v r A. CGRfiEPiEX QWERNETEC MGBEL GP ENTREPEESOML CBMEUNXCATION The“: for "ta Degree; .95 M. A. ‘ mcmm 3-? TE: UNIVERSITY Joseph; N. Cappeifa. i973 THESIS IIIIIIIIIIIIIIIIIIIIIIIIIIII ‘ w-. u” I'L'W.”"-t; f..\ I. u‘! ,--\ 3-- ‘1‘ . . c. o w. . . C ‘»"‘“' :3. . - " ‘ 4 ' ' r E; ... - - .11 ' . I 9‘! .. fifi3n‘. ."-~*I"': f: j 4" 2 .- . . j a... .4. - :- 4-.- .- a: . , . ‘ .\ ;" % .‘ll‘d". . E”, . ,.' -“. s . . _ .. - 1 '3‘- “5' .94 .4 44y ' 1 w w .9.» “if '— fl- fan-Ir? MSU LIBRARIES .‘IIIIIEIIIIII. RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. J 057:3 1&4 at)» o" - 1‘ _ my?" r ’4 {5 GH'?‘ ,— -\ w I’ . I r 1 J 1 .4‘ ABSTRACT A COMPLEX CYBERNETIC MODEL OF INTERPERSONAL COMMUNICATION BY Joseph N. Cappella A dynamic and cybernetically—oriented model of interpersonal com— munication is developed which takes into account both stability and. growth dimensions of interpersonal processes. It is argued that most system perspectives emphasize homeostatis and equilibrium rather than growth and change, and that such an emphasis is inimical to the success and longevity of any system, in particular interpersonal systems. A balance between stability and flexibility (the stable-positive case mathematically) is assumed to insure satisfaction and longevity in in- terpersonal communication systems. Because the methods for treating simultaneous stability and flexi— bility in cybernetic systems are not well-developed in the social sciences, the necessary conditions for stability-flexibility are derived. Given causal loop structures among variables in a model, it is found that the conditions for flexibility-stability can be readily applied. However, the analysis of the nature of interpersonal communication systems strongly implies the adoption of a cybernetic model. To accomplish both ends, a transformation is presented which permits cybernetic systems to be modeled in a linear causal framework. The model of interpersonal communication to which the above methods are applied begins with the assumption that individuals seek to coordinate their activities interpersonally and that this coordination necessitates the transfer of symbolic information. With these assumptions and an ex- tended version of the coorientation model, a set of eight interrelated propositions are generated. These constitute the heart of the model. Under certain assumptions, the set of propositions can be rewritten as a mathematical system of first-order, linear differential equations with constant coefficients. Applying the flexibility-stability conditions to these equations results in a pair of system hypotheses whose validity depends on the validity of the model as a whole. They are that the effects of communication will be to increase actual consensus and perceived con- sensus more in high than low satisfaction groups. There is a similar hypothesis for high and low predicted longevity groups. The hypotheses were tested by self—administered questionnaires on a self—selected sample of 32 married couples. The results indicated certain errors in one of the underlying propositions and hence, the hy- potheses failed. Upon correction of the model and reformulation of the derived hypotheses, it was found that the best fit for the data was ob- tained under conditions of overall stability rather than flexibility- stability. Consequently, some doubt is cast on the assumption of the importance of stability ang_flexibility in interpersonal communication systems. A COMPLEX CYBERNETIC MODEL OF INTERPERSONAL COMMUNICATION By Joseph N. Cappella A THESIS Submitted to Michigan State Univers ity in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1973 To my parents whose life—long encouragement has in no small way made this work possible. ii ACKNOWLEDGMENTS I wish to express my most sincere appreciation and intellectual in- debtedness to a number of people. First and fbremost the members of my committee, Dr. Gerald Miller, Dr. Vincent Farace, and Dr. Donald Cushman, provided me the freedom and encouragement to pursue what certainly must have appeared at first to be a somewhat tentative analysis. In parti- cular, I cannot fully detail the influence which Donald Cushman has exerted on my thinking and analysis of communication behavior in general. His support and sharp critical abilities have helped hone many of the ideas and their underpinnings presented in this thesis. In addition, many fruitful hours of brainstorming and mind—picking were spent with a colleague and friend, Peter Monge. Much of my initial thinking on dynamics of systems was spurred by him. Lastly, my wife, Elena, must be included. It was she who first interested me in communi— cation, and it was her insistence on relevance which keptthis topic from becoming more and more esoteric. Further, her many hours of typing and editing, and her careful reworking of much of the questionnaire with me cannot be easily repaid. iii TABLE OF CONTENTS Chapter Page I A CYBERNETIC MODEL OF GROWTH AND STABILITY IN INTERPERSONAL COMMUNICATION . . . . . . . . . . . . . . . . 3 A. The Problem of Morphostatis and Morphogenesis . . . . . . 3 B. Casting Cybernetic Relationships in Causal Frameworks . . 8 C. The Nature of Interpersonal Communication Systems . . . . 18 l. The antecedents to coordination . . . . . . . . . . . 21 2. The system state vectors . . . . . . . . . . . . . . . 28 D. Derivations from the Consensus Model of Inter- personal systems 0 O O C O O O O O O O I O O O O O O O 35 E. Summary and Implications for Theory Construction . . . . . ”5 II METHODS AND PROCEDURES . . . . . . . . . . . . . . . . . . . . #9 A. Testing Synthetic Deductive and Explanatory Deductive Theories . . . . . . . . . . . . . . . . . . #9 B. The Necessary Conditions for Coorientation . . . . . . . . 52 C. Procedural Rules as Objects of Coorientation . . . . . . . SH D. General Methods and Procedures . . . . . . . . . . . . . . 56 l. Pretesting . . . . . . . . . . . . . . . . . . . . . . 57 2. Sampling . . . . . . . . . . . . . . . . . . . . . . . 57 3. Administrative procedures . . . . . . . . . . . . . . 58 E. Instrumentation and Indices . . . . . . . . . . . . . . . 59 1. Measuring actual system consensus (X ) and perceived individual consensus (X snd X5) . . . . 59 2. Measuring A's and B's communication IX and X6) . . . 6M 3. Measuring satisfaction and prediction . . . . . . . . 65 F. Testing the Data 0 O O O O O O O O O O O O O O O O O O O O 66 iv Chapter Page III RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 A. Descriptive Statistics for the Individual Measures . . . 70 B. Correlations Descriptive of the Entire Sample . . . . . . 72 C. Data Display for the Satisfaction Groupings . . . . . . . 7n D. Data Display fbr the Prediction Groupings . . . . . . . . 76 IV INTERPRETATION OF RESULTS, CONCLUSIONS, AND IMPLICATIONS . . 79 A. Validity of the Sample and Instruments . . . . . . . . . 79 B. Testing the Underlying Propositions: Estimates of the aij . O O O O C O C O O O O O O O O O O O O O O O 81 C. Evaluating the Derived Hypotheses . . . . . . . . . . . 85 D. Explanation and Reformulation . . . . . . . . . . . . . . 88 E. Conclusions and Directions . . . . . . . . . . . . . . . 92 Table 10. 11. 12. 13. 14. 15. 16. 17. LIST OF TABLES Summary of Three Orders of Perceptions and Their Relation— ship to Coorientation Variables . . . . . . . . . . . . Intrapersonal and System Variables Redefined . . . . . . . The Sixteen Possible Situations Based on Different Communi- cation Styles . . . . . . . . . . . . . . . . . . . . . . Primary Format fOr Data Display and Testing . . . . . . . . Means and Standard Deviations for Satisfaction . . . . . . Means and Standard Deviations for Prediction of Longevity . Means and Standard Deviations for Communication . . . . . . Means and Standard Deviations for Consensus Measures . . . . Correlations among Communication and Dissensus Measures fbr the Entire Sample . . . . . . . . . . . . . . . . . . Estimates of aij from Total Sample Correlations . . . . . . Communication and Consensus Correlates of Marital Prediction and Adjustment 0 O O O O O O O O O O O O O O O O O O O O Satisfaction Groups for Total Sample with H and W Controlled Satisfaction Groups for Total Sample with H and W Randomized Satisfaction Groups for Extreme Scores with H and W controlled 0 C C O O O O I O O O O O I O O O O I O O O O Satisfaction Groups for Extreme Scores with H and W Ran domi z e d O O O O O O O O O O 0 O O 0 O O O O O O O O 0 Mean Satisfaction Scores for High and Low Groups of Tables 12, 13, 1”, and 15 O O I O O O I O O O 0 I O O O O O O 0 Prediction Groups for the Total Sample with H and W contrOIled O I O O O O C O O O O O O O O O O O O O 0 vi Page 28 3O 39 66 7O 71 71 71 72 73 73 75 75 75 76 76 77 Table Page 18. Prediction Groups for the Total Sample with H and W Randomized O O I I O O O O O O O O O O O O O O O O O O O O O 77 19. Prediction Groups for Extreme Scores with H and W controlled 0 O O O O C O O O O C O O O O O O O O O O O O O O 77 20. Prediction Groups for Extreme Scores with H and W Randomized O O 0 C O C O O O O C O O O O I I O O O O I O O O 78 2l. Mean Prediction Scores for High and Low Groups of Tables 17, 18, 19, and 20 O O O O C O O O O O 0 O O O O O O O O O O 78 22. Patterns of Sex Differences for Low Satisfaction Groups on a2” and a26, and as” and a36 . . . . . . . . . . . . . . 85 23. A Reformulated Set of Communication Styles . . . . . . . . . . 89 24. Expected Results for the Reformulated Flexibility- Stability cases 0 O C O O O O O O O O O O O O O O O O O O O 90 vii LIST OF FIGURES A Mutual Causal Analysis of Population Growth . . . . A Simple Mutual Causal Loop Involving Communication and Agreement 0 I O O O O O O C O O C O O O O I A Modified Mutual Causal Loop Involving Communication and a Discrepancy Variable . . . . . . . . . . . Range of Applicability of Propositions u' and 5' A Schematic Representation of Propositions l' - 8' A Schematic Representation of the Restricted Model . Page 1” 15 32 33 35 INTRODUCTION Within the developing communication sciences, the terminology, con— cepts, and approaches of systems theory and of one of its special cases, cybernetics, have become part and parcel of the intellectual tools of the practitioners of theory and research. However, even a cursory review of the attempts to apply systems approaches to the phenomenon of communication (see Monge, 1972) reveals either superficial applications involving little more than terminological translations or purely theoretic discussions which stop short of actual application. ,The primary purposes of this essay shall be to develop a modified cybernetic modeling technique, to apply the tech- nique to modeling long-term interpersonal communication situations, and finally to empirically test the hypotheses derived from this procedure. In accomplishing these ends, several other propositions shall be demon- strated which are pointed out below for the purposes of emphasis. It shall be shown that: l. Causal models are well-suited to the development of conditions of growth and stability. 2. Cybernetic relationships differ markedly from causal relationships. 3. The nature of interpersonal communication systems necessitates a cybernetic modeling approach. u. Under certain conditions, a cybernetic model can be transformed into a causal model. 5. The nature of interpersonal communication systems suggests an informa— tion-theoretic modeling approach. 6. The problems of infermation theory are circumvented through the coori- entational paradigm. 7. The necessary conditions for the application of the coorientational paradigm can be satisfied by choosing ”procedural rules" as the objects of coorientation. 8. A systemic hypothesis concerning interpersonal communication systems can be derived and tested, which is identical in form to the more usual reductionist two-variable hypotheses. With regard to fermat, Chapter I shall bear the burden of analysis and proof as it focuses on Propositions 1 through 6, and the derivation of the systemic hypotheses. Chapter II will focus on Proposition 7 and the methodology surrounding the test of the systemic hypotheses. Chapters III and IV will present results of the test and the interpretation of those results respectively. The final chapter will investigate implications of the earlier discussion and of the empirical results for theory construction, explanation, and the current understanding of interpersonal communication systems. CHAPTER I A CYBERNETIC MODEL OF GROWTH AND STABILITY IN INTERPERSONAL COMMUNICATION SYSTEMS The Problem gf_Morphostasis and Morphogenesis Perhaps the most unfortunate result of the adoption of various system approaches within the social sciences has been an overemphasis on permanence and preservation of the status quo, rather than on growth, change, and alter- ation. Monge (1972, p. 116) has reviewed the three most influential systems perspectives--cybernetics, structural-functionalism, and general systems theory-~as they have been applied to problems in human communication, and concluded that Any description of a communication system which accounts only_for structural preservation is insufficient. This emphasis on change is important because it permits the study of communication as a complex adaptive system rather than a static enduring structure. The structural-functional approach, which grew out of Merton's and Parson's writings, attempts to provide techniques for analyzing the mechanics within a system which maintain the goal-state of the system within a certain acceptable range. These procedures and methods were severely criticized by other sociologists such as Gessous (196 ) for failing to account for social change and societal development. Reactions to this approach, for example, by Coser (1956), have sought to emphasize the functional nature of apparently dysfunctional social conflict. The obvious point of these reactions to pure equilibrium analyses of society is that society continues to exist not only because a set of mechanisms maintain a certain goal-state despite disturbances, but also because societal systems change and alter themselves in reaction to disturbances. Within the interpersonal area, Jackson (1957) and Watzlawick, Beavin and Jackson (1967) have described family systems based directly on cybernetic principles and mechanisms for achieving and maintain— ing homeostasis. This approach also has an equilibrium emphasis in that it fails to account for change and alteration in healthy families. Speer (1970) in reviewing various theories of family systems, has argued that purely morphostatic or structural analyses, like Jackson's, are insufficient for understanding, predicting, or counseling family behavior. Rather, morpho- genetic or growth-oriented considerations must also guide the development of theory in family systems. Clearly, the problem of accounting for morphostasis as well as morpho- genesis in modeling social and social-psychological systems is not a new one. However, the few methods for dealing with both growth and stability simultaneously have shown persistent problems. One such method has been provided by Maruyama (1963). His approach applies only to relationships which can be assumed to be mutually causal and which form loop structures, so that no one variable can be labeled as the exogenous variable. The sign of the correlation between each pair of variables is established either empirically or theoretically. If the product of the correlation coefficients around a 100p is positive, the loop is deviation-amplifying; if negative, then it is deviation-counteracting. Number of People + (P) (l) \ + + Modernization (M) (6) Amount of + Garbage (G) (2) Migration to City + (c) (5) Bacteria + Number of Per Area > Diseases (B) (3) (D) (u) Figure l. A Mutual Causal Analysis of Population Growth For example, in Figure 1 the loop "population-————+ garbage -—-—r bacteria.