 107 proportion of the disputed recommendations in each case; i.e., regardless of attitude. Using the same comparison technique there was no significant difference found between groups homogeneous On complexity (i.e., XI - Riv). Another set of comparisons was made to test for significant differences between heterogeneous and homogeneous group types (group types II and III, and I and IV respectively). Groups composed Of heterogeneous dyads, one high complexity and low complexity should show greater differences on proportion of "wins" between dyads than groups homogeneous on complexity. In this case either Of two null hypotheses could be used: ‘P = (xI - XII) + (xIII - XIV) = 0 (1) ‘I‘ (2) II C) (XIII — XI) + (XIV - XII) Essentially we are testing if the difference in proportion 'Of wins for liberals in homogeneous dyads is different from that in heterogeneous dyads. By using both comparisons simultane- ously, we double the number of groups and increase the power of the test. In both null hypotheses we are saying that the sum of the differences between homogeneous and heterogeneous groups should be equal to 0 if the predicted differences do not equal 0. Null hypothesis (1) tests whether or not the difference between the proportion Of "wins" for "liberals" in group type I and group type II (predicted to be negative) plus the difference between group types III and IV (also predicted to be negative) will equal zero. If the differences are cloase to zero or if one is in the wrong direction the null hypothesis will be sup- ported and our prediction will not be verified. Null hypothesis 108 (2) tests for the same relationship, but from the point Of view of the "conservatives." In both cases the differences were significant (p < .01), thus providing evidence that the change from homogeneous to heterogeneous groups brings about important differences in group recommendations.ll These differences involve a change from similar to equal division of "wins and losses" among homogeneous groups to unequal dis- tributions of "wins and losses" among groups with subjects differing in their level of cognitive complexity. Before discussing these results it is necessary to account for the possible effects of attitude. Attitude is not a major Hinterest to us for its independent effects, but it is important insofar as it interacts with group composition. A significant interaction between group composition and attitude would call into question our finding on the effects of group composition on decisions. Since financial stringencies forced us to adopt an incomplete factorial design, we have no Way of testing' for the effects Of within group distributions of attitudes toward drugs upon group decisions. It is equally impossible to draw any conclusions concerning interaction effects of attitude and complexity distributions by using the regular analysis of variance techniques. The previously discussed finding that, for individuals, attitude toward drugs has a very Weak relationship to cognitive complexity and to differentiation (Tables II, III and IV) does not answer the question about the separate or combined effects of attitudes and complexity, or Of within group distributions Of theSe variables, upon group outcomes. Nor does a similar weak relationship (r = .13, p > .05) 109 between complexity level and individual recommendations for solution of the drug problem help to elucidate the issue. TABLE VIII Mean Proportion Of Recommendations Won by Liberals Sex 'Male Female Column Mean Number of Groups Mean Number Of Disagreements** *Group Types are: HH* .560 .539 .550 29. HH - HL - LH - LL - GroupyComposition (Type) HL* ‘ LH* LL* Row II III IV Mean .909 .669 .526 .536 .990 .631 .563 .552 Overall Mean .997 .650 .595 .598 9 9 9 1 19.3 19.5 15.1 All high complexity subjects 2 liberals and 2 conservatives 2 high complexity liberals and 2 low complexity conservatives 2 low complexity liberals and 2 high complexity conservatives All low complexity subjects 2 liberals and 2 conservatives **The situation that domination is mathematically more likely in groups with a smaller number of issues in conflict is not reflected in the results. Group type IV shows least tendency toward extreme dominance by one or other dyad, while having the smallest mean number of conflicts. Group type I .is somewhat higher, but the greatest dominance exists in the middle range groups, Types II and III. 110 However, a special application Of the Scheffe method of pp§£_ppg contrasts makes it possible to test a hypothesis of interaction between attitude distribution and complexity upon 12 If there is interaction the effects of group decisions. cognitive complexity on "liberals" should not equal the effects of complexity on "conservatives." The null hypothesis used to test for such a relationship may be stated in two general ways: II C) (3) Hoc‘ (”I -lltII) " ”(III ’AIV) (9) Hoe‘ (AI -/[III) S H H I H < v I D We shall work with the first equation only, since either will provide the apprOpriate test. The A 's in these hypotheses refer to the column means from Table VIII. The Roman numeral subscripts refer to our four group types. Essentially, null hypothesis (8) states that the effects on the difference in the proportion of "wins" for "conserva- tives" should be the same in situations where "Conservatives" are high and low complexity regardless of whether Or not their attitude Opposite dyads are high or low complexity. Thus, when "conservative" dyads are high complexity with high complexity partners/II and low complexity with high complexity partners I ’Atll’ the difference between high and low complexity "conservatives" with low complexity partners (jkIII - /‘(IV). Null hypothesis (9) presents the same contrasts for "liberals." With the help of Figure IV, below, we shall explain how these particular contrasts provide answers to our critical questions. lll v‘\ 04 ..»(\. om ..v(\ . O H O H Hmnmbwq onHAV‘ B + B l 3 19‘ 1H a JV‘ PCfl'IU‘ +H' It" Om «HA 9‘ >¥flXGHQ§OU 301H Hmswan .Lmr\ use 8-0 5. -H I +wum>nmmcoo m>wum>pmmcoo zuwmeeEoo sou. hvwxmamsoo swam mmcsuflpp< cam Hm>mq zpwxmaesoo cmmzpmm meanwcowvmamm manwmmom >H mMDme 112 The Figure represents the possible relationships between the complexity level Of group members and attitude Of those members. The Roman numberals within cells designate our four group types. We have appropriately subscripted the cell and marginal means in order to clarify once more what these means represent. The capital letters H and L stand for the presence .in the groups of high and low complexity subjects (dyads), the lower case letters 1 and c for the presence of "liberals" and "conservatives" with respect to drug issues. In each cell is an expanded version of the terms used in null hypotheses (3) and (9). The null hypothesis to be used (3) states that, absent interaction, the absolute values of the two row effects must be equal to one another and that, similarily, the absolute values Of the two column effects must be equal. In each cell the means are expanded into their component parts, and the entire formula (3) is shown in expanded form below (3a). The mean Of Observations in cell 1, for example, is made up Of the grand mean/A[, plus the effects of high complexity"liberals", “1’ high complexity "conservatives", “c’ and the interaction of the two, c‘int' The other cells are similarily composed. Thus: (3a).