————e disease-————a»population" is deviation—counteracting, so that an initial increase in population leads to an eventual decrease in population. However, Maruyama's most important insight concerns deviation- amplifying loops. In a system of several closed loops, unless at least one loop is deviation-amplifying the system described by the loops cannot grow or alter its behavior; such a system must be homeostatic. This conclusion is obvious since if all loops are deviation-counteracting, then agy_dis- turbance will be countered by a force whose direction is such as to reverse the effect of the disturbance. Although a detailed critique of Maruyama's methods need not concern us here, a discussion of two shortcomings will serve our interests. First, Maruyama fails to distinguish between types of deviation—amplifying and deviation-counteracting loops. This is crucial since both amplifying and counteracting loops can be stable or unstable. If we are interested in the conditions under which systems exhibit overall stability as well as growth, then the type of deviation amplification or counteraction of the loops cannot be ignored. From Blalock (1969, p. 82), the fOIlowing situa- tions indicate all the mathematically possible alternatives and provide the necessary distinctions: I. Deviation-amplifying A. Unstable positive feedback 1. With an initial increase leading to eventual explosion 2. With an initial decrease leading to eventual decay B. Stable positive feedback 1. With an initial increase leading to a new stability point at a higher value 2. With an initial decrease leading to a new stability point at a lower value II. Deviation~counteracting A. Unstable negative feedback 1. With an initial increase leading to eventual decay 2. With an initial decrease leading to eventual explosion B. Stable negative feedback 1. With an initial increase leading to restoration of initial value 2. With an initial decrease leading to restoration of initial value Second, under the assumption that the "first cybernetics" has dealt only With deviation—counteraction and control, Maruyama presumes to call his methods the "second cybernetics." This is unfortunate, not because Maruyama fails to handle growth in addition to control, but because his analysis is based On the assumption of mutual causal loop structures. According to Buckley (1967, p. 63), for a system to be cybernetic it must include: l. A goal parameter or referent signal set in a control center 2. Infbrmation on the current state or action of the system 3. A test or comparison between the referent signal and the current state of the system, resulting in 4. An error signal based on the comparison which determines the direction and magnitude of the 5. Action taken by a control mechanism. The mutual causal assumption satisfies 2 and 5 but certainly not 1, 3, or n. Thus, while attempting to take cybernetics a step further with an analysis of growth in addition to control, Maruyama's cybernetics has suppressed the goal-seeking behavior of truly cybernetic systems. This difference is more than a semantic quibble over the "real" meaning of cybernetics. If one wishes to model real systems which are goal-seeking and at the same time take advantage of the insights into growth processes that Maruyama provides fOr mutual causal systems, then one must account for this difference. The reason Maruyama was forced to the mutual causal assumption is signi- ficant. With the mutual causal condition, some rather straightforward con- ditions for deviation-amplification and deviation-counteraction could be derived. If all of the above conditions for a cybernetic relationship had to be satisfied, the conditions for deviation—amplification and deviation— counteraction are not simple deductions. However, if a set of conditions can be fbund fOr transforming cybernetic relationships into a form amenable to mutual causal analysis, then growth and stability conditions in goal- seeking systems would be a possibility. That is, if the phenomena to be UKKieled are purposive and goal-directed, then the logic best suited to thc>se phenomena is that of cybernetic systems as defined above. However, the most powerful logics are found in causal analyses. Thus, if cybernetic Peléationships can be transformed to fit the causal framework, then we can develop a modeling technique which has the deductive power of causal systems but the "fit" of cybernetic relationships to goal—directed, purposive phenomena. Casting Cybernetic Relationships in_Causal Frameworks In order to obtain relatively precise and analytic answers to questions concerning growth and change in complex systems, it is necessary to develop and employ mathematical formulations of the dynamics of the system under investigation. Without this step the higher-order deductions required con- cerning system dynamics would be impossible. However, since much of the theorizing and proposition generation of the social sciences is within the verbal realm only, it is necessary to investigate the strategies available for transforming sets of verbal pr0positions into systems of mathematical equations. Fortunately, Blalock (1969) has already made a significant step in this direction. The key problem (prior to the question of mathe- matization) is the degree of precision which verbal propositions manifest. If we are to be capable of mathematizing our verbal theories (or, in fact, to develop useful scientific theories of any fbrm), then our propositions must be analytically precise. Blalock provides several useful guidelines, but Zetterberg's (1965, pp. 69-72) criteria for verbal propositions are more definite. According to Zetterberg, every proposition must be stated such that it is clear that the proposition is l. Reversible or irreversible 2. Sequential or coextensive 3. Deterministic or stochastic H. Necessary or substitutible 5. Sufficient or contingent. Thus, every verbal proposition must be able to be characterized by five bin- ary choices if it is to manifest the precision necessary for mathematization. Now, if the propositions are to be cast in a mathematical fOrmulation of linear differential or difference equations, then we must assume that the relationship between the variables is sufficient and deterministic. Though such assumptions may not be valid with certain propositions, they must be made for the sake of the mathematical formulation. The hope is that the power of the mathematical formulation will far outweigh the error introduced in possible violations of the assumptions. In addition, the reader might wonder if the propositions mu§£_be cast into a linear system and if the equations must be differential (or difference) in form. The answer to both of these questions is definitely NO! However, the most significant problem with mathematical modeling is generally fOund in the assumptions which are made. The greater the number and extent of the as- sumptions, the more powerful the deductions and the lower the likelihood of meeting the assumptions empirically. Thus, by making a limited number of simple assumptions, the deductive power of our methods is decreased but errors due to unmet assumptions are also decreased. Furthermore, (l) the theory of linear equations is well-developed, well-known, and the simplest theory of equations available mathematically, (2) most mathematical modeling which is in its early stages adopts first—order (linear) approximations rather than higher—order approximations, and (3) the linear model is readily adapted to linear measurement (statistical) models (for example, path ana— lytic) for the purposes of testing and verification. Finally, the only reason differential or difference equations need to be employed is that the problem to be answered concerns the dypamics of the system, and hence the time rate of change of the variables. 10 Once the set of theoretically important variables has been determined and the relationships among them specified in terms of precise verbal prop- ositions indicating direction and sign, it is simple to transform these propositions into a set of linear differential assume that the relationships of the schematic from verbal propositions. tions in the fbrm: 1. dP/dt 2. dG/dt 3. dB/dt u. dD/dt 5. dC/dt 6. dM/dt where the variable i attributable to variable j. for which a12’ a13’ tions is: 1. dP/dt 2. dG/dt 3. dB/dt u. dD/dt 5. dC/dt 6. dM/dt all a21 31 ul a a61 P+ P+ P+ P+ P+ P+ a12 a22 a32 auz a52 a62 G+ G + G + G+ G + equations. For example, Figure l are the translations there are six Since variables, we write six equa- coefficients aij’ i # j, represent the amount of covariation in no link exists between variables j and i. 616’ a23’ a2u’ a25’ a26 Clearly, those aij will be zero This is the case with among others. The correct set of equa- a D + a C l“ 15 The coefficients aii are always included, since as Blalock (1969, p. 95) indicates: 11 [ai-J summarizes our ignorance of the details of the feedback process, and ... the term [aiixi] stands as a surrogate fbr other terms that could be brought into the equation explicitly in order to explain the feed- back process .... the notion that 'xl causes xl' might be used to substitute for a causal argument involving one or more feedback loops, as for example, xl-———> x2-———+ X3 -——+ xu-———+ x1. Of course, as our ability to explain the variation in x1 increases, then the value of aii will become smaller since the loops feeding back on xi will have become explicit. Once having set down the system of linear differential equations, there are basically four types of analysis that one can undertake (Simon, 1957, p.128); 1. An explicit solution of the equations can be obtained, giving the time path that any one variable would follow from some set of initial conditions. 2. The equilibrium positions can be determined and predictions made as to the behavior of the system at or near equilibrium. 3. The conditions for stability can be obtained and predictions made as to the susceptibility of the system to external and internal perturbations. u. Alterations in equilibrium and stability conditions can be determined as a function of changes in system constants and exogenous variables. This is known as the method of comparative statics. Since our main concern in this paper is with the dynamics of complex cyber- netic systems cast in causal formulations, options 1, 2, and 3 are of most interest. We have already argued that equilibrium in and of itself is not the most significant nor the most useful descriptor of complex systems; rath- er, the stability of those systems overall and their simultaneous capacity for growth are our major concerns. The necessary conditions for stability 12 of a set of n linear differential equations are presented by Blalock (1969, p. 107) to be:1 V: n aii < O (l) 4. i=1 (lai.” >> 0 if n is even (2)a 3 where flai-H is the deter- minate of %he system. "ah.“ < o if n is odd (2)b The ability or inability to satisfy these conditions can provide useful hypotheses about the system under analysis. We shall see how this works with our model. However, the stability conditions alone do not provide information as to the kind of stability: negative, indicating equilibrium or homeostasis, or positive, indicating growth and change. This situation is particularly unfbrtunate for we are searching fOr descriptors of systems which are stable but exhibit potentialities for growth; that is, "stable—positive" systems. The conditions under which the system manifests stable—positive, stable— negative, or unstable feedbacks can be determined if we were to take Simon's option 1 above and explicitly solve the set of linear differential equations to obtain their time paths. This is very simple in the case of a few vari- ables, but for several variables the situation is much like that for the 1Since these conditions are necessary but not sufficient, their satis— faction allows us to conclude nothing about the stability of the system. However, since an exhaustive list of necessary conditions constitutes sufficiency, we can at least conclude that satisfaction of the necessary conditions increases the probability of stability. Sufficient conditions are available (see W. J. Baumol, Economic Dynamics, New York: MacMillan Co., 1959, pp. 118-122), but are so complex as to be useless when one is dealing with algebraic rather than numerical solutions. Thus, as estimates of the aij are obtained, sufficient conditions can be applied. 13 sufficient conditions for stability; that is, analytic solutions are avail- able2 in principle fOr any set of n variables, but as n increases the prac- ticality decreases very rapidly. Once again, if numerical estimates of aij were available, explicit solutions of the time paths would be readily available (with computer aid) and, hence, the precise nature of the stabi- lity available also. Fortunately, all is not lost since we can once again specify at least necessary conditions for stable-positive feedback. Since well-defined con- ditions for system stability are available, the set of alternative condi- tions which result in overall system stability is established. Within this range of alternatives leading to stability, some or none of the 100ps will be deviation-amplifying. If Eg_loops are deviation-amplifying, then the system cannot be stable-positive overall since every alteration in the value of a variable will initiate a response which returns that variable to its initial value (stable-negative). Furthermore, as the number of deviation— amplifying loops increases while maintaining overall stability, the prob- ability of achieving stable-positive feedback also increases. Thus, the presence of deviation-amplifying loops is a necessary condition to insure stable-positive systems. It is not sufficient however, since even with overall stability the mere presence of deviation-amplifying 100ps could still result in a stable-negative condition. Having stipulated and derived the necessary conditions for stable-posi- tive mutual causal systems, we are now in a position to develop conditions *For example, see Coleman, "The Mathematical Study of Change" in Methodology 22 Social Research, H. M. Blalock and A. B. Blalock (New York: McGraw—Hill, 196871 1H for transforming cybernetic relationships into mutual causal relationships. Let us proceed inductively from the rather simple (and hypothetical) ex- ample represented schematically in Figure 2. In this example, perceived Amount of B's Persuasive Communication A's Perceived B's Perceived Agreement \ i / Agreement Amount of A's Persuasive Communication Figure 2. A Simple Mutual Causal Loop Involving Communication and Agreement. agreement is nothing more than how much A thinks B has similar cognitions to his own; similarly for B. According to the criteria for cybernetic re- lationships set down earlier, the loop is mutual causal and not cybernetic. However, suppose on theoretic or empirical grounds one argued that the nature of the phenomena represented by this loop was purposive and goal-directed rather than purely causal. For example, one might argue: the fact that B thinks he and A disagree on some topiC'does not necessarily imply that he will actively seek to persuade A of his position. Rather, the import- ance 2£_agreement gn_the topic to the successfu1 continuation of the re- lationship between B and A, in addition to the amount of disagreement, determine whether B will seek to persuade A. Hence, in Figure 3, a modi— fied loop is presented in which the amount of communication depends on the discrepancy between importance of agreement on some tOpic and the individual's perception of the amount of agreement. 