[(/( +031 +ac +«int) - (H +a1 etc-«1,191 - l:(lt-‘xl +°‘c -°‘int) — (A-o‘l -°‘c +uint” = 0 by cancelling terms of opposite signs in this equation we have (3b). "“int = 0 113 Note that the algebraic result would be the same if any one iof-the alphas, any combination of two of them, or all three were taken to be negative numbers. (Ifa'int were negative, the equation _(Bb) would Of course read 'uc'int = 0). It is, further, obvious that equation (9), if appropriately expanded by decomposition of the means, would yield the same result. Hence, if the null-hypothesis had to be accepted, we would not ‘be able to assert that there is interaction between complexity and the participation of "liberals" and "conservatives" in the experimental groups. The results Of this contrast show that the differences are not significantly different from 0, p > .1, thus indicat- ing no significant interaction between attitude distribution and complexity.13 This means the effects of complexity and attitude are exerted independently and the original finding of a significant effect for complexity is valid. The results of our experiment in the area of control of recommendations made by groups indicate that in groups having individuals differing in complexity level there is an advantage to those individuals of low complexity in getting their atti- tudes represented. Those outcomes do not reflect a complete domination_by the low complexity subjects since the proportion of variance explained by our findings is only .27. In addition, the proportion won varies not more than 15% from the overall mean indicating compromise on the part of low complexity sub— jects as well as high complexity subjects. Other factors are Obviously Operating here, including possible experimental 114 design problems, which may add significantly toward explaining how the final recommendations are formed. Specifically, the unexplained direct effects of attitude and the effect of atti- tude on homogeneous group types are not included in our experi- ment. The attitude test itself could have allowed many individuals with particular differences, on one or more of the dimensions used, to be classified as overall "conservatives." The fact that college students generally seem to have a more liberal attitude toward the use of drugs may have induced many "conservatives" to agree to changes in drug laws, etc., to which a more general sample of subjects would not have 'agreed. For example, many "conservatives" did not appear very firm in their convictions that marijuana should not be legalized, a question which was an important public issue between student and non-student residents of East Lansing at the time of the experiment. A second design factor may have been a greater pressure to compromise in the instructions thanwas intended. The possibility of dissent was provided, but seldom used. There were only six dissenting Opinions on all recommendations made. But our findings that given like complexity the more liberal individual will win and given mixed complexity (evenly divided) the lower complexity individual will win cannot be disputed. The relative explanatory power of cognitive complexity in any situation is dependent upon the importance of the issue being considered. Our rough measure Of "wins" and "losses" does not take into account the priority, urgency or significance of the issue being considered. Thus the role 115 of complexity can vary greatly depending on the issue being discussed. A future experimenter might want to build those kinds of considerations into his measurement Of compromise. Group Level Differences in the Complexity of Recommendations In our second hypothesis (II) we prOposed that the complexity Of the group recommendations would vary with the proportion of high complexity subjects in the group. This hypothesis will be tested with a one-way analysis of variance design since we are concerned with only the gross differences between groups based on the prOportion Of high complexity subjects in the groups. If a significant difference should occur pp§£_hpg comparisons would be used to find if the differences are those we predicted. This null hypothesis may be expressed as ’ltI II¥KII =/‘(III :’£(IV' The mean in this case is the mean number of new aspects introduced by each of the different group types, those of all high complexity, two high and two ‘low complexity, and all low complexity. A new aspect is measured in the same way it was for in-group discussion except that is l” The resultant means is applied to the group recommendations. were as follows: Group Type I : Mean = 9.56 ll H O 0" .fi Group Type II : Mean Group Type III: Mean = 10.25 Group Type IV : Mean = 9.33 There are not significant differences between these means. They are, in fact, remarkably similar. 116 The high variance within group types is one reason for the failure to find significant results. The results for groups II through IV are in the predicted direction. One possible explanation for a failure to find differences is the time limit placed on group discussions. A grOup with complex individuals could possibly have Spent more time discussing details and failed to get those details transfered into pro- posals. Their initial prOposals, however, did not reflect a particularly great concern with detail. Experimental Results: Individual and Group Process Behavior Patterns Cooperation and Conflict: Hypotheses III A - E prOpose essentially that the amount Of COOperation and conflict behavior will vary with complexity level of subjects and group type. Again, differences will be treated on different levels of analysis. For overall group differences we should find that cooperation and conflict should vary with the prOportion of high complexity subjects in the group. It will vary directly in the case Of cOOperation, the higher the prOportion of high complexity subjects the greater the amount of COOperation, and inversely for conflict. In our analysis Of group type we expect to find that in heterogeneous group types a great pro- portion of COOperation and a smaller proportion of conflict will be exhibited by high complexity Subjects. In homogeneous groups the divisions should be about equal. Finally, in our individual analysis, we expect a difference in the behavior of high and low complexity subjects regardless Of the nature of 117 their group partners. High complexity subjects will exhibit greater COOperation and less conflict, while low complexity subjects will show the Opposite behavior. As we have stated in Chapter III the low intercoder ,reliability in scoring COOperative behavior makes it impossible to effectively test these hypotheses. We were able to achieve a reliability of only .99 among our three.coders. Using a mean cooperation score for groups may have partially overcome the unreliability within groups, but does not allow us to assume the scores are truly representative Of cooperative behavior. The high variance between groups also raises questions about the measurement of COOperation. The fairly equalgroup type means for COOperation, presented in Table IX, are unrepresentative due to the high variance within each group type. Type I ranges from 3.6 to 20.3 (mean number Of cOOperation behaviors for three coders); type II from 5.0 to 17.0; type III from 9.3 to 20.0; and type IV from 2.7 to 21.0. We have no means of sorting out these diverse results which could be a product of our inability to measure cooperation accurately or other group and personality factors not within the scope Of our study. The pattern Of COOperation does not seem tO be related to other behavior Observed in the experiment with the exception of new aspects introduced. Table IX shows that the amount Of COOperative behavior exhibited by individuals is significantly correlated with only the number of new aspects introduced. Given the lack of coding reliability, however, such a finding 118 TABLE IX Group Behavior: COOperation and Conflict GPOUP Type Inter coder . - reliability Group mean for I II III IV (Pearson's R) COOperation 9.95 12.08 9.9 9.62 .99 Conflict 11.86 9.12 16.56 10.81 .75 N= 9 9 8 9 Adjusted Conflict* 11.87 7.59 15.08 6.53 N: 7 7 6 7 *Highest and lowest group removed from each group type could be the result of coincidence, especially since COOperation does not correlate significantly with any other variable. For conflict we Obtained a higher reliability (Pearson's R = .75), but little more success in verifying our hypotheses. In our individual analysis the relationship between level of complexity and the number Of conflict behaviors exhibited dur- ing the experiment (Hypothesis IIIB) is not significant (r = .05, p > .05) (Table IX). The general differences between groups show that homogeneous high and low complexity group types generate virtually the same number of conflict behaviors. Group type III has a considerably higher mean, but in a One— way analysis of variance this difference proved to be nOnsignifi— cant (F3,31 = .661, p > .1). A large prOportion of the higher mean in group type III is accounted for by two groups which had 91 and 28 conflicts each. As with COOperation there is very high within group variance which makes interpretation of the findings extremely difficult. One possible factor producing 119 the high variance is the presence of one or possibly two very aggressive individuals who raise the level of conflict signifi; cantly. There is us reason, however, to expect that aggrese .siveness is related to complexity level in this situation. At the other end Of the scale were groups with four quiet subjects and those who were juSt not interested in what was going on. This situation produced very little conflict because subjects were too "shy" to disagree or simply did not care. Owing to the small number of cases in each group type in our sample, these extreme scores have large effects on final outcomes. For example, if we eliminate the higheSt and lowest conflict grOups from each group type, as we have done in Table IX, Row IV, we find a somewhat different pattern emerg— 15 The pattern shows conflict decreasing with a higher ing. prOportion of low complexity subjects except for group type III. This pattern contradicts our predicted behavior (Hypothesis IIID). 16 It It does follow the behavior pattern predicted by Stager. may be true, then, that the increased number of different ideas _generated by high complexity subjects produces more conflict despite a tendency Of high complexity subjects to compromise. Unfortunately our data on new aspects do not bear out this hypothesis. There is no significant difference in the intro- duction of new aspects between group types (see discussions below). Also, we must be very cautious in interpreting data produced by the rather arbitrary procedure used. 120 .ea. A :Oevmamspoo ewes mo. v e .03H 0 2e eee. eee. eme. eee. eee. eme. mme. Hee. Hee.. eee. eee. flee. Ammo sxoam . .me9 9:00 90m eee. eee. eee. eee. mme. eee. eee. eee.- emH. eme. mee. gees madam mmEeH sumo hem mmo.| Hos. ema. omo. esm. mmo. mac.) Hmm. cam. omo. AHHV pOOOOOhycH myooem< 3mz eee.- eee. eee. eee. mHH.- eme. eee.- eee.- eee.- gees anagram coprEQOmcH eee.- eme. eee. eee. eee.- eee. eee. mes. Ame npamasmn< Hmoemoq eme. eee. eee.- see. eee.- eee.: eee.- eev mesmesmae mdfim> eme.- eee. flee. eee.- eee.- eme. Aeo pomeesoo eee.- mee.- eee.- eee.- eee. Ame coeesamaooo eee.- Hes. eee. eee. Ame ease msmpsmm .D.m.: mac. oHo.| meo. Asv xmm mmm. «ma. Ame cowvmesomcH _ msso mmo. Ame mmseo opmzoe mpspepp< AHV hvexmaasoo O>prcwoo AmHV.AmHV.nHHV.AOHV - Ame .Amv nee Ame, Amy 4 are Ame . Ame may sow>mcwm macaw mo mmmma HH< cam zpwwaeEoo O>prcmoo mo xepvmz coepmHmshoo < x mqm .05). The small amount of information 3 seeking behavior results in an unduly large effect on the group type mean by just one individual. In this case one person asked six information seeking questions which caused a large increase in the mean for group type III. The group in which this occurred is not one of the groups with an unusually high amount of conflict discussed above. The pattern of group type III differing from other groups seems attributable to a very few aberrant individuals rather than to the effects Of the particular group structure (complex conservatives and simple liberals). Our analysis of variance on the effect Of group type shows no significant differences (p = .73) indicating that in mixed groups high complexity subjects did not seek a significantly greater quantity of information. TABLE XII Group Behavior: Information Seeking Group type intercoder I II III IV reliability Mean Information Seeking Behavior 2.79 2.5 3.26 2.72 .65 Individual differences in information seeking were related, however; zero-order correlations between information seeking and attitude and between information seeking and drug information scores (Table X) were negative and at about -.29. Conservative subjects with lower drug information scores were somewhat more likely to seek information. This relationship 125 is reasonable since one might expect those who do not know about drugs to ask questions and in light Of the fact that drug information scores and attitude are significantly related (r = .328). However, information seeking is not related to complexity level (r = -.02) as we had predicted. Thus neither hypothesis IXA nor IXB was verified in the experiment. One possible explanation for failure to find other hypothetical relationships was the Operation of the experi- mental sessions. Each group had only an hour long session. This situation was not conducive to information seeking. There were no available materials nor was there a recognized expert who could be used to answer questions. In addition, the previous research on information search contains many qualifications to the simple hypotheses we Offered.18 Information search is affected by such factors as the amount Of environmental stress and the location of initiation. The former we presumed to be constant and the latter rested with the individual in our experiment. Previous research suggests this last point is crucial to our failure to find a significant relationship.19 New Aspects: Hypotheses XA and XB suggest that the greater the individual's complexity the more likely he is to have more new ideas to inject into-the discussion of solutions to the drug problem. At the individual level there is a slightly positive, but insignificant, correlation between individual complexity and the introduction of new aspects, r = .090 126 (Table X). Hypothesis XA must therefore be rejected. Table X does show several other interesting relationships. Both attitude toward drugs and information about drugs are signifi— cantly related to the introduction of new aspects. Liberals would be in the position Of leaders in the introduction Of program suggestions since most of the group discussions cen- tered around the problem Of whether or not to change existing laws or modify existing treatment and education programs through expansion. The conservative, among most of our subjects, was a stand pat individual rather than one pressing for stricter controls, etc. This would explain the positive correlation between attitude and new aspects. Subjects with more infor- mation about drugs might also be expected to contribute more ideas since they are generally more familiar with the whole "drug scene." TALBE XIII Group Behavior: The Introduction Of New Aspects Group Type Intercoder I II III IV Reliability Mean Number of New Aspects Introduced in 32.08 31.91 33.87 32.56 .89 Group Discussion Table XIII shows no significant effects (one-way analysis of variance) for groups on the introduction of new aspects. The higher proportion Of complex subjects does not increase the introduction Of new ideas within the group, just as was the case for the complexity Of final recommendations. 127 There are also no main effects for group type in our ,two-way analysis. High and low complexity dyads in hetero- geneous groups do not appear to behave differently on this variable. However, there is a significant interaction between complexity and sex in the introduction Of new aspects. Table XIV, the analysis of variance for new aspects, shows the interaction to be significant at the .099 level with the simple effects Of complexity and sex being far from significant. TABLE XIV Analysis of Variance Table for Effects of Cognitive Complexity (Group Type) and Sex on the Introduction Of New Aspects in Groups Degrees Mean Sum Sum of of of Signifi— Squares Freedom Squares F cance Complexity (A) .098 3 .016 .886 .961 Sex (B) .0001 1 .0001 .006 .939 (A X B) .161 3 .059 2.98 .099 Error .987 Total .689 Using Scheffe ppst hoc comparison techniques it was determined that the only significant differences occur between the means for males and females of group type I (all high complexity) and the means for group type III (males) and II (females) (Table XV). 