15 Amount of B's Persuasive Communication A's Perceived Agreement K\\\\\\\\‘Bvs Perceived Agreement Relative to Importance /////////;.Re1ative to Importance of Agreement of Agreement Amount of A's Persuasive Communication Figure 3. A Modified Mutual Causal Loop Involving Communication and a Discrepancy Variable. In the modified case, the relationship between the variables is still causal, but there are several important differences between the case of Figure 3 and that of Figure 2: (l) The definition of the variable "per— ceived agreement" has become a discrepancy between some current state of the system (an individual's perception of agreement) and a goal state (the magnitude of required agreement). (2) The generation of persuasive com- munication depends not on the value of perceived agreement but on the magni- tude and direction of the difference between that value and the goal state. (3) Persuasive communication acts to control not merely the correlation between perceived agreement at time t and at time t + l, but also to con- trol the relationship between perceived agreement and the goal of required agreement. All of these differences together indicate that the modified causal loop of Figure 3 is cybernetic in its most important sense; its logic is isomorphic to a purposive and goal-directed empirical phenomena. BefOre setting down the principles fbr transforming cybernetic rela- tionships to mutual causal ones, we need to recognize an important point. The process that we have just gone through is quite the reverse of that normally carried out in theory construction. That is, beginning with a 16 causal model, we sought to transform it to a cybernetic one. Rather, one should begin with empirical and theoretic justification for the nature of the phenomena being modeled and then seek out a logic which best fits the conceptualization. The point of our example is that a phenomenon essentially goal-directed need not compromise that goal orientation for the sake of a causal logic. Rather, cybernetic relationships can be cast into a causal framework without too much loss of power when: 1. At least one variable in the loop is a "comparator variable" whose value is determined by either the ratio or difference of two other variables: one being the referent variable (or goal state), the other being the action variable (or current state) of the system. 2. The comparator variable is one variable in the causal loop so that its value at one point in time ultimately affects its value at a later point in time. 3. The value of the comparator variable, which shall be termed the error signal, determines the value of some control variable. u. The function of the control mechanism is to alter the error signal by increasing it if the loop is deviation—amplifying, or by de— creasing it if the loop is deviation—counteracting. This can be accomplished by altering the action signal, the referent signal or both. 5. The comparator variable and the control variable together con— stitute the control center of most cybernetic models. Now, when conditions are such that cybernetic relationships can be cast into mutual causal loops, the loops can be cast into mathematical formulations amenable to analysis for conditions of stability and growth. That is, since conditions for growth and stability can be established for mathematical systems, the above technique represents a method for analyzing growth and stability in cybernetic systems when the nature of the phenomena require a cybernetic logic. However, the method just developed has certain shortcomings. For one thing, the comparator variable can be changed by alteration of either 17 the action state or the goal state. By including both in one variable, it is impossible to determine which state changes and, hence, important information on the behavior of the system is lost. It is our hope that this loss of information will be far outweighed by the logical and ana- lytic power of the methods developed herein. A further shortcoming is the limitation on the type of growth and change with which we can deal. In his analysis of social change, Applebaum (1970, pp. 7-8) distinguishes three dimensions of change: magnitude or scale, time span, and effect on the changing unit. The magnitude of the change can be large or small and can be distinguished by the size and centrality of units affected, the proportion of the units affected, the degree of alteration involved, and the suddenness of the change. The duration of the change can be long or short term. And, most importantly, the effects of the change can be processual, which are merely alterations in the output of substructures, or structural, which involve alteration of the very relationships among substructures, or the disappearance of old substructures. Clearly, our methods have dealt only with processual change and the degree of alteration. In no sense has a comprehensive theory of change been presented and, in fact, our position rests on the assumption of continuing structural identity rather than structural change. Before applying these methods to interpersonal communication systems, it would be useful to summarize the ground we have covered. A set of nec— essary conditions for the overall stability of causal feedback loops as well as necessary conditions for stable-positive loops have been stipulated. These conditions include the satisfaction of certain mathematical conditions (equations (1), (2)a, and (2)b) on the set of linear differential equations 18 describing the system of causal interactions while_maintaining the maximum number of deviation—amplifying 100ps. In addition, the conditions for transforming cybernetic relationships into mutual causal relationships was established, thereby making possible the analysis of conditions for growth and stability in cybernetic systems. The Nature g£_Interpersonal Communication Systems The first step in developing an understanding of interpersonal communi- cation systems should be to distinguish among types of systems which consist of two or more people engaging in the transfer of symbolic information. In a certain broad sense, a family, a dating couple, the local Elks Club ex- ecutive committee, and an assembly line crew at General Motors could each be conceived to constitute an interpersonal communication system. However, at the current stage of development of communication theory, any attempt to include such diverse situations under one model would at best be very generic. Rather, a very precise and limited definition of an interpersonal communication system will be offered in order to construct a more specific, and, hopefully, more powerful model. Whenever two or more individuals are present to each other and in some sense interdependent, they will need to be capable of predicting one another's actions if they are to coordinate their activities and remain interdependent. One basis fer distinguishing among these interpersonal-interdependent sys- tems is in terms of the type and soyrce of information which the individuals use to make predictions about the behavior of the other. If the type and source of information is attainable gnly_through direct observational or symbolic interaction with the other, then this information shall be termed 19 "interpersonal." Clearly, information predictive of the other's behavior is often obtained indirectly (for example, through a third party) or on the basis of the other's membership in some class of individuals with accepted characteristics (role, stereotype, cultural types). Because of the indirect- ness of these sources of information, the directness of the communicative link with the other is removed and the commitment concomitant with that link is also absent. That is, if the source of information is indirect, coordina- tion among the individuals is possible but limited only to the types in which individuals from the same statistical classes would be interchangeable. If the source of information is provided directly by one person, but its nature is normative, conventional, or stereotypic, then, once again, the types of coordination possible are very limited. Miller, Nunnally, and Wackman (1971) suggest that three types of information are crucial when individuals interact: topic, person, and relationship statements. Topic statements refer to items of general interest but peripheral to the relationship between the individuals. Person statements refer to the self or to the other person in the relationship. Relationship statements refer to both individuals as a unit. Each of these infbrmation types can be normative, conventional, and stereotypic, or idiosyncratic and highly individuated, making possible dif- ferent levels of coordination. Adapting Miller, Nunnally, and Wackman's position somewhat, I contend that when topic, person, and relationship statements have a high (as opposed to low) disclosure potential3 and a 3Disclosure potential is defined as the "amount of information provided by a statement and/or the amount of infermation potentially elicited from another person by a statement" (Miller, Nunnally, and Wackman, 1971, p. 67). 20 high risku or uncertainty, then those statements are interpersonal. That is, when the task facing a pair (or more) of individuals is such that to coordinate activities on that task requires types and sources of informa- tion which have been labeled interpersonal above, then that situation will be called an interpersonal communication system regardless of whether the information transmitted is actually stereotypic or idiosyncratic. Thus, we have carved out for analysis as interpersonal communication systems those situations requiring a level of coordination on tasks attainable only through certain types of information transfer. This definition does not exclude from analysis those interpersonal situations which are incap- able of coordinating on these crucial tasks, but rather includes them as objects of analysis. There are three immediate implications of this orientation to inter— personal communication systems: (1) It is eminently a communication per- spective since it focuses on the origin and content of messages. (2) Because of the source and type of the messages required, it suggests that the nature of interpersonal communication relationships is highly idio- syncratic and individualized and, hence, manifests a high variability of interpersonal styles. (3) The necessary goal of overall coordination plus the variability of interpersonal styles suggest that the specific goals and styles of achieving them are themselves highly variable and individua- lized. If this last implication is valid, then the attempt to model an interpersonal communication system cannot impose goals from without, but “Risk is defined as "the degree to which responses elicited by a statement are predictable ... the range of responses afforded the lis- tener is minimal" (Miller, Nunnally, and Wackman, 1971, p. 70). 21 rather must discover the goals and incorporate that discovery into the characterization of the system. In a broader framework, Toulmin (1969) has argued that any action which is purposive and goal—directed cannot have categories imposed upon it without destroying the essential nature of the phenomenon. Thus, our analysis must incorporate the distinction between discovery and imposition. Furthermore, that which must be discovered are the goals of the in- terpersonal system. To remain interdependent the system must coordinate, but what to coordinate on and how to achieve that coordination is deter- mined by the individuals involved. Therefore, because of the goal—directed nature of interpersonal systems, a cybernetic model which does not impose goal-states would seem to preserve the nature of interpersonal communica- tion systems. That is, the isomorphism between the phenomena and the model would be upheld if the logic of the model did not violate the goal-directed- ness and choice of goals which seem characteristic of interpersonal communi- cation systems. The antecedents to coordination Coordination is nothing more than the ability of each individual in the system to behave so as to insure the solution of problems, the comple- tion of tasks, and the removal of exigencies facing the system. Such an ability can be understood by attributing to each individual a set of alternative behaviors which are physiologically, psychologically, and socially possible fer him, and by associating a probability with each alternative determined by the task—situation and by the need for coordin— ation. In such a conception, coordination is possible if and only if 22 each person can predict the most probable alternative of the other and can be assured that the other can predict his choice. Furthermore, these pre- dictions are possible only through the transfer of symbolic information since without such information accurate predictions at an interpersonal level would be highly improbable. If this analysis of the nature of in- terpersonal situations is correct, then accurate prediction at an inter- personal level would be impossible with information gathered through ob— servation of similar others. Allport (1962) has termed the development of correct predictions of the behavior of others the "complementarity of expectations." If person A does not know what to expect from B and cannot be sure what B expects from him, the chances of A and B coordinating actions at anything other than minimal tasks are highly unlikely. Thus, our first proposition for in- terpersonal systems becomes: Proposition 1: The greater the complementarity of expectations, the greater the ability to coordinate activities. Because the complementarity of expectations is conceptualized in terms of a set of alternatives and probabilities for those alternatives, an infor— mation-theoretic measurement model would seem most applicable. If suCh a model were adopted, two glaring difficulties would be immediately con- fronted: (l) the determination of the set of behavioral alternatives for A and B, and (2) the determination of the subjective probabilities fer each alternative. However, the coorientational paradigm of Newcomb (1953), Chaffee and McLeod (1968), Chaffee, McLeod and Guerrero (1969), Scheff (1967), and Laing, Philipson, and Lee (1965) allows us to cir— cumvent both of these problems by directly determining the accuracy of 23 predictions. As shall be shown below, the coorientational approach will allow us to retain our conception of interpersonal systems and to avoid the problems found with an infermation theoretic analysis. The remainder of this section shall be concerned with laying out and modifying the co- orientational approach to fit the nature of interpersonal systems as they have been conceived. Newcomb (1953) began the study of coorientation with his balance model of interpersonal systems. This model suggested that individuals simulta- neously oriented themselves not only to the other, but also to the other relative to some object, X, where X is anything to which the individual can refer. The key variables in Newcomb's model included (1) agreement or the similarity between person A's orientation to X and person B's orientation to X, and (2) perceived agreement or the similarity between A's (or B's) orientation to X and A's (or B's) perception of B's (or A's) orientation toward X. For example, if X is "where to spend our vacation" and if Mrs. A thinks "Visiting my mother" and Mr. A thinks "Fishing in the mountains," then Mr. and Mrs. A disagree. But if Mr. A thinks that Mrs. A thinks "Fishing in the mountains," then he perceives agreement. 