128 TABLE XV Cell Means for the Effects of Cognitive Complexity (Group Type) and Sex on the Introduction of . New Aspects in Groups (Means are the prOportion of Liberal New Aspects) sex I II III IV Mean Male .688 .500 .987 .698 .583 Female .988 .669 .569 .620 .587 , ' Grand Mean Mean .589 .583 .527 .635 .582 N 9 9 8 9 Explained variance = 29% In groups II and III we expect that high complexity subjects will dominate the introduction of new aspects. Keeping in mind that we are dealing with an overall mean Of .58, we find that in three of four Of these groups this is the case (although this relationship is not statistically significant by itself). For group type III the mean propor- tion of new aspects introduced is less for low complexity subjects than for high complexitysubjects. ,In group type II this is true for females, but not males where the result is in the Opposite direction. From.these results we may draw some evidence for Hypothesis XB, but only weakly, since the effects of complexity are confounded with those Of sex and are not strong enough to be statistically significant. I can offer no explanation as to why type II heterogeneous male groups should behave differently than female groups. In group type I we could expecteach cell to be divided .approximately equally in the introduction of new aspects. 129 However, for males the liberals seem to dominate the introduction Of new aspects, while for females the conserva- tives dominate. This latter finding is in contradiction to the overall results for the effects of attitude which show that liberals should dominate the introduction Of new aspects (Tables X and XV). Talking: Our final hypothesis predicts that complex subjects will engage in a greater amount of talking than simple sub- jects. A confirmation of this minor hypothesis may add some evidence to the argument that high complexity subjects Speak in greater detail and are more likely to use complicated arguments in attempting to present their arguments. Our group type analysis of variance does not show significant results, although the findings for heterogeneous groups are in the predicted direction.(Tab1e XVI). TABLE XVI Cell Means for the Effects Of Cognitive Complexity and Sex on the Proportion of Group Discussion (Means are the prOportion Of liberal discussion) sex I II III IV Means Male .620 .510 .987 .563 .596 Female .520 .599 .990 .562 .593 . Grand Mean Means .570 .552 .988 .562 .593 N 9~ 9 8 9 On the individual level Table IX shows a small significant correlation between complexity level and the amount of 130 speaking, thus providing another modicum of evidence for our hypothesis. The correlation between attitude and per cent Of-time spoke is also small, but significant. This might be expected for the reasons discussed above, particularly that it is the liberals that must carry the fight for change, while the conservatives are mostly content to stand pat. Summary and Discussion Our central hypotheses about the effect Of group types on the outcome Of group recommendations were supported by our data. In heterogeneous groups, mixed high and low com- plexity, cognitively simple dyads were more likely to "win" in cases where recommendations were disputed than were cog— nitively complex dyads regardless of attitude (Hypothesis IA). In homogeneous groups the prOportion of "wins" by "liberals" was close to the experimental mean (Hypothesis IB). Thus, the distribution of complexity levels among grOups has a significant influence on the nature of the groups' decisions. In cases where attitudes and complexity level differ the lower cOmplexity subjects are more likely to have their views dominate the final outcome of group negotiations. In instances where complexity levels are the same, there appears to be no bargaining advantage and we cannot predict the outcomes on the basis of the complexity variable. These findings suggest that one Of the psychological information processing variables which Simon prOposed as being important elements in our ability to predict the behavior of nonrational_groups is cognitive complexity. The increased information processing 131 ability of the cognitively complex individual apparently leads to behavior which affects the outcome Of group discussions. Support for the concept of cognitive complexity as an independent factor capable of influencing behavior is Obtained from the diScriminant validity of our measure of complexity. Level of cognitive complexity was found not to be signifi- cantly related to either I.Q., as measured by several different college aptitude tests (S.A.T., C.Q.T., A.C.T. and the M.S.U. Reading Test), or information, as measured by a test on drug information develOped eSpecially for this study. This was true for the total screening test sample (N = 1278) and the 190 experimental subjects. We also expected cognitive complexity to be an indicant Of behavior at several levels within the group processes occurring in our experiment. We expected cognitively complex individuals to be more COOperative, to exhibit less conflict behavior, to use more logical arguments, to use fewer value arguments, to seek more information about the subject, and to introduce more new ideas into the discussion. The cognitively simple subjects would behave in the Opposite manner. Confirma- tion of hypotheses about these behaviors would increase the validity of our conceptualization of cognitive complexity. These behaviorsare not logically related to the Operation of the complexity variable as it effects group outcomes. Failure to find support for our hypotheses in this area does not diminish our other results. We do not directly measure the process of changing or failing to change reCommendations, but 132 only the results of that process. Positive results on these hypotheses would have increased validity by showing that other behavior theoretically related to complexity level also varies in the predicted manner. Failure to establish these results raises questions about that validity, but does not indicate a rejection of the general theory. Positive findings were, unfortunately, not forthcoming in this experiment. We tested for the effects of these group process variables at three levels: (1) the overall differences between groups based on the prOportion Of high complexity subjects within the group, (2) the difference between group types, where we were primarily interested in the effects of homogeneous and hetero— geneous group structure on the behavior of high and low com- plexity subjects, and (3) individual differences in complexity level regardless of group structure. Hypotheses about group process were examined at one or more of these levels in each case. A second predicted group result was a direct relationship between prOportion Of high complexity group members and the (complexity of recommendations. The results showed mean :recommendations to be about equal across the four group types, ‘thus the hypothesis was not confirmed. Several factors in the (design of the experiment may have affected the initial lrypothesis. Of particular importance was the time limit. If liigh complexity subjects discuss prOposals in greater detail ‘the number of prOposals they produce could be limited. The 133 The argument that such proposals would be more complex is weakened by the procedures used to record recommendations. A more elaborate procedure for recording recommendations and a longer experimental session, perhaps multiple sessions, may provide a better testing ground for this hypothesis. We were able to measure conflict successfully and found, after some manipulation of the data, that the trend was probably more in a direction Opposite to that we have predicted. High complexity groups had greater conflict. A similar and signifi— cant finding was Obtained when we measured the number of issues in conflict in initial dyad recommendations. 