0n the other hand, if Mrs. A predicts Mr. A. will say "Fishing in the mountains," then she perceives disagreement. Newcomb predicted that there would develop a strain toward symmetry in the event that the set of orientations (A to B), (B to A), (B to X), and (A to X) were in imbalance. Balance or imbalance of the dyad is determined by the products of the valences (+ or -) of the four orienta— tions, and is balanced if the product is positive and unbalanced if nega- tive. With the assumption of strain toward symmetry (balance), Newcomb 2“ predicts that in an unbalanced system one or both individuals will act so as to reintroduce balance. In this analysis, communication becomes one possible response to system imbalance and functions to re-establish bal- ance by changing the other's cognitions. That is, communication serves a persuasive function. Quoting an earlier article by Festinger (1950), Newcomb (1953, p. 399) is careful to hypothesize that it is "perceived symmetry, viewed as an independent variable, which is obviously a deter- minant of instigation to symmetry-directed communication" (emphasis added). Thus, it is an intrapersonal variable which predicts individual behavior and, by analogy, it must be interpersonal variables which predict system variables (as is the case with Proposition 1). Newcomb's classic work provides an overall framework for interpersonal communication systems (the coorientation paradigm) and provides our second proposition: Proposition 2: Given some pressure to agree, the greater the perceived agreement, the less the amount of persuasive communication. This proposition shall be modified slightly in the following section. Now, Chaffee and McLeod (1968) and Chaffee, McLeod, and Guerrero (1969) have a more general coorientation strategy. In addition to characterizing interpersonal systems by agreement and perceived agreement, they emphasize a second—order perception which is defined as one individual's prediction of the other's orientations. When this prediction is correct (as deter- mined by the other's actual orientation), the individual is accurate. For example, Mrs. A of our earlier situation was accurate in perceiving dis- agreement while Mr. A was inaccurate in perceiving agreement. Chaffee and McLeod recognized that communication could increase accuracy as well as increase agreement. In fact, a comparative study by 25 Wackman and Beatty (1971) showed greater increases in accuracy (rather than in agreement) as a result of communication. In other words, com- munication can function to increase the accuracy of A's predictions about B's most probable alternatives. Unfortunately, while recognizing this information function of communication, Chaffee and McLeod failed to pro- vide antecedent conditions to communication as infermation as Newcomb provided antecedent conditions for communication as persuasion. Actually such antecedents are unavailable within the coorientational strategy as presented thus far. A third order perception is needed. However, the emphasis on the interpersonal variable, accuracy, is crucial since it is a precise formulation of a second dimension of the complementarity of expectations and, hence, will be related to the system variable coordi- nation according to Proposition 1. The third-order perception necessary for develOping antecedent condi— tions for informative communication was developed and used by Laing, Philipson, and Lee (1965) and Scheff (1967), although for other purposes. It is defined as person A's prediction of person B's prediction of A's orientation to X. A person's third-order perception in combination with the other's second-order perception is termed realization, and is defined as one individual's ability to predict the other's accuracy. In other words, if A can predict What B will predict about A, then A realizes whether or not B is accurate. For example, if Mrs. A says "Mr. A thinks that I want to go fishing in the mountains," then she realizes what Mr. A thinks because she has correctly predicted his second-order perception. Conversely, if she says "Mr. A thinks that I want to visit my mother," then she fails to realize what Mr. A thinks because she has ircorrectly predicted his second-order perception. 26 As Scheff points out, this spiraling reciprocity can go on indefinitely although certain logical and semantic problems arise with fourth and higher order perceptions. However, these higher orders are not necessary since linking the third-order perception to the first-order perception yields perceived accuracy, which then becomes an antecedent condition to informa— riygcommunication.5 That is, if A perceives B to be inaccurate, then A will generate symbolic information whose purpose is to inform B of A's true orientation. Back to Mrs. A: If she says "Mr. A thinks that I want to go fishing in the mountains," then she perceives Mr. A. to be inaccurate be— cause she already knows what she actually thinks. Now, because she feels that Mr. A does not know what she thinks, she will inform him of her wishes. Hence, our third proposition becomes: Proposition 3: Given the pressure to be accurate, the greater the per- ceived accuracy, the less the informative communication. Like Proposition 2, this one will also be modified slightly in the next section. Thus, this highest order perception is necessary since it provides, on the one hand, one of the components for an antecedent variable to informa— tive communication, and on the other, an additional dimension of the com- plementarity of expectations which should be a powerful indicator of the ability to coordinate. This latter claim needs some explanation. In re- viewing the development of the coorientational paradigm, two classes of variables, intrapersonal and interpersonal, were separated. The former should be predictive of individual behavior within the dyad (Propositions 5Note the similarity to perceived agreement as an antecedent condition to persuasive communication. 27 2 and 3) and the latter predictive of system variables (Proposition 1). The interpersonal variables agreement, accuracy, and realization were suggested as dimensions of the complementarity of expectations. Why are these dimen- sions important in predicting coordination? Perhaps an analogy can help argue the claim. Suppose that a computer programmed to play chess is competing with a human opponent. In order for the human to win, he must correctly predict the next move or series of moves of his machine opponent and foil the machine's predictions of his own moves. Without knowledge of the programming instructions, including evaluation criteria, depth of search, etc., the human will probably be very unsuccess- ful. However, if he has learned and stored the computer's programmed in- structions, his chances improve. The reason is that if the criteria for choice among alternative moves are the same for the human and the computer (that is, agreement) if the human knows this (that is, accuracy), then it is a simple task to predict the computer's next move. Furthermore, if the human knows what the computer's prediction for his criteria for choice among alternatives is (that is, realization), then the human can foil the computer's predictions simply by violating his own criteria. Thus, in this competitive situation, success is contingent upon agreement, accuracy, and realization. In a cooperative situation, success is once again con- tingent on agreement, accuracy, and realization except that realization allows the fulfillment, rather than the frustration, of the other's ex- pectations of self. In summary, then, the characteristics of interpersonal communication systems must include: 1. Agreement, denoted AGR 2. Accuracy fer A and B, denoted ACCA and ACCB respectively 28 3. Eealization for A and B, denoted RELA and RELB respectively H. Perceived Agreement for A and B, denoted PAGRA and PAGRB respectively 5. Perceived Accuracy for A and B, denoted PACCA and PACCB, respectively These nine values are derived from three orders of perception: l. let-order perceptions of X for A and B, denoted P A and P B respectively 1 l 2. 2pd—order perceptions of X for A and B, denoted P2A and P2B respectively 3. 3rd—order perceptions of X for A and B, denoted P3A and P3B respectively. Table 1 summarizes these definitions and their relationships. Table 1. Summary of Three Orders of Perceptions and Their Relationship to Coorientation Variables. AGR = P A - P B l l ACCA = P2A - PlB ACCB = P2B - PlA RELA = P3A — P2B RELB = P3B - P2A PAGRA = P2A - PlA PAGRB = P2B - PlB PACCA = P A - P A PACCB = P B - P B 3 l 3 l The system state vectors Earlier in this chapter, it was argued that while the overall goal of coordination characterized interpersonal communication systems, the specific objects of coordination were determined by the individual system and, hence, needed to be discovered rather than imposed. Juxtaposing this claim with the pr0posed relationship between coordination and agreement, accuracy, and realization, the question becomes: "What goals has the in- terpersonal system set for itself with regard to the amount of agreement, accuracy, and realization required for coordination?" If our analysis of the nature of interpersonal communication systems is correct, then any 29 attempt to establish these goals §_priori by an outside observer must be doomed to failure. In order to avoid this problem, three other perceptions must be introduced: 1. Perceptions of the agreement required on X for A and B, denoted GlA and GlB respectively 2. Perceptions of the accuracy required on X for A and B, denoted G2A and G2B respectively 3. Perceptions of the realization required on X fOr A and B, denoted G A and G B respectively. 3 3 With these goal-perceptions characterizing each system, the intra- personal and system variables defined earlier can be redefined so as to develop cybernetic relationships from our causal propositions. The intra- personal variables change from PAGR and PACC to: l. Perceived agreement relative to the amount of required agreement, or $2 = Gl - PAGR, and 2. Perceived accuracy relative to the amount of required accuracy, or ¢ = G - PACC. 2 2 The system variables change from AGR, ACC, and REL to 1. Agreement relative to required agreement, or Y1 = (GlA - AGR, GlB - AGR) 2. Accuracy relative to required accuracy, or 4/2 = (G2A - ACCA, G2B - ACCB) 3. Realization relative to required realization, or Y3 = (63A - RELA, G3B - RELB). Table 2 presents a summary of these new system and intrapersonal variables. It should be noted that \r’, \k;, and \i/é are two dimensional vec- tors since we have implicitly assumed a two-person system. Two advantages of a vector notation are that the number of persons in the system can be incorporated directly into the dimensionality of the vector, and that maximal infbrmation can be retained with a relatively simple notation. 30 Table 2. Intrapersonal and System Variables Redefined. Intrapersonal ¢1A = GlA - PAGRA ¢1B = GlB - PAGRB ¢2A = G2A - PACCA ¢2B = G2B - PACCB System ‘Fi = (GlA - AGR, G13 - AGR) ‘Vfi = (62A - ACCA, G2B - ACCB) \y5 = (G3A - RBLA, 93B - RELB) Because vector notation may be unfamiliar, making interpretation difficult, Appendix A presents a more thorough explanation and a worked example using this notation. It should be now be obvious that the scalar l/rf12, derived from the state vector = if: + \Yé +'fl/é, is a measure of the system's'bomplemen- tarity of expectations." The greater the value of 14\r12, the greater the complementarity of expectations. In order to avoid future possible con- fusion, we shall refer to l/FVY2 as the degree of actual system consensus. Similarly, we shall refer to ¢l and ¢2 as perceived individual agreement and perceived individual accuracy respectively; or, when referring to both simultaneously, simply as perceived individual consensus. Although this multiplies the terminology, it will help in distinguishing these variables from earlier ones. Rewriting Propositions l, 2, and 3 to include the redefined variables, we have: Proposition 1': The greater the degree of actual system consensus, the greater the degree of coordination. Proposition 2': The greater the degree of perceived individual agree- ment, the less the amount of persuasive communication. Proposition 3': The greater the amount of perceived individual accuracy, the less the amount of informative communication. 31 In discussing the work of Newcomb, Chaffee and McLeod, and Wackman and Beatty, no propositions about the effects of the amount of communica— tion on intrapersonal or system variables were offered. This was done not because those authors failed to provide such propositions, but rather be- cause this author maintains that the sheer amount of communication (meas- ured in terms of number of words, length of time talking, length of associa— tion, or whatever) should not necessarily result in greater perceived in- dividual consensus or greater actual system consensus. We are not arguing that accurate information transfer or effective persuasive styles could not accomplish these ends, but rather that such a crude measure of communica— tion as the amount of this or that symbolic form bears 22_re1ationship to the intrapersonal or system variables defined above. There is some em- pirical evidence for this contention. A study by Udry, Nelson, and Nelson (1961) fOund no significant correlations between various measures of the amount of communication and agreement, understanding, or perceived agreement. As a result, the fOllowing propositions will relate amount of communi— cation to the variables of interest: PrOposition u': The greater the amount of communication generated within the system, the greater or lesser the amount of actual system consensus. Proposition 5': The greater the amount of persuasive communication, the greater or lesser the amount of perceived in- dividual agreement.6 6These propositions imply either a negative linear or a positive linear relationship. This linear relationship cannot reasonably hold over the entire range in the negative case. If it did, then it would be possible for an essentially negligible amount of communication to produce large amounts of perceived and actual consensus. This is completely at odds with all that we have argued thus far. To circumvent this problem, we assume that the negative relationship is linear only after a certain threshold has been reached and that these propositions apply only within that range. Figure u should make this clear. 32 Measures of System and Perceived "——Range of Linearity-——+ Consensus //////' i‘\\\\“\\\\\\\\\\\\ Amount of Communication “-_- Figure 4. Range of Applicability of Propositions H' and 5'. Proposition 6': The greater the amount of informative communication, the greater or lesser the amount of perceived in— dividual accuracy. Essentially, these three propositions treat relationship between "amount of communication" and the various consensus states as variable and, as we shall see, predictions about the behavior of the system will be determined from various values of this variable relationship. A final proposition will close our system of variables. The ability to actually solve a problem, complete a task, respond to an exigence, or, in general, to physically coordinate activities should produce positive changes in perceived individual consensus. Hence, we have: Proposition 7': The greater the degree of coordination, the greater the degree of perceived individual agreement. Proposition 8': The greater the degree of coordination, the greater the degree of perceived individual accuracy. Assuming these propositions to be causal, a schematic model linking the ten variables with one another according to the propositions can be developed, as in Figure 5. Furthermore, assuming the propositions to be linear in addition to sequential and sufficient, the schematic model can be represented in an isomorphic set of ten linear differential 33 A's Perceived B's Informative __+ Accuracy (X7) ‘ Communication (X10) __ A's Informative B's Perceived Communication (X8) , Accuracy (X9) Degree of Degree of Actual Coordination AL_ System Consensus , (x1) (x2) A's Persuasive B's Perceived Communication (X9) ; Agreement (X5) A's Perceived B's Peisuasive L_, Agreement (X3) . Communication (X5) Figure 5. A Schematic Representation of Propositions 1' - 8'. equations according to the techniques suggested earlier in this chapter. The set of equations are: l. Xm/dt = allxl + al2X2 2. dX2dt = a22X2 + azuxu + a26x6 + a28X8 + a210X10 3. dX3/dt = ale1 + a33X3 + a36X6 U. qu/dt = auBX3 + auuxu 5. dXS/dt = a51Xl + asuxu + aSSXS 6. dXs/dt = aGSXS + a66x6 7. dX7/dt = a7lX + a77X7 + a7lOXlO 8. dXB/dt = a87X7 + a88x8 9. ng/dt = aglxl + ag8x8 + aggxg 10' dio/dt = alogxg + alOlOXlO (3) (u) (5) (6) (7) (8) (9) (10) (11) (12) 34 This set represents a first-order approximation of the time rate of change of interpersonal communication systems as we have defined and analyzed them, and it shall be our object of analysis for the remainder of the chapter. What remains to be shown is that this model is a cybernetic represen- tation of a set of what appear to be causal relationships. The reasons are threefold: (l) the redefinitions of the intrapersonal and system variables, (2) the control fUnction that is played by communication, and (3) the relationships between (a) perceived individual agreement and per- suasive communication and (b) perceived individual accuracy and informa- tive communication. First, the definitions of perceived individual con— sensus and actual system consensus differ from earlier definitions of conceptually similar variables in that they are expressed as differences from some expected or desired value, the goal state Gi' This value is the "referent signal" of most cybernetic analyses and is the value against which perceived levels of consensus are compared. Second, because the relationship between communication and perceived individual consensus and O 7 I 0 actual system consensus has been left variable, that relationship_con— trols the overall time rate of change of the interpersonal system. This relationship becomes the "control mechanism" of most cybernetic systems except that the mechanism need not be homeostatically oriented but can be deviation—oriented-—a significant departure from the emphasis on control in cybernetics. Third, because perceived individual agreement and per- ceived individual accuracy have been defined as difference variables, 7 O O O O O O 0 If the relationship is pOSitive, then discrepanCies from goal states are decreased; if negative, then discrepancies are increased. 35 their values are discrepancies or errors from the goal states; and it is these errors which determine the generation of communication. Thus, these two intrapersonal variables serve the "comparator function" of most cy- bernetic systems which, in turn, activates the control mechanism, per— suasive and informative communication. TherefOre, the conditions for transfbrming cybernetic relationships into mutual causal ones have been satisfied and the essentially purposive and goal—directed phenomenon-- interpersonal communication-- has been modeled with a cybernetic logic adapted to fit causal frameworks. Derivations from the Consensus Model 2§_Interpersonal Systems In this section we shall analyze the conditions for positive-stable feedback fer our cybernetic model. ~This shall be accomplished according to the techniques discussed in the second section of this chapter. In order to make our already complicated analysis a bit more tractable, we shall restrict the mathematical analysis to only the first six variables of equations (3) through (12) of the previous section. This restriction modifies the schematic representation of the model as in Figure 6, but it does not in any way invalidate the model since A's perceived individual +/- B's Amount of A's Perceived Consensus (X ) 4 3 ‘ o 0 Communication (X6) + +/_ Degree of + Actual System Coordination (X1) 6—----Consensus (X2) J /< "’ A's Amount of —% B's Perceived Consensus (X ) o o +/"' 5 Communication (Xu) Figure 6. A Schematic Representation of the Restricted Model. 36 agreement and A's perceived individual accuracy can be summated to A's perceived consensus; this can also be done for B and fer persuasive and infermative communication. Such a summary reduces the number of variables and equations to: l. 2. Xm/dt = alle + al2X2 dX2/dt = a22x2 + aQuXu + a26x6 an/dt = a3le + a33X3 + a36X6 qu/dt = auax3 + auu u dXS/dt = asuxu + aSle + aSSXS dXS/dt = a65X5 + a66 6 Thus, the necessary conditions for stability become 5:1 aii < O and (l) "aiju ‘>' 0 since n = 6. (2a) UnfOrtunately, there is little else that can be done with condition (1) except to hope that it is satisfied. This is accomplished if each aii is less than zero or if some combination of a.. terms less than zero is 11 greater in absolute value than the remaining aii terms. Otherwise, we can h0pe that most of the variability in any one variable is already ex— plained by the variability in other real variables so that each aii is essentially zero. This latter situation seems unreasonable given the present state of the art in the social Sciences. 80, in order to pro- ceed let us assume that each aii is less than zero. 37 Condition (2a) can be investigated by evaluating the determinant a.. through standard techniques.8 are presented in Appendix B. The results are: ”ain = alla22a33auua55a66 aiia22(a43a35a65a5u) 633a4u(a12a26365a51) a55a66(ai2a2uau3a31) ’ ai2azuau3336a65a51 ' ai2a2eaesa54ausasi which must be greater than zero purposes of later reference let 1' aiia22a3334ua55a66 = t1 2' ’ aiia22(au3a36a65a5u) = 3' ' a33au4(ai2a26a65a51) = ”' ' a55a66(ai2azuau3a31) = 5' ' ai2a2uau3a36a65a51 = t 6' ‘ ai2a26aesasuau3a31 = t for overall system stability. us label For the We also see that because we assumed that each aii is less than zero, t must be greater than zero and al 1822’ assauu’ ass ass are each greater than zero. What remains to be determined, then, are the signs of the 8Marcus, M. and Minc, H. Co., New York, 1965, Chapter 2. Introduction £2_Linear Algebra, MacMillan 1] The details of the technique for this case 38 three terms in parentheses in t t3, and t and the signs of the two six- 2’ H element terms, t5 and t6. For the purposes of interpretation, it should be noted that in Figure 6 there are three four—variable loops corresponding to terms t , and t”, 2’ '3 two six-variable loops corresponding to t5 and t6. Each of the six vari- ables appears in four of the loops. Hence, the effects of any variable on itself at a later time is due to the combined effects of four different loops. The overall effect of any variable on itself is determined by the sum of the products of the correlation coefficients around the loops in which the variable appears. For example, X1 appears in t t t5, and 39 ”9 t6 but not in t2: t6: Xl—-* X3-—’ Xq—b X5—4 X6—-9 X? X1 125: Xf—* X5—’ Xé—’ X? xii—4‘ X7 X1 and the parenthetical terms of t ' X-—+ X-—4 X-—+ X-—+ X H' l 3 4 2 l t : X-—~’X-v-X--*X-—*'X 3 l 5 6 2 l but not 1=2= Xr’ X? X? x3_—’Xu. Obviously, Maruyama was not far afield in his analysis. It is also worth noting that condition (2a) would be greatly simplified if either all’ a33, a55 all were equal to zero, or if a negligibly small. In either case only t5 and t6 would remain. Thus, it appears valuable to have variables whose values are influenced by as few 22, an”, a66 all equalled zero or were other variables possible, which is equivalent to saying that the stability conditions will be greatly simplfied when some obviously sufficient re- lationships are included. 39 The immediate question, however, is the conditions under which “ain will be greater than zero. As yet, we have not applied the proposed sign relationships on a 51, and al2’ and we need not since sta— 31’ ans’ ass’ a bility conditions can be develOped without this constraint. However, one of the purposes of this analysis is to take advantage of "known" relation— ships to deduce other hypotheses representative of a whole system of inter— relationships. That is, certain assumptions about the internal relation- ships among variables (the propositions) are made in order to derive hy- potheses about overall system characteristics. So, assuming the proposed relationships on a 51, and a hold in Figure 6, we are 31’ aus’ ass’ a 12’ left to specify only four other signs: , and a Since asu’ a36’ a2u 26' each of these four terms can take on only two values, + or —, then there are 24 or sixteen possible situations detailed in Table 3. Table 3. The Sixteen Possible Situations Based on Different Communication Styles. 1 2 3a3b u 5 6a6b 7a7b 8a8b 9a9b lOalOb 5'24 + - - + - + + + - + + - - - + - as” + - - + + - — + + + - - + - — + a36 + — + - + — + - + + - - - + + _ a26 + - + - - + + + + - - +' - - — + Itshould be noted that of the sixteen possibilities, only ten are truly different. Six are degenerate due to the interchangeability of person A and person B. In turn, this interchangeability implies that a65 = a”3 and a31 = aSl' no Now if we assume that the proposed relationships hold, and 1. if 26 > 0, then "aij" > 0 if and only if a2u’ asu’ a36’ a It1+tu+t3'>'t2+t5+t6'; 2. if < 0, then "aij" > 0 if and only if 324’ asu’ a36’ a26 |tl])[t2 + t3 + tn + t5 + t6! ; 3a and 3b. 1f 1:) o and a 36( 0, then “51in > 0 1f asu’ a2 26’ a and only if Itl + t2 + t3 + t5 + 126' > ”4' ; '4. if > O, a < 0, then "Eli—j") 0 if and only asu’ a36 2n; a26 if |’cl + t5 + t6l)|t2 + 1:3 + tu' ; 5. if as“, a36 < O, a > 0, then Ilaijll>0 if and only 2u’ a26 . + . 1f Itl t3 + t” + t5 +-t6|)|t2| , 6a and 6b. if > o, a5u< 0, then Ham") 0 if a2u’ a26’ ass and only if ltl + t2 + t3 + t” + 1:6. > Itsl 3 7a and 7b. if a2u< 0, a26, a36, as”) 0, then llaijll> 0 if and only if [1:1 + t3 + t6|>|t2 + t1+ + tSI ; 8a and 8b. if a2” > o, < 0, then Ilaijl ) o if a26’ ase’ asu and only if |tl + t3 + t6| ) |t2 + tn + tsl ; 9a and 9b. if < o, and as”) 0, then llaij|l> o a2u’ a36’ a26 if and only if |tl + t2 + t5] > 'ts + t” + t6! 3 10a and 10b. if a , a36 ) O and a II > 0 2n su’ a2e if and only if [t1 + t2 + tuI>‘t3 + t5 + t6| . < 0, then "ai L_.Io Before attempting to provide some specific interpretation to this mass of results, some general remarks on the nature of aij can be made. If terms t through t l were positive, then certainly aij O and 6 Ml condition (2a) would be satisfied. However, to meet the condition in this fashion, all the loops would have to be deviation—counteracting (that is, negative, since each is multiplied by a minus) and, hence, no "growth" or deviation-amplifying loops would be present. Thus, the system would be stable-negative or homeostatic overall. This result is general and im- plies that growth and stability are competitive processes—-an intuitively obvious point now made rigorous. Blalock (1969, p. 119) supports the above argument indirectly when he indicates the general stability con- dition for a single k-variable loop: + (—l)k-l alla22°"akk a2la32au3"'akk-lalk. If k is even the above term must be greater than zero (condition (2a)), k-l . but (-1) W111 be odd so that a2la32...a1k must be less than zero or deviation-counteracting. If k is odd the above term must be less than zero, but (-l)k-l = 1 so that a a ...a 21 32 must again be less than zero lk or deviation-counteracting.9 The conclusion, both for the general case and for our particular example, is that to simultaneously satisfy con- ditions fer both growth and stability will require not only the signs of the products representing the causal loops but also the relative magnitude of those products. In interpreting the results presented in Table 3 we find that cer— tain of the situations need not be considered because they are empirically implausible; they include situations 5, 6, 8, and 10. These situations are implausible because each posits for A or B or both that communication 9It is important to note that this analysis assumes that I a a a a 3.11322 ... nnl <.| 21 32 "' lnl ' 1+2 can simultaneously increase discrepancy values for perceived individual consensus while decreasing discrepancy values for actual system consensus. While it is possible to conjure examples of these situations, it would be difficult to defend any of them as representative of persistent and long- term communication styles within interpersonal relationships. Thus, al- though predicted through the logic of enumeration, these situations must be rejected on empirical grounds. Situations 4, 7, and 9, on the other hand, represent empirical situa- tions in which A or B or both adOpt communication styles which reduce dis- crepancies in perceived individual consensus but amplify discrepancy values for actual system consensus. These situations are akin to interpersonal relationships which attempt to "keep the peace" rather than resolve per— ceived disagreements and misunderstandings. Mathematically, each of these situations can achieve simultaneous growth and stability if the necessary conditions indicated earlier are satisfied. Once again, though, situa~ tions 4, 7, and 9 represent unusual communication styles and, as a result are difficult to exemplify or imagine. The more common situations are those in which communication styles act similarly on both actual and per— ceived discrepancies. These are represented by situations 1, 2, and 3; let us examine each in turn. Situation 2 is the most readily interpretable. If the communication styles of A and B always increase discrepancy values for perceived in— dividual consensus, then the system can achieve an overall stability only if the magnitude of tl is greater than the sum of the magnitudes of the other five loops--a highly improbable state of affairs. Because of the improbability of stability, this system could exhibit high levels of #3 coordination and consensus if the direction of the initial kick was toward more coordination or greater intrapersonal or system consensus; that is, the system is growth-oriented with five deviation—amplifying loops. How— ever, because of its instability, this system is highly susceptible to decay and disintegration.10 Situation 3 represents the case in which only one of the participants adopts a communication style which seeks to reduce discrepancies. It should be noted that this need not be the same individual from time unit to time unit since A and B are interchangeable. What is of particular interest in situation 3 is the high probability of stability due to the four deviation-counteracting loops; and this results from a relationship in which only one of the participants seeks to reduce discrepancies. Of course, as a result of the high probability of stability, the number of deviation-amplifying loops is one and probability of growth minimal. Hence, situation 3 is the direct opposite of situation 2, exhibiting high probability of stability rather than growth. Situation 1 can be viewed as the optimal system state somewhere between situations 2 and 3. In this case, all communication acts to reduce discrepancy values for perceived individual consensus and actual system consensus. The probability of achieving overall stability is greater than that of situation 2 but less than that of situation 3, while the chances for a stable-positive system (growth-oriented) is less than that of situation 2 but greater than situation 3. Thus, we are confronted once again with the competitive nature of growth and stability processes 0In the present case, decay and disintegration mean very low levels of coordination and little intrapersonal or system consensus. at: and the obviously sensitive balance which must be maintained between the two if a system is to survive the instabilities of growth and decay and escape the rigidity of unaltering homeostatis. Having completed the analysis and interpretation, we can now draw some general conclusions and state a specific hypothesis for this model of interpersonal communication systems. First, the mathematically astute reader may be unsatisfied