'The number of issues in conflict was greater between high com- plexity dyads than between low and high complexity dyads (heterogeneous groups) and all low complexity dyads. Although this pattern was not reflected in the group discussion it may have had an important effect on other aspects Of the group ‘process, Such as, the production of a larger number of recommendations (complexity Of outputs), where more debate over'conflicts could have reduced complexity of Outputs. Low complexity may produce disagreement, but the disagreement may be handled in a different way. This trend was not sustained in our analysis of group types, where high .and low complexity subjects were not significantly different in heterogeneous groups, or in our analysis of individuals, xihere there was an insignificant correlation between conflict .and complexity level. Although we do not have systematic data <3n this behavior, in several groups there was a tendency to .avoid discussing areas of greatest disagreement. If a subject 139 was brought up by one group it was Often drOpped after a short discussion or even ignored. The most common subject to receive this silent treatment was the legalization of "hard" drugs. In other instances topics of seeming importance to the issue being discussed were simply never raised. Again, legalization was an obvious omission. Given our measure of the number of conflicts, rather than their intensity, we could have overlooked much Of the conflict that existed. Redesigned hypotheses in this area should probably take into account the intensity of conflict. The problem of measuring this intensity is, of course, the major limitation in gathering useful data. We were also unsuccessful in verifying our hypotheses about information seeking. Again, a design problem could have been the major factor. For example, had we used two group sessions and allowed a greater Opportunity for subjects to seek information we would have provided a better test for these hypotheses. As it was, the one-hour group sessions did not give the subjects a great Opportunity to demonstrate this kind of behavior. The rare occurrence Of information seeking is indicative of this problem. The introduction of new aspects into the group discussion showed a slight relationship to complexity in our group type analysis, but it was tempered by the differences between the sexes. Women exhibited the predicted behavior characteristics, 'though not enough to be statistically significant, while men had mixed results. There is nothing in the literature on differences in male—female performances in groups to suggest 135 why this should have been the case. In addition, the role of sex was not significant in any of our othermeasurements of group behavior. The broken pattern occurred for high com- plexity liberal men in heterogeneous groups. Liberal men did not exhibit that pattern in homogeneous groups. There seems to be no immediately Obvious explanation for the exception that did occur. Nor can we Offer an explanation for the deviance of high complexity liberal women in homogeneous groups that occurred in the same variable. Differences between groups and between individuals were not significant. Three of the major measures Of group behavior were lost due to a failure to Obtain either a sufficient volume of data (the use of value and logical arguments) or'a high enough inter-coder reliability (COOperation). The reliability problem was also the initial reason for the limiting of value and logical argument coding. We have, therefore, no real test of hypotheses concerning behavior in these areas. Certainly, any future replication of this experiment should include more reliable measures of these behaviors. The failure to find predicted behavior for the internal functioning Of experimental groups raises questions about the previous findings in other research, the predictive power of complexity in those areas, and the design of our experiment. The kind of behavior measured was not crucial to how groups decided about recommendations. They were measures of how simple and complex individuals should behave in group situations. They were apparently independent of the recommendations produced by the groups, although better measuring procedures 136 would provide a much clearer answer to that particular question. What they fail to provide are additional guages for the validity of our test for complexity. The failure of these hypotheses seems due more to methodological problems than the nature Of the_theory or the test Of complexity. Such results suggest a replication based on revised procedures for assessing internal group behavior. The nature of these changes as well as some possible explanations for the process which produced our positive results will be discussed in the Conclusion. 9! FOOTNOTES 1 Donald T. Campbell and Donald A. Fiske, "Convergent and Discrimanent Validation by the Multitract-Multimethod Matrix," Psychological Bulletin, LVI, 2 (1969), pp. 81-105. 2Ibid. 3Joseph Vannoy, "Generality of Cognitive Complexity— Simplicity As A Personality Construct," Journal of Personality and Social Psychology, II, 3 (1965), pp.-385-396. 1"Although the tests were administered to the same subjects at the same time, the study was conducted by a different researcher and the complicated procedures involved in processing these data have delayed final analysis. 5In a pre-test it was found that the SAT and the CQT were not significantly different in their relationship to cognitive complexity. Nor were they significantly (using p .05 as the cutoff point) related to complexity. (Total N = 59) For this reason the three scores used in the final screening test were not differentiated for analysis. 6This measure was formed in an attempt to isolate inte— gration from differentiation. It was not used in the experi- ment and was created only to see if it differed from differen- tiation or was related to any of the other screening test items. The measure is derived by dividing the differentiation score (number of aspects mentioned) into the total score. The range was from .9 to 6, with the higher score indicating the greater integrative ability. 7We use this measure Of I. Q. because it is the only :measure common to all subjects and it also correlates highly at least 7, with all other indicants (See Table I). 8Milton Rokeach, The Open and Closed Mind, (New York: Basic Books, 1960). 9The Analysis Of Variance model was taken from William Gunther, Analysis of Variance, (Englewood Cliffs, New Jersey: Prentice—Hall, 1969T. The particular procedure used in the (mamputer was the Least Squares (LS) program in STAT manual no. :18, Michigan State University. The computer used was the CDC 3600. 137 138 10These post hoc comparisons were made using the Scheffe technique as described in Henry Scheffe, The Analysis Of Variance (New York: Wiley, 1959). All subsequent comparisons will use these procedures. The above comparison was set up gm a cgnfidence interval with the null hypothesis being XII ‘ XIII = 0 The formula for the confidence interval is XII _ YIII ' 33$ 5' V ‘ 2II ' 7111 + 33'? where X andXIII are the means for the effects of the different group types and 83.1) = .91} (k - 1) (N - k) x MSEé'(ci2) 327' and is calculated for whatever degree of significance is desired. For p = .05 with df = 1, 15 on the data Obtained. The value is .187. Thus:- "" "'" _ AA _ XII - XIII - -.203 and Sony - .187 then -.203 - .187 g \V .<_ - .203 + .187 Since 0 is not an included value the difference between these means is significant. 11... Ho' I ' XII + (XIII ‘ XIV) = 0 x XI " XII + (XIII " YIV) " S'i’s‘ys XI ' xII + -- AA .550 - .997 + (.650 - .595) + .1600 54’; .550 — .997 + (.650 - .595) - .1600 .208 + .160 5V; .208 - .160 This equation was calculated for a p Of .01, thus the difference is significant at the level of p s .01. Using the second formula would produce the same results. 121 am indebted to Mrs. Janet Eyster, Statistical Consul- tant for the Statistics Department, Michigan State University for suggesting and setting up these contrasts. 139 13The formula for the contrast is as follows [WI'/(11)'A(AIII'/(Iv)]' 53"? 3 "' 5 [WI ‘/711) ‘ (.650 - .595) + .0760 .078 5 T < .079 fig”. 3,31 x MSE)(O_:) j. So» > .9 u = /2.28 x .0181 (l + l + l r}-) , 9 9 9 9 = .076 2.28 is the F value for p = .1 at 3 and 31 degrees of freedom ll'See Chapter III, p. (measure new aspects) 15There is no statistical test applied to these data since the manipulation used to Obtain it did not meet statistical assumptions. 16Paul Stager, "Conceptual Level as a Composition Variable in Small-Group Decision-Making," Journal of.Personality and Social Psychology, VIII, 2 (1967), pp. 152-161. 17The problems centered around what amounted to support- ing logical evidence for a particular argument. My coders often knew more about drUgs than the subjects. This meant that they could spot fallacious arguments used by subjects who were not even aware they were false. It was difficult for the coders to see them as logical. 0n the other hand statements beginning with the "I believe. . ." approach were Often true in fact and it was difficult to call them value statements. There was also the problem Of statements which the coders could not identify as being factually true or false. The coders were not drug experts. This left only the most Obvious statements, logical or value, to be coded and even here there was disagreement. 18See Chapter II, p. 32. 19Siegfried Streufert, and Carl H. Castore, "Information Search and the Effects of Failure: A Test Of Complexity Theory," Journal Of Experimental Social Psychology, 7 (1971), pp. 125—193. CHAPTER V Conclusion Our experiment has provided evidence that in groups with high and low complexity subjects of opposing attitudes the low complexity subjects have a bargaining advantage. When there is an issue in conflict the attitude position of the low complexity subjects is more likely to prevail. This finding at least partially answers questions raised by Simon in his discussion Of decision-making processes.1 Individual differences in the psychological limitation Of information processing capabilities (cognitive complexity) provide an indication of what decisions will be made by a group com- posed Of these subjects. Hence, we are better able to pre- dict the result if we know the group is composed Of individuals differing in their complexity level and attitudes toward the subject being considered. Hypotheses predicting these outcomes were derived from our conceptualization of cognitive complexity and had little support in the previous literature which focused primarily on group processes. Our own hypotheses about group process failed to illuminate the problem of how behaviOr inside the group leads to the results we Obtained for group recommendations. At this point, then, we have a Skinnerian S — R model with no 190 191 hard evidence about what, if anything, goes on inside the "black box" which would resolve this issue. However, it is possible to speculate on what was happening in the groups based on some unsystematic Observa- tions by the experimenter. We know that a greater propor- ition of high complexity subjects in a group produces a larger number Of conflicts between dyads within the group, at least when initial recommendations are used as the indi— cator. This greater issue conflict is not directly reflected in the group discussion. It may have the effect of depressing the number of recommendations made by high cOmplexity groups. It may also lead to different means of handling conflict within groups. A major problem in attempting to Observe the processes by which decisions were made was the use Of unspoken pro— cedures. In some of the 99 experimental groups (35 in the actual experiment and 9 in the pre-test), certain issues were simply not discussed. The question of legalization of hard drugs, for example, may never have been mentioned by a subject who favored it because he realized that those dis- agreeing with him would Oppose it. Thus, Open conflict was avoided.2 This "behavior" is difficult to code, however, because we cannot be sure why the particular suggestion was not considered. It may have been that time was a factor or that the prOposal was just not important enough to its originator. Systematic Observations could be included in a future experiment by having all subjects rank their proposals in importance, by Observing, during the group sessions 192 whether or not the question was raised, and by using a post test questionnaire to find out why it was not raised. A related pattern occurring in some groups involVed having an issue raised, then immediately dropped or ignored by the subjects. This seemed to happen in three kinds of situations; where the issue was highly controversial or where it was of little importance to the participants or both. Again, the presence Of conflibt in the group process would be minimized in this pattern. Ranking of proposals and careful Observation Of how each new suggestion is treated are means to test propositions about this treatment Of new issues. There was for the most part an absence of intense disagreement in the group discussions, even among subjects in homogeneous low complexity groups where it seemed most likely to occur. This may have been the result of the experimental situation and the absence of strong disagreement on the issue being discussed, but it seems at least partially the result of the two means of handling disagreement discussed above. Another means of handling conflict was extended debate. JBecause there was no control Of group debate some groups tended 'to be rather unsystematic and discussion deteriorated into ibull sessions. An issue may arise, be discussed for a period, “then just fade into another issue, never being resolved. Sometimes the question would be reOpened when the group was (iictating its recommendations and at other times it was 3 imply ignored . 193 There are, then at least three different means of resolving conflict situations which are not immediately evident and which do not lend themselves to reliable scor— ing. Each is an indicator of behavior that might be expected of high or low complexity subjects and may, therefore, be important in the process of group decision making. Thus, we might expect high cOmpleXity individuals to withhold the introduction of a prOposal they favor if they perceive it will not be well received (avoiding conflict). 'Low complexity individuals, on the other hand, are more likely to ignore or perhaps not even perceive other peoples' proposals lead- ing to a situation where new ideas are presented and just ignored. These propositionsoperate in addition to more '"normal" channels for handling conflict in which issues are debated and one or both sides compromise to produce the final proposal. In all cases a very careful monitoring of debate is required to determine just when each type Of ‘conflict resolution takes place and if it has anything to do ‘with.what individuals finally agree to in group recommendations. It may be possible that none Of the intra-grOup debate is .indicative of how particular individuals will behave when it comes to agreeing or disagreeing with final recommendations. Another factor not considered in our original experiment ;is the relative importance of the proposals recommended. It :is possible that many important issues are treated in the ways discussed above, while only unimportant proposals are fully'debated and reSOlved. It is also possible that the 199 higher proportion of "wins" achieved by low complexity gsubjects are on unimportant issues and that on really significant issues they do not maintain their "advantage." Or the Opposite situation may prevail, where the simple individual is even mOre rigid as the issues are relatively more important to him. It is the last alternative which seems both theoretically and empirically more likely. In the recommendations prOposed in our experiment those about education and treatment seemed to have the widest acceptance. Proposals about radical changes in the legal status Of drugsand drug users were less widely accepted. The latter_ type of change also seemed to provoke the most vehement debate when it was considered. .Thus, had we measured issue salience we may have found that low complexity subjects were :more adamant on more salient issues. Such a proposition is consistent with findings discussed in Chapter II, that high complexity individuals were less likely to take extreme positions on their attitudes. ' Other behavior likely to be indicative of change in a group'discussion follow from the behavioral characteristics cxf high and low complexity individuals. A high complexity iJniividual, for instance, would be more likely to exhibit Ixile flexibility and be better able to take the other's position. He should also be better able tO see nuances. But both. of these behaviors are extremely difficult to measure in the context of an open discussion of issues. Controlling an experiment to sharpen measurement is one Ineaxus to overcome the problems we have discuSsed. Subjects 195 .might be required to rank their prOposals orthe experimenter might provide a standard list Of prOpOsals so that all sub« jects could be directly compared on the amount of change' they exhibited. But, although these proposals would improve’ measurement accuracy, they would also make the experiment less realistic and could put less complex subjects on a par with higher complexity subjects on the introduction Of new ideas. We suspect that a set list of proposals rated by all subjects before the experiment would narrow the range of the rdebate if the list were short and limit its depth if the list were long. It is Obvious, however, that some control is i needed if we are to measure accurately that part Of the group process which reflects the expected behaviOr of low and high complexitysubjects. Such control might also be accompanied by a control group Of subjects who do not :participate in a group at all, but only make recommendations on the basis of written versions Of attitude Opposites' proposals. ' Our experiment has provided some evidence that cognitive complexity is or can be an important factor in the decisions made by groups. It has not answered questions about possible (effects Of attitude on groups differing in complexity level, .although we do know there was no interaction between com- plexity and attitude. Our experimental design precluded (Mfr attempting to answer this question despite the possibility (If a real world relationship. Such queStions should certainly -Ix3 taken up in future research. The exact pattern of complexity and attitude interrelationships could be extremely important 1&6 in determining both what happens within groups and the kinds of decisions groups produce. There are two additional rather extensive areas yet to be examined; the nature and importance of which issues are compromised, and the exact process by which these compromises (or non-compromises) take place. We have made several rather general suggestions as to how future research in these areas might proceed and hopefully other researchers will replicate our own findings and try to answer questions we have not been successful in answering. FOOTNOTES 1See Chapter I. 2Gamson offers a possible explanation in his model of group behavior called the "anticompetitive theory." In the situation where resources of potential coalitions are rela- tively equal (as in our experiment) there is a tendency to follow "the path of least resistance" and promote agreement "with a minimum of Haggling and unpleasantness." William A. Gamson, "Experimental Studies of Coalition Formation," cited in Leonard Berkowitz, ed., Advances in Experimental Social Psychology, Vol. I, (New York: The Academic Press, 1H7 BIBLIOGRAPHY '91,“ BIBLIOGRAPHY BOOKS Abelson, Robert P., Aronson, Elliot, McGuire, William J., Newcomb, Theordore M., Rosenberg, Milton J., and Tannenbaum, Percy H. Theories of Cognitive CODSistency: A SOurcebook. Chicago: Rand McNally and Company, 1968. Adorno, T.W., Frenkel - Brunswik, Else, Vevinson, Daniel j., Sanford, R. Nevitt. The Authoritarian Personality. Part One and Two. New York: John Wiley 8 Sons, Inc., 196”. ‘ Anastasi, Anne. Psychological Testing. London, The Macmillan Company, Collier-Macmillan Limited, 1968. Angoff, William H. The College Board Admission Testing Program. New York: Colege Entrance Examination Board, 1971. Apter, David E. (ed.). ‘IdeOlogygand3Di3content. New York: The Free Press, 1964. " I . - Bales, Robert F. Interaction Process Analysis. Cambridge, Mass: Addison-Wesley, 1951. Bannister, Donald (ed.). Perspectives in Personal Construct Theory. London, Academic Press, 1970. ‘ Barber, James D. Power in Committees: An Experiment in Governmental PrOcess. Chicago: Rand McNally, 1966. Bass, Bernard M. Leadership, Psychology and Organizational Behavior. New York: Harper and Row, 1960. Berelson, Bernard and Steiner, Gary A. Human Behavior. (New York: Harcourt, Brace and World, Inc., 196M). - -~Berkowitz, Leonard. ’AdvanCes in Experimental Social jEsyghOlogy. Volume M. New York: AcademicPress, 1969. - ,.17. (ed.). 'Advances in Experimental Social P‘sy’ch‘o‘l‘o‘gy. Volume 5. New York: Academic Press, 1970. Bettinghaus, Erwin P. ~PerSuasive Communication. New ‘Yorfin Holt, Rinehart and Winston,l968. 1H8 __ 1’49 Biddle, Bruce J., and Edwin Thomas. Role Theory: Concepts and Research. New York: Wiley, 1966. Bieri, James, Atkins, A.L., Briar, S., Leaman, R. L., Miller, H. and Tripodi, Tony. Clinical and Social Judgment: The Discrimination of Behavioral Information.. New York: John Wiley, 1966. Blalock, Herbert. Social Statistics. New York: McGraw Hill, 1960. Borgatta, Edgar F. and Lambert, William. Handbook of Personality Theoryand Research. Chicago: Rand McNally, 1968. 'and Crowther, Betty. A Workbook for the Study of Social Interaction Processes. Chicago: Rand McNally afia Company, 1965. Bower, Gordon H. and Spence, Janet Taylor (eds). The Psychology of Learning and Motivation. Volume 3. New York: .Academic Press, 1969. Brodbeck, May (ed.). Readings in the Philosophy of the Social Sciences. London: Macmillan, 1968. , Brown, Roger. Social Psychology. New York: The Free Press, 1965. Bruner, Jerome , e1.” a_1_'. (eds .'). Contemporary Approaches 'to Cognition. Cambridge: Harvard University Press, 1957. and Krech, David. Perception and Personality. NeW'York: Greenwood Press, 1968. , Olver, Rose R. and Greenfield, Patricia M., 33. EEL' Studies in Cognitive Growth. New York: John Wiley 8 Sons, Inc. 1966. ., ,.Goodrow, Jacqueline J., Austin, George A. A Eltudy ofThinking. New York: John Wiley 8 Sons, Inc. 1956. Buros, Oscar Krisen (ed.). Personality Tests and Reviews. Iiighland Park, N.J., The Gryphon Press, 1970. Campbell, Angus, Converse, Philip E., Miller, Warren E. arui Stokes, Donald E. The American Voter. New York: John Wiley 8 Sons, Inc. 1960. Campbell, Donald T. and Stanley, Julian C. Experimental and (pmasi—Experimental Designs for Research. Chicago: Rand fiéNally 8 Company, 1963. Cartwright, Dorwin and Zander, Alvin. Group Dynamics. New York: Harper 8 Row, Publishers, 1968. 150 Cattell, Raymond B. (ed.). Handbook of Multivariate Experimental Psychology. Chicago: Rand McNally 8 Company, 1966. and Warburton, Frank W. Objective Personality and Motivation Tests. Urbana, Ill: University of Illinois Press, 1967f Collins, Barry E. and Guetzkow, Harold. A Social Psychology of Group Processes for Decision-Making. New York: John Wiley 8 Sons, Inc., 196”. Cronbach, Lee J. and Gleser, Goldine C. ngchological Tests and Personnel Decisions. Urbana, Ill: University of Illinois Press, 1965. Cyert, Richard M. and March, James G. A Behavioral Theory of the Firm. Englewood Cliffs, New Jersey: Prentice Hall, 1963. . Davies, James C. Human Nature in Politics. New York: Wiley, 1963. Edwards, Allen. The Measurement of Personality Traits. Eulau, Heinz. The Behavioral Persuassion in Politics. New York: Random House, 1963. Evans, Richard and Rozelle, Richard. Social Psychology in Life. Boston: Allyn and Bucon, 1970. Feldman, Kenneth A., Newcomb, Theodore M. The Impact of .Qollege on Students. Volume I. San Francisco: Jossey-Baso Inc., Publishers, 1969. Festinger, Leon. Conflict, Decision and Dissonance. , Stanford, California: Stanford University Press, 1964. , Back, Kurt, Schachter, Stanley, Kelley, Harold H., and Thibaut, John. Theory and Experiment in Social Communica- tion. Ann Arbor, Michigan: Edwards Brothers, Inc. 1952. Fisher, Roger (ed.). International Conflict and Behavioral Science. New York: Basic Books, 1969. Fiske, Donald W., and Maddi, Salvatore R. (eds.). Functions (xf'Varied Experience. Homewood, Illinois: The Dorsey Press, 1961. i I' Fromm, Erich. Escape From Freedom. (New York: Rinehart 8 Company, Inc. 19u1). Furth, Hans G. Piaget and Knowledge. Englewood Cliffs, INeW'Jersey: Prentice-Hall, Inc. 1969. 151 Guilford, J. P. The Nature of Human Intelligence. New York: McGraw-Hill Book Company, 1967. Glass, Gene V. and Stanley, Julian C. Statistical Methods in Education and Psychology. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. 1970. Golembiewski, Robert T. The Small Group: An Analysis of Research Concepts and Operations. Chicago: University of Chicago Press, 1962. Greenstein, Fred I. Personality and Politics. Chicago: Markham Publishing Company, 1969. Greenwald, Anthony 6., Brock, Timothy C. and Ostrom, Thomas M. Psychological Foundations of Attitudes. New York: Academic Press, 1968. Grey, Alan L. (ed.). Class and Personality in Society. New York: Atherton Press, 1969. Guenther, William C. Analysis of Variance. Englewood Cliffs, New Jersey: Prentice-Hall, 1964. Guetzkow, Harold. Groups, Leadership and Man. New York: Russell and Russell, 1951. _ , Alger, C.F., Brody, R.A., Noel, R.C. and Snyder, R.C. Simulation in International Relations. Englewood Cliffs, N.J.: Prentice-Hall, 1963. Hare, A. Paul, Borgatta, Edgar F., and Bales, Robert F. (eds.). Small Groups. New York: Alfred A. Knopf, 1966. Harvey, O.J., Hunt, David E., and Schroder, Harold M. Conceptual Systems and Personality Organizations. New York: Wiley, 1961. (ed.). Experience Structure and Adaptability. New York: Springer PuBiishing Company, Inc., 1966. (ed.). Motivation and Social Interaction, New York: The Ronald Press, 196?. Hays, William. Statistics. New York: Holt, Rinehart and Winston, 1963. Hellmuth, Jerome (ed.). Cognitive Studies. Volume I. New York: Brunner Mazel Publishers, 1970. Hopkins, Terence. The Exercise of Influence in Small Groups. Totowa, N.J.: -The Bedminster, 196k. 152 Horst, Paul. Personality: Measurements of Dimensions. San Francisco: Jossey-Bass, Inc., Publishers, 1968. Hovland, Carl 1. and Janis, Irving L. (eds.). Persuasibil— ity. New Haven: Yale University Press, 1959. Insko, Chester A. Theories of Attitude Change. New York: Appleton-Centruy-Crofts, 1967. N Jackson, Douglas N. and Messick, Samuel (eds.). Problems in Human Assessment. New York: McGraw-Hill Book Company, 1967. Katz, Daniel and Kahn, Robert L. The Social Psychology of Organizations. New York: Wiley, 1966. Kelly, George A. A Theory of Personality. New York: W.W. Norton 8 Company, Inc. 1963. . Kelman, Herbert C. (ed.). International Behavior. New York: Holt, Rinehart and Winston, 1965. Kessel, John, Cole, George and Seddig, Robert. Micro- olitics: Individual and GroupyLevel Concepts. New York: Holt, Rinehart 8 Winston, 1970. Kirk, Roger E. Experimental Design: Procedures for the Behavioral Sciences. Belmont, Cal.: Brooks Cole, 1968. Kirscht, John P. and Dillehay, Ronald C. .Dimensions of .Authoritarianism: A Review of Research 8 Theory, Lexington, Kentucky: University of Kentucky Press, 1967. Krech, David, Crutchfield, Richard S. and Ballachey, Egerton L. Individual in Society. New York: McGraw-Hill Book Company, Inc. 1962. Lane, Robert E. Political Idelogy. New York: The Free Press, 1962. Political Life. New York: The Free Press, 1959. Political Thinkingand Consciousness. Chicago: Markham , 19 6 9 . Lazarus, Richard S. ,Psychological Stress and the Coping Process. New York: McGraw-Hill, 1966. Lennard, Henry L. and Bernstein, Arnold. Patterns in Human IInteraction.‘ San Francisco: Jossey-Bass, Inc., 1969. ' Lewin, Kurt. A Dynamic Theory of Personality. New York: rthraw-Hill Book Company, Inc., i935. 153 . Field Theory in Social Science. New York: Harper 8 Brothers Publishers, 1951. . Resolving Social Conflicts. New York: Harper 8 Brothers Publishers, 1908. Lindzey, Gardner (ed.). Handbook of Social Psychology. Reading, Massachusetts: Addison-Wesley, 1950. and Aronson, Elliot (eds.). The Handbook of Social Psychology. Reading, Massachusetts: Addison-Wesley Publishing Co., 1968. Lyden, Fremont, 3:. El’ (eds.). Politics, Decisions and Organization. New York: Appleton-Century-Crofts, 1969. Maddi, Salvatore R. Personality Theories: A Comparative Analysis. Homewood, Illinois:. The Dorsey Press, 1968. Madron, Thomas Wm. Small Group Methods. Evanston, Ill: Northwestern University Press, 1969. Maher, Brendan A. (ed.). Progress In Experimental Per- sonality Research. Volume 2. New York: Academic Press, 1965. . Progress in Experimental Personality Research. Volume u. New York: Academic Press, 1967. Maier, Norman R.F. Problem Solving and Creativity. Belmont, California: Brooks Cole, 1970. March, James G. (ed.). Handbook of Organizations. Chicago: Rand McNally 8 Company, 1965. and Simon, Herbert A. Organizations. New York: Wiley, 1958. Milbrath, Lester W. Political Participation. Chicago: Rand McNally 8 Company, 1965. Nunnally, Jum C. Psychometric Theory. New York: McGraw- .Hill, 1967. Parsons, Talcott, Bales, Robert F. and Shils, Edward A. VhorkingPapers in the Theory of Action. Glencoe, Illinois: The Free Press, 1953. Petrullo, Luigi and Bass, Bernard (eds.). Leadership and IInterpersonal Behavior. New York: Holt, Rinehart and Winston, 1' 9731, pp. 201—215. Piaget, Jean. The Construction of Realityin the Child. New York: Basic BooE, Inc., 1951}. 15” . Logic and Psyphology. New York: Basic Books, Inc., I957. ‘ . The Mechanisms of Perception. New York: Basic Books, Inc. Publishers, 1969. Structuralism. New York: Basic Books, Inc., 1970. , Rapoport, Anatol and Chammah, Albert M. Prisoner' s Dilemma. Ann Arbor, Michigan: The University of Michigan Press, 1965. Rieselbach, Leroy N. and Bulch, George 1. Psychology and Politics. New York: Holt, Rinehart and Winston, Inc., 1969. Robinson, John P. , Rush, Jerrold G. and Head, Kendra B. Measures of Political Attitudes. Ann Arbor, Michigan: Institute for Social Research. September, 1968. and Shaver, Phillip R. Measures of Social Ps cholo ical Attitudes. Ann Arbor, Michigan: Institute for Soc1a1 Research. August, 1969. ‘ Rokeach,.Milton.' The Open and Closed Mind. New York: Basic Books, Inc., 1960. -Rosnow, Ralph L. and Robinson, Edward J. Experiments in Persuasion. New York: Academic Press, 1967. Schroder, Harold M., Driver, Michael J. and Streufert, Siegfried. Human Information Processing. New York: Holt, Rinehart and Winston, Inc., 1967. Shaw, Marvin B. Group Dynamics. New York: McGraw- Hill, 1971. and Wright, Jack M. Scales for the Measurement of Attitudes. New York: McGraw-Hill Book Company, 1967. Sherif, Muzafer and Sherif, Carolyn W. Interdisciplinery Relationships in the Social Sciences. Chicago: Aldine, 1969. Simon, Herbert. Models of Man. New York: Wiley, 1957. Singer, J. David (ed.). Human Behavior and International Politics. Chicago: Rand McNaIIy, 1965. . Quantitative International Politics: Insights Iand.Evidence. New York: The Free Press, 1968. Smith, M. Brewster, Bruner, Jerome S. and White, Robert W. 08inions and Personality. New York: John Wiley 8 Sons, Inc. , 155 Snyder, Richard C., Bruck, H.W. and Sapin, Burton (eds.). Foreign Policnyecision-Making: An Approach to the Study of International PoIitics. Glencoe, I11.: The Free—Press, 1962. Stagner, Ross. Psychological Aspects of International Conflict. Belmont, California: Brooks CoIeIPuinshing Co., 1967. Thibaut, John and Kelley, Harold. The Social Psychology of Groups. NewYork: Wiley, 1959. Torgerson, Warren S. Theory and Methods of Scaling. New York: Wiley, 1958. Verba, Sidney. Small Groups and Political Behavior. Princeton, New Jersey: Princeton UniverSity’Press, 1961. Walker, Helen M. and Lev, Joseph. Statistical Inference. New York: Holt, Rinehart and Winston, 1953. Werner, Heinz. Comparative Psychology of Mental Develop- ment. New York: International University Press, Inc., 1998. Winer, B.J. Statistical Principles in Experimental Design. New York: McGraw-Hill, 1962. Witkin, H.A. 23, él- Personality Through Perception. New York: Harper 8 Brothers, 1959. . Psycholpgical Differentiation. New York: John Wiley and Sons, Inc., 1962. DISSERTATIONS Cegala, Donald J., "Cognitive Complexity, Cognitive Similarity and Sex in Dyadic Communication." Unpublished Ph.D. dissertation, Department of Psychology, Florida State University, 1972. Cox, Gary 8., "Cognitive Structure:' A Comparison of Two 'Pheories and Measures of Integrative Complexity." Unpublished IHI.D. dissertation, Department of Psychology, Duke University, .1970. Crown, Barry M., "An Evaluation of Selected Cognitive and Social Dimensions on Poverty Intervention Project Participants." Ihrpublished Ph.D. dissertation, Department of Social Work, IFlorida State University, 1969. ' Fielder, John F., "The Relationship Between Intensity of JBelief'and Level of Cognitive Compelxity." Unpublished Ph.D.