with the rigor with which the necessary condi- tions for stable-positive feedback were developed and applied. As was pointed out in the second section of this chapter, more precise condi— tions could be used if numerical estimates of the aij coefficients were available. Given the rather rough and intuitive state of verbal proposi- tions about interpersonal communication systems, such estimates are not available and if they were probably ought not be trusted. Hence, we have attempted to compromise tight logical requirements with empirical con- straints. Bailey (1971, p. 69) upholds this approach to theory building, concluding his discussion of empirical versus logical requirements with the maxim: "Surely imperfect theory is better than no theory at all." Second, given numerical estimates of aij’ a first-order approxima- tion of the degree of stability of the system can be obtained by deter— mining how much greater than or less than zero "ain is and how much greater than or less than 0 2::aii is. This approximation can then be used as a predictor of the susceptibility of the relationship to rigidity (no growth) or to disintegration (no stability). Third, throughout this chapter, there has been an implicit assumption that all systems must establish and maintain mechanisms whiCh insure over- all stability as well as the capability of adaptation to the random dis- turbances which impinge on the system. Many systems' theorists have argued us this position (for example, Buckley (1967). Furthermore, Speer's (1970) critique and review of theory in interpersonal situations intimated that, both in therapy for disturbed families and in prescriptions for increasing the longevity and success of "normal" families, growth as well as stability ought be emphasized. Granting the (untestable) postulate, Proposition 9': The longevity and success of complex systems depends on the maintenance of mechanisms for overall stability and mechanisms insuring the capability of growth, the following hypothesis on communication styles can be derived: Hypothesis: The greater the correlation between the amount of com- munication and perceived individual consensus, and the greater the correlation between the amount of communica— tion and actual system consensus for both A and B, the greater the probability of system longevity and success. This result is basically a statement of the conditions of situation 2 and can be viewed as a hypothesis relating a certain communication style (as measured by the correlation coefficients) to system longevity and success. In the next chapter, we turn to questions involving the test of this hypothesis. Summary and Implications fer Theory Construction Because of the length and complexity of the foregoing discussion, it might be wise to summarize the ground that has been covered thus far and to make explicit the general implications of the techniques developed. To summarize the major developments in one sentence: Given a set of variables related in causal loop structures, then the methods of this chapter can be used to derive communication hypotheses dependent upon the particular system of variables and their structural characteristics. That is, system hypotheses can be derived and tested with communication as the system parameter of primary interest. ”6 This method for developing such hypotheses can be summarized as follows: 1. A set of verbal pr0positions is developed either through a review of pertinent literature, or through reasoning to a set of plausible postulates. This set must be closed, assumed to be causal, and specific enough to indicate the direction of relationship between variables in all propositions. 2. The propositions are transtrmed into a schematic representation (such as that of Figure 6, page 35) so that the number and inter— connection of the loops can be clearly identified. 3. From the schematic diagram, the rate of change of any one variable (say X) is assumed to vary linearly and additively with all other variables which have causal arrows pointing to X. The rate of change of each Xi, dX1/dt can then be written as a first- -order, linear differential equation with constant coefficients by equat- ing each Xm/dt with a linear sum of all other variables which have arrows pointing to Xi- u. Evaluate the determinant of coefficients, Haijfl , fer the set of conditions which lead to stability, and within each condition determine the number of deviation-amplifying loops. 5a. If the theorist assumes that stability alone is the fundamental necessary condition for system success, satisfaction, or longevity, then the relationship between the stability conditions and the output measures becomes the system hypothesis to be tested. 5b. If it is assumed that an Optimal balance between flexibility and stability is the fundamental necessary condition for system success, satisfaction, or longevity, then the relationship between the flexibility-stability conditions and the output measures becomes the system hypothesis to be tested. If it can be also shown or assumed that the relationship between the amounts of communication and other dependent variables is itself variable across output levels (as in Propositions u', 5', and 6'), then communi- cation becomes the parameter which determines the stability or stability- flexibility characteristics of the system. With this added assumption, the derived, testable hypothesis will always relate communication as the parameter of stability or flexibility-stability to the output meas- ures of satisfaction, successfulness, or longevity. M7 The technique described above has certain advantages over the more common approaches to theory construction which characterize communication study. First, the hypotheses generated are clearly systemic hypotheses in that they depend upon the §e£_of verbal propositions and the inter- relation among the members of the set. If the causal arrows of Figure 6 or the signs associated with those arrows were altered, the same de- rived hypotheses could not be obtained. That is, the hypotheses are derived about the system described by the propositions and their inter- relation. In this way, a system's analysis is much more than a purely theoretic analysis with no pragmatic import and cannot be subject to the usual criticism that the analysis is a mere verbal translation of standard techniques. Second, the derived hypotheses cannot be deduced through the loose syllogistic reasoning (Zetterberg, 1965; Costner and Leik, 1964) typically used for making deductions in social theory. Rather, the de- rived hypotheses depend upon first, the mathematization of the verbal theory and, then, upon certain powerful theorems from the theory of simultaneous, first-order linear differential equations, Hence, the results cannot be obtained through simpler deductive procedures. Third, the technique developed makes possible a qualitative analysis of the dynamics of social systems without the difficult methodological problems associated with over-time data gathering. Obviously, this method is not equivalent to gathering data on each variable over time but it is a significant improvement over the usual static, "point-in—time" hypotheses which characterize social research. Fourth, the method is relatively flexible since it can meet the assumptions of mutual causal systems (as described by Maruyama) or the assumptions of modified cybernetic 48 systems (as described on pagesliithroughlt7), and since the theorist can opt for stability only or stability plus flexibility conditions. Further- more, the method can be applied to any content area for which the assump- tions of causal loops among variables, a closed system of variables, and linear, additive relationships among variables can be met. On the other hand, the approach provides enough of a "cookbook" style to permit even the mathematically unsophisticated to apply the methods to substantive problems. Although we have focused solely on the implications for theory con— struction in this summary of Chapter I, the significance of the model of interpersonal communication should not be underplayed. However, we shall refrain from comment on the implications of the model itself for inter- personal communication until the final chapter when a data—based appraisal can be offered. CHAPTER II METHODS AND PROCEDURES This chapter presents the procedures, operationalizations, and meth- ods of statistical analysis developed for putting the model of Chapter I to an empirical test. Although the discussion here will be primarily descriptive, three issues of general concern will be argued. The first concerns the testing of synthetic versus explanatory deductive theories. The second concerns the characteristics of the general class of objects of coorientation (the X's of Chapter I) which can rightfully be chosen for the model. Third, certain of the difficulties arising in interpret- ing coorientational data will be treated. Testing Synthetic Deductive and Explanatory Deductive Theories The recent sociological literature is giving considerably more at- tention to the techniques and problems of theory construction and valida- tion than in the past (Blalock, 1969; Bailey, 1971; Costner and Leik, l96u). One issue of concern to this study involves the pros and cons of testing deduced hypotheses of’a theory versus testing the propositions underlying those hypotheses. We shall argue that the model presented here is best tested through testing the deduced hypotheses. Bailey (1971, p. 58) distinguishes two general classes of theory: synthetic deductive and explanatory deductive. The two are different ug 50 in terms of their method of formulation and in terms of their testability. Explanatory deductive theories take "known" empirical generalizations and seek to set down postulates from which empirical generalizations can be deduced. Usually the postulates are themselves untestable. On the other hand, synthetic deductive theories take propositions (both empirically established and stipulated) and deduce testable hypotheses from them. The propositions, themselves, may or may not be testable. New, the model which has been developed here is not a pure example of either type of theory. It has elements of the synthetic deductive theory since the underlying propositions are, themselves, testable. Also, elements of the explanatory deductive type are present at two levels. To deduce the derived hypotheses, a postulate about system stability and growth is necessary (Proposition 9'). Without this postulate, the de- duction would not follow. This constitutes an untestable postulate char- acteristic of explanatory deductive theory. At another level, the test— able, but largely unverified, prepositions which constitute the structure of the model (see Figure 5) are themselves derived from unspecified assump— tions. For example, it is assumed that the accurate transfer of symbolic information is a necessary and sufficient condition for interpersonal predictability. Thus, it appears that there are two levels at which the model of Chapter I can be tested: at the level of the underlying prepositions and at the level of the deduced hypotheses. The fermer level represents the explanatory deductive aspects of the model and serves primarily as a test of the reasoning (from the non-explicit postulates) on which the prepositions are based. The latter level represents the synthetic de- ductive aspects of the model and serves primarily as a test of the 51 underlying propositions and the postulate on which it is based. Let us consider the merits and demerits of each approach. To adequately test the model of Figure 5 or even the simplified model of Figure 6 requires not merely a test of the eight propositions each in isolation, but also of their overall interrelation. This interrelation, as represented in equations (3) - (12), is very complex because it relates the time_£§£2_2£_change of each variable to the value of the other vari- ables. Classical design methodologies (Campbell, 1963) are not suited to this complex situation. Although the techniques of path analysis applied to panel data suggest themselves as a method (see for example, Heise, 1971), they too are inadequate. The model even in its simplified form is neither recursive nor block-recursive (as required by path anal— ysis) and, hence, cannot be attacked with that methodology. There has been some effort within econometrics to treat over-time data in non- recursive systems (Koopmans, 19u9), as is the case here, but developments are limited and relatively unknown among the other social sciences. Thus, although the individual propositions can be readily tested, their over- all interrelationship cannot. For this reason, validating the underlying propositions would not constitute validation of the derived hypotheses given the postulate.ll On the other hand, a test of the derived hypotheses is at least an in- direct test of the entire model. It is only an indirect test because even if the derived hypotheses are validated, then any set of propositions llActually, even if the model were validated, the derived hypotheses would not be since the conditions for stability and growth are only nec- essary and not sufficient conditions. Hence, the truth of the proposi- tions does not insure the truth of the consequences of the propositions. 52 leading to the same hypotheses would also be validated. As the number of variables increases, the number of possible models consistent with the derived hypotheses also increases. In a similar fashion, failure of the hypotheses does not necessarily indicate that the theory is false in its entirety but can mean that one or other of the propositions is in error. On more pragmatic grounds, simply testing the propositions in isola— tion would require three separate studies, whereas one study would suffice for both derived hypotheses. Thus, fer practical, theoretic, and method- ological reasons, we will seek to test the model by testing the derived hypotheses. In so doing we are obtaining indirect evidence on the valid- ity of the model as a whole. The Necessary ConditiOns for Coorientation Throughout the earlier discussion of coorientation, no attempt was made to explicate what persons in the interpersonal situation coorient about. Obviously, not any X can be input as an object of coorientation if we expect the propositions to be predictive. For example, let person A be a department chairman and B a graduate student in A's department. If X is "Country music," then high levels of system consensus could not possibly predict coordination on the tasks normally carried out by graduate students and their chairmen. Furthermore, low levels of per- ceived consensus on this X would be very unlikely to eventuate in com- munication activity designed to reintroduce consensus. Thus, fer this particular model of coorientation the first criterion fer choice among possible objects of coorientation is that they be objects which are per- tinent to the general set of tasks carried out in interpersonal situations. 53 For example, if one chooses to study married couples, then the objects of coorientation chosen must be pertinent to the tasks requiring coordination in marriage (decision—making, handling quarrels, role-distribution, etc.). If the criterion of pertinence, as I shall call it, is not met then our predictive propositions are likely to fail. Chaffee (1971, pp. 3—u) has developed other criteria which are more general than the pertinence criterion and also necessary conditions for the application of the coorientation strategy. When one or several of these criteria are not met, then the possibility of generating "pseudo- data" (as Chaffee calls it) is greatly enhanced. The criteria include: la. 1b. Ha. Mb. 5a. 5b. 6a. 6b. 7a. Person A is simultaneously oriented to B and to some object X. Person B is simultaneously oriented to A and to some object X. The elements of X perceived by A are identical to those in B's orientation to X. The cognitive and affective dimensions of judgment of X in A's orientation are identical to those in B's orientation to X. A is oriented toward B, cognitively and affectively. B is oriented toward A, cognitively and affectively. A is oriented toward PlB' B is oriented toward PlA. A sees PlB as relevant to PlA and his A-B orientation. B sees PlA as relevant to PlB and his B-A orientation. PlB must be communicable to A. PlA must be communicable to B. 5” With some obvious modifications to include the third order perceptionsl2 these seven plus the pertinence criterion provide the rule for choosing among possible objects of coorientation. Criteria two and three can be satisfied by careful design and question construction. Criterion four can be satisfied by a judicious choice of dyads. In the present study, married couples will be used and it will be assumed that they are indeed oriented toward one another, cognitively and affectively. Yet even within these constraints, there exist several classes of X's satisfying the remaining criteria: 1, 5, 6, and 7. For example, role prescriptions and role expectations (Levinger and Breedlove, 1966), or relational statements (Laing, Philipson, and Lee, 1966) would suffice but attitude items on topics of general interest would not. The point is that those classes of X's which satisfy Chaffee's criteria and the pertinence criterion can be used without fearing the generation of psuedo-data. Procedural Rules a§_Objects‘2£_Coorientation The pertinence criterion suggested above evaluates items as objects of coorientation relative to the tasks in which couples normally engage. The primary nature of these tasks is symbolic in that they generally involve coming to agreements, resolving disagreements, clarifying mis- understandings, making decisions, and the like. In short, the ability 128a. A is oriented toward P2B. 8b. B is oriented toward P2A. 9a. P2A must be communicable to B. 9b. P2B must be communicable to A. 10a. A sees P2B as relevant to P A and his A—B orientation. 10b. B sees P2A as relevant to PlB and his B—A orientation. 9 55 to coordinate activities in marriage necessitates completion of a set of tasks which are essentially symbolic. A recent position put forward by Cushman and Whiting (1972) argues that conjoint, combined, and coordinated human action is facilitated by the transfer of symbolic infermation when that symbolic information itself is governed and guided by content and procedural rules. A rule is a prescription for action which indicates what ought, must, or might follow in a specific set of circumstances. In particular, procedural rules are those implicit contracts which govern and guide the w§y_interactants carry out their communication activities. Procedural rules are meant to govern not the meaning of the messages generated but, for example, how or when messages might be sent, what sequences are permitted, what responses are acceptable, and the like. In principle, then, procedural rules fit the criteria fer objects of coorientation quite well. That is, they meet the pertinence requirement by permitting coordination on the symbolic tasks of marriage, and Chaffee's requirements by virtue of the fact that they govern and guide the £312: tionship between A and 3.13 Within the interpersonal communication literature itself, Miller, Nunnally, and Wackman (1972) have argued that consensus on the what_or why of communication is much less important than consensus on the hgw_of communication. In fact, in their interpersonal training for married couples, they teach that requiring interpersonal consensus on a broad range of "what's" (such as political attitudes, interests, etc.) can be detrimental by stifling individual autonomy. Rather, the way in which 3Several studies in organizational communication have recognized at an intuitive level the importance of agreement, accuracy, and per- ceived agreement on procedural rules of communication (Russell, 1972; Berlo, _e_t_. 31:, 1971). 56 problems are to be handled, hgw_decisions are to be made and interactions carried out require consensus. In other words, the rules governing inter- action are the primary objects of consensus. However, the actual specifi— cation of such rules is at a very elementary level in their work. Never- theless, the applicability of procedural rules as objects of coorienta- tion cannot be disputed. As a class they satisfy the criteria fer objects of coorientation provided above. In addition, they are closely tied to the theoretic framework of Chapter I in that the criteria for choosing among operationalizations is dictated by the theory. This latter bonus is of utmost importance in empirical research since the operationaliza- tion of a variable or concept can be achieved in numerous ways. The choice among alternative ways should be dictated by criteria other than happen-stance or ease. When the choice is guided by theory, then a stronger test of the theory is possible in that failure cannot be ex- plained away by appealing to the inappropriateness of the operationali- zations. Just as the measurement of mass became possible only after Newton's laws were established, so any operationalization is best carried out in the context of the theory for which it is relevant. General Method and Procedures The derived hypotheses were tested in a nonexperimental setting on a self-selected sample of married couples. The sole source of data was self-report in form, obtained through self-administered questionnaires. The five sections of the questionnaire sought to tap system and perceived individual consensus on rules, marital satisfaction and a prediction of marital longevity, and amount of communication generated by the husband and by the wife. 57 Pre-testing A small, and relatively homogeneous group of graduate student couples served as subjects for a pre-test of an earlier form of the questionnaire. Because of the length and complexity of the questionnaire, results were used primarily to shorten and simplify the questions of sections I, IV, and V. The final ferm of the questionnaire is presented in Appendix C. At the time of the pre-test only one subject pair indicated any discom- fort or anxiety concerning the information sought. Sampling The final sample consisted of 35 married couples from Madison, Wisconsin, and Lansing, Michigan, who volunteered to participate in the study. The subjects were run between mid-November, 1972 and mid—February, 1973. They were solicited from church groups (one in each city), from university housing in Madison, from a low-to-middle income housing cooperative in Lansing, and from undergraduate communication courses at the University of Wisconsin, Madison, and at Michigan State University, East Lansing. A copy of the leaflet circulated to the university housing in Madison and to the housing cooperative in Lansing are included in Appendix D. Of the 35 dyads, three sets of questionnaires were not included in the final data analysis, leaving the final sample size at 32. One ques— tionnaire had inadvertently left out section V in the collating process. In the other two sets, one of the partners failed to complete a significant number of questions in sections I and IV. It was felt that the data for these pairs could not be validly salvaged through the usual missing- data techniques. 58 Although the sample appears to represent a broad range on age and years married, and a less broad but variable range on socio-economic status, the self-selecting nature of the sample places severe limita— tions on the generalizability of the results. The violation of the as— sumption of randomness is not unique to this study but, in fact, seems to be characteristic of studies carried out on married couples (for ex— ample, Levinger and Breedlove, 1966). Unfortunately, there is no simple solution to this sampling problem when the information desired from re- spondents is personal and highly salient. In sum, the small sample size and its non-randomness requires that the data be treated primarily as exploratory and of limited generalizability. Administrative procedures Once subjects had been contacted and had volunteered their time, individual and group administration times were arranged at the conveni- ence of the respondents. Sixteen of the pairs were administered the questionnaire individually and sixteen of the pairs in groups of two or more couples. In all cases husbands and wives filled out the question— naires simultaneously and great care was taken to separate the spouses during this time. An administrator was always present to insure that the questions were answered separately. The typical individual session took place in the home of the respond- ents. The administrator talked through the general instructions, a copy of which is presented as a cover sheet in Appendix C. The pair then opened their individual envelope, took one questionnaire each, separated, and began working. Upon completion, the subjects returned their question- naires to their envelope, sealed it, and returned the envelope to the 59 administrator. At this point the pair was debriefed. Throughout the ad- ministration care was taken to give the subjects confidence that their anonymity was being guarded. Instrumentation and Indices In testing the deduced hypotheses, it is necessary to obtain or develop measures of the husband's and wife's communication activity, procedural rules to serve as objects of coorientation (ultimately to be used as indices of perceived individual consensus and actual system con— sensus), and measures of marital satisfaction and predictions of marital longevity. In this section, we shall indicate the choices made in each of the above areas, the validity and reliability of the measures when available, certain of the problems Which arise with each measure, and the specific indices developed. Measuring system consensus (X1) and perceived consensus (X and X5) 3 Attempting to develop procedural rules applicable to a broad range of married couples and yet fit the criteria laid out by Cushman and Whiting (1972) proved to be no simple task. The literature employing a coorientational or person perception scheme tends to use rather simple scales for first and higher order perceptions (for example, Laing, Philipson, and Lee, 1966). Theory requires our Operationalizations to be more complex. The best sources of items for procedural rules was found to be observational rating scales used in the evaluation of family inter— action (Behrens, 35: gl:, 1969), in the evaluation of counselors' empathy (Kurtz, 1968) and in the evaluation of family 6O quarrels (Bach and Wyden, 1970). Observational scales from these sources attempted to tap a single dimension for each item, to provide a detailed verbal description for each level of the scale, and, presumably, to tap an important interpersonal skill. This combination made the adaptation of the scales to self-report questions relatively straightforward. The final set of items used in this study is to be found in Appendix C, Section I of the questionnaire, questions 1a, 2a, 3a, Ha, 5, 6a, and 7a. It should be noted that each question does give an indication of circumstances in which the rule is operative, as the Cushman and Whiting formulation re- quires, and a set of alternatives descriptive of the couple's behavior in the situation. The questions fall into four categories: (1) rules on how to quarrel (la, 2a, 3a), (2) rules on discussing sensitive tOpics (4a), (3) rules on making decisions (5, 6a), and (u) rules on what topics can be brought up for discussion (7a). It is h0ped that these four categories are simply dimensions of a more general construct described as procedural rules for transferring symbolic information. This set is obviously unique to interpersonal communication situations. In order to obtain measures for P2 and P3, it was necessary to repeat the same questions two more times. Questions la, 2a, 3a, 4a, 5a, 6a, and 7a of Section IV (see Appendix C) ask the subject to answer the same set of questions in terms of what his spouse thinks. Questions lb, 2b, 3b, Ab, 5b, 6b, and 7b of the same section ask him to predict what his spouse thinks that he thinks. In order to minimize the tendency to project one's own orientation to the other, Section IV did not follow Section I but has: separated from it by other questions. Questions 1-6 in both sections were: coded l to 7 by superimposing a clear plastic sheet which had pre- maI’ked units equal in length to the vertical scale lines. 61 In order to obtain a measure of the importance of agreement and ac- curacy on each rule (constants we denoted as G1 and G2 in the previous chapter), the "b" questions of Section I and the "c" questions of Section IV were developed. Both attempt to measure a general intensity of feel- ing regarding disagreement or misunderstanding in terms of a propensity to communicate either persuasively (the "b" questions) or informatively (the "c" questions). These important measures, G and G2, will be used 1 to weight the agreement and perceived agreement, accuracy, and perceived accuracy scores respectively, as indicated in the previous chapter. Developing an index of a pair's actual system consensus and per- P P G and G measures 1’ 2’ 3’ 1 2 is, unfortunately, not a trivial matter. According to Wackman (1969), ceived individual consensus based on the P the general approach used to obtain indices of relational variables, such as agreement and accuracy, has been to use the so-called D2 score. If X1 is (PlA)i for the ith item and Yi is (P2B)i for the ith item, then 2 N N is a measure of B's accuracy, where N equals the number of items. But, as both Wackman and Cronbach (1955) have shown, D2 is not a pure measure but is confounded by individual response set factors. In fact, Wackman (1966, Appendix A) shows that 2 D = (Y-T)2 + (s — s )2 + 23 s (l—r ) (13) x y x y xy Where sx is the standard deviation for the (PlA) distribution, and sy is irkre standard deviation of the (P2B) distribution. The response set 62 differences give rise to the first two terms on the right in equation (13) and only the final term represents a "pure" measure of accuracy. One way to obtain such a measure is to use rxy as an index of similarity rather than D2. In this way, we are assured that X'= T and 3x = Sy’ and response set effects are controlled. On the other hand, the standard error of estimate of the z-value cor- responding to rxy is 6’2 = l/(N-3)l/2, where N equals the number of items (McNemar, p. 157). In the present case with thirteen items, O’z equals .316 and, hence, rxy as a measure of the degree of similarity is highly unstable. Furthermore, the measure of accuracy that is needed according to Chapter I is one in which the accuracy on each item is weighted by the importance of that item, in particular by G This destroys the possibility 2. of interpreting rxy in any of the usual ways. Lastly, Wackman (1969, p. 17) presents some data to support his contentions as to the effects of response set on the D2 index. The data are by no means unequivocal. The greatest correlations are between D and (l—rxy) and they are significant ath.= .05. The correlations between D and (X'- T)2 and D and 2sxs are low and not significant. The correlation between D and (Sx — sy)2 is significant, but only for person 1 and not for person 2. In sum, the ‘4 data only confirm that the rejected D score in fact correlates highly with the proposed index, rxy. For present purposes, were it not for the problem of instability and the necessity of weighting each Xi’ Yi pair, wachman's approach would be the safest. But there is some condolence in the fact that, according to Wackman's own data, the greatest percent- age of the variance in D2 is explained by (l - rxy). Thus, using the D2 approach, the following indices were developed. AS 61 Ineasure of the amount of disagreement, 63 2 _ i3 GlA + 313 . 2 As a measure of the amount of inaccuracy, _ 13 2 2 Dace -Z M «321% ((13211)i - (lei) + (c223)i ((323) - (plzm >. As a measure of the degree of failure to realize, 2 _ 13 _ 2 _ 2 rel -:E: i=1 (((P3A)i (P2B)i) + ((PBB)i (P2A)i) ) And, the inverse of actual system consensus, D, depends on these indices as fOllows: It should be remembered throughout that D is a measure of lack of consensus and not of the amount of consensus. As a measure of perceived disagreement for A, 2 _ 13 2 Z i=1 ((3111)i ((PlA)i - (P2115) . DpagrA - The amount of perceived inaccuracy for A is given by 2 _ 13 2 paccA - Ei:i=1 (GQA)i ((P3A)i - (PlA)i) . And, as before, the inverse of the perceived consensus for person A is _ 2 2 1/2 DPA ' (DPAGRA + DPACCA) There are similar indices for B's perceived consensus. 6'4 Measuring A's and B's communication (Xu and X6) The instrument used to measure the amount of communication generated by A and B is an adaptation of the Primary Communication Inventory (PCI) (Navran, 1967) used widely in family and marital research (Petersen, 1969; Locke and Sabagh, 1956). The instrument is a self-report measure of verbal and non-verbal activity filled out by both spouses. The reliability of the instrument has not been reported at this time. No tests of its valid- ity are reported either. In order to obtain a measure of the amount of communication generated by A and B separately, three sub-indices from each questionnaire were formed. The first sub-index IiA is a measure of A's report of his own communication activity (questions 6, 10, l2, l7, 18, 21, and 23 of Section V; see Appendix C). The second, I A, is A's report of his spouse's com— 2 munication activity (questions 5, 7, 9, ll, 16, 19, 20, 22, and 2a). The third, I3A, is A's estimate of the couple's communication activity (ques- tions 1, 3, H, 8, 13, lu, 15, and 25). A similar set of sub-indices are calculated for B. To obtain an index of A's communication, the sub-indices were added as follows: COMA = I A + I B + (I3A + I33). 1 2_"2"— That is, A's communication is derived from an estimate by A of his own activity, an estimate by B of A's activity, and an average estimate by A and B of their communication together. A similar index is calculated for B. 65 Measuring satisfaction and prediction The derived hypothesis of Chapter I is actually two hypotheses. The first concerns the satisfaction or adjustment of the marriage at a point in time. The second concerns the longevity of the marriage over time, that is, its long run successfulness. Instruments to measure these exact variables have been available since the early 1950's (Burgess and Wallin, 1953). These instruments, widely tested and validated, had been developed for use primarily in family therapy and counseling and did not find their way into research literature because of their length and complexity (up to 246 questions). However, Locke and Wallace (1959) have developed a short form of the tests which conforms to the constraints of research situations. Several studies have used this measure successfully in both experimental and field settings. Locke and Wallace (1959, p. 254) report that the short-form adjust- ment test shows a split—half reliability corrected by Spearman-Brown for- mula of .90. Furthermore, of the sample of 236 married couples tested by Locke and Wallace, 48 were known to be maladjusted from psychiatric case study data. Only 17% of the maladjusted received scores of 100 or greater, while 96% of the adjusted group received scores of 100 or greater. Thus, the satisfaction test seems to be both reliable and valid. The test is reproduced in Section II of the questionnaire (see Appendix C). The prediction test shows a reliability corrected by Spearman—Brown formula of .84. There was no longitudinal study carried out to validate the short-form prediction test. However, it showed a correlation of .47 with the adjustment test which is essentially the same as that between their longer counterparts. Hence, the prediction test has some degree of valxidity as well. The test is reproduced in Section III of the questionnaire. 66 The indices formed from each test are straightforward sums of the scores on each question, where the special scoring system developed by Locke and Wallace was used but is not indicated in Appendix C. On the prediction test, scores can range from 0 to 532 for men and from 5 to 502 for women. On the adjustment test, scores can range from 3 to 149. Testing_the Data Since marital satisfaction and longevity are at least conceptually distinct if not operationally distinct variables, then the derived hy- pothesis ought to be tested separately for each. The reasoning develOped in Chapter I suggested no difference between satisfaction at a point in time (the time of administration) and longevity over time, and, hence, none should be expected. Now, a clear-cut test of the derived hypotheses for satisfaction would pit a low satisfaction group against a high satisfaction group on each of the estimates of a 36’ and a26 (see Table 4). If in 24’ asu’ a Table 4. Primary Format for the Data Display and Testing. Hi Satisfaction Lo Satisfaction est a2” est a2n est a5” est a5n est a36 est a36 est a26 est a26 each of the four cases the estimate of the aij coefficient is significantly greater in the high satisfaction than in the low satisfaction group, then 67 the derived hypothesis for satisfaction will have unequivocal support. A similar argument holds for the high longevity and low longevity groups. It may not be clear that this constitutes a test of the derived hy- pothesis for satisfaction. The hypothesis in essence states that when the correlations between measures of communication and certain measures of consensus are all positive, than flexibility-stability is optimally balanced, while a system with all negative or with two negative (a2 and u as”) and two positive (a2 and a36) correlations favors growth in the 6 former case and stability in the latter case. Assuming that both satis- faction and longevity vary directly with an optimal balance of flexibility— stability, then the greater the satisfaction, the greater the observed correlations. The derived hypothesis does pg£_expect that the correla- tions of the high satisfaction group will be positive and the low group all negative but that the correlations in the high satisfaction group will be greater than those of the low satisfaction group. In studies involving marital adjustment and longevity, groups are usually pp£_divided into high and low on the basis of the median. In order to insure a successful "manipulation," extreme groupings are usually used with middle range scores left out. This will be our primary mode of data display. unfortunately, the measures of aij will be correlational /2 and, hence, dependent on 1/(N—3)1 for their stability. Thus, using extreme groupings will increase the instability of our estimates by decreasing N while also increasing their purity. The estimates of a 36’ and a2 will of necessity be first— 24’ asu’ a 6 order partial correlations rather than product moment correlations. For is the degree of variability in X3 attribut- According to Figure 6, this variability example, an estimate of a36 able to the variability in X6. 68 can arise through the direct causal relationship between X6 and X3 and through the direct causal relationship between X6 and X3 and through the indirect causal relationship from X6 to X2 to X1 and, finally, to Since only a measure of the direct influence between X6 and X3 is desired, the effects of the indirect path need to be controlled. Since X3. no measure of X1 is available, r63 2 is the best estimate of a36' Simi- larly, the best estimate of a5 is r 4 54.2' In an analogous way, the correlation between X2 and X” in Figure 6 can arise through the path XE—+~X§——’Xé—aix2 as well as directly. Hence, the best estimate of a controls for the indirect path and is r 24 24.5' The best estimate of a26 is r26.3. The data display of Table 5 will then consist of partial correla— tion coefficients which are independent between the high and low groups but correlated across levels since they are based on the same sample. Unfortunately, no significance test exists for testing the overall dif- ferences between treatments and across correlated levels. But our pri- mary interest is in the overall pattern and direction of the relation- ships, not in their mere differences. However, the significance of difference between partial correlation coefficients can be tested (Hays, p. 576; Blalock, p. 406) within one level. The test involves trans- forming the partial r's to 2'3 by Fisher's Z-transformation and taking their difference relative to the standard error of estimate for difference between two z—scores. That is, z - z' 69 where Ni = the number of observations in group i, and k = the number of variables controlled in the partial correlation coefficient (one in this case). Obviously, the value of Z is directly dependent on the square root of the sample size. In fact, to obtain a Z = 1.96 (minimum value for 0‘ = .05, two-tailed) for the maximum group size in this study (N1 = N = 16) with k=1, 2 - z' must be at least .800. Obviously, with the 2 small sample size of this study, significant differences will be diffi- cult to obtain. CHAPTER III RESULTS In this chapter data will be presented which is pertinent to testing the model of Chapter I, describing the overall characteristics of the sam- ple, and evaluating the strength of the data itself. The chapter proceeds by presenting statistics descriptive of the individual measures, correla— tions among variables for the entire sample, and, finally, data for testing the satisfaction and the longevity hypotheses. Descriptive Statistics for the Individual Measures The means and standard deviations for average satisfaction, and husband and wife satisfaction are presented in Table 5. The difference between Table 5. Means and Standard Deviations for Satisfaction. Mean Std. Dev. Average satisfaction 101.39 26.89 Husband satisfaction 100.94 30.03 Wife satisfaction 101.84 28.57 husband's and wife's mean satisfaction is not significant (t = —.l2, df = 62). In addition, the correlation between husband's and wife's satisfac- tion is 0.741 and, hence, the mean is probably a valid measure of the couple's satisfaction. Table 6 presents means and standard deviations for husband's, wife's, and average prediction of longevity scores. 70 71 Table 6. Means and Standard Deviations for Prediction of Longevity. Mean Std. Dev. Average prediction 328.02 59.94 Husband prediction 318.56 71.56 Wife prediction 337.47 74.90 The means on husband's and wife's prediction scores are not significantly different (t = -l.02, df = 62). Although the correlation between husbands' and wives' prediction scores was only 0.339, this is to be expected from the individual nature of the personality and demographic items which con- stitute this index. Table 7 provides data similar to that of the previous tables but for communication. Husbands' and wives' communication is not significantly Table 7. Means and Standard Deviations for Communication. Mean Std. Dev. Average communication 94.70 11.81 Husbands' communication 94.84 12.86 Wives' communication 94.56 11.82 different (t = .09, df = 62) and, in fact, these items correlate 0.900. This very high correlation casts some doubt on the independence of the measures of the amount of communication generated by the husband and by the wife. Table 8 summarizes the means and standard deviations for the various cxnisensus measures. It should be remembered that the D measures are Table 8. Means and Standard Deviations for Consensus Measures. Mean Std. Dev. D 18.59 4.83 DPA (Husband 10.01 3.41 DPB (Wife) 9.22 4.34 72 measures of dissimilarity and of the lack of system and perceived individ- ual consensus. Husbands' and wives' are not significantly different on perceived dissensus (DPA, DPB) measures (t = .80, df = 62). Correlations Descriptive gf_the Entire Sample In order to obtain some indication of the strength of relationship among variables crucial to our model over the entire sample, various cor- relations among communication and dissensus variables were calculated and are presented in Table 9. Using the test presented in McNemar (1969, p. Table 9. Correlations Among Communication and Dissensus Measures for the Entire Sample. COMH COMW D DPH DPW COMH (4) — COMW (6) .900 — D (2) —.403 -.377 -— DPH (5) “.377 —.387 .659 --- DPW (3) -.264 -.304 .266 .204 -- 156) for the significance of a correlation coefficient we find that any r greater than 0.348 will be significant at an CL = .05 confidence level fOr df = 32 (two-tailed test). It is clear that all the r's in Table 9 are 31% significant at CC: .10. Now the correlations of Table 9 can be used to obtain estimates for the aij of the model as indicated earlier but here based on the entire sample. These estimates are presented in Table 10. It should be care- fully noted that it is the absolute value of these partial correlations which represent the estimates of the aij since D, DPW, and DPH are dis- sensus measures. Of the first four correlations in Table 10, all indicate a slight positive relationship between communication and the consensus 73 Table 10. Estimates of aij from Total Sample Correlations. aij Estimated Estimate EEQfiEi a24 rCOMHxD’DPW ”'183 a26 rCOMWxD’DPH "361 a45 rCOMHxDPw-D "l78 a36 rCOMWxDPH’D "199 ass rDPWxCOMW ”'30” a34 rDPHxCOMH ‘°377 measures. Of these, only the estimate of a2 is significant (ci = .05, 6 two-tailed). This should not be construed as a sex difference since essen- tially identical results obtain when husbands and wives are randomly assigned to the A.and B role. The estimate of a3” is significant (CK==.05, df = 30, two-tailed) and that of a near significance (less thancx = .10, df = 30, 56 two-tailed) but are opposite in sign to that suggested in propositions 2' and 3' of Chapter I. This fact will be of some significance later. In sum, the estimates of the aij in Table 10 represent the best estimates of the aij for the entire population. Before turning to the data of primary interest, Table 11 presents several correlates of marital prediction and adjustment which are of Table 11. Communication and Consensus Correlates of Marital Prediction and Adjustment. Average Satisfaction Average Prediction Average Satisfaction ----- .640 Average Prediction .640 ----- Communication Average .710 .334 D -.400 -.445 DPH -.616 -.440 DPW —.412 -.222 74 general interest in the family and marital research literature. Once again all correlations greater than 0.348 are significant (01 = .05, df = 30, two-tailed). Data Display for Satisfaction Groupipgs In this section, the data pertinent to testing the derived hypothesis for satisfaction is presented and the appropriate statistical tests per- formed. The same statistics will be presented in two major divisions, each with two sub-divisions. The first division presents the total sample (N=32), first with husband and wife measures controlled, and second with husband and wife randomly assigned to the person A or person B role. The second division presents only the extreme scores in high and low groups, first with sex controlled and second with male and female randomly assigned to the A or B role. It was necessary to display the data in both forms since preliminary analyses indicated the possibility of strong sex dif- ference. Since the possibility of sex differences is of interest in it— self but cannot be accounted for by the model, it is valuable to retain that control, but it is only fair to evaluate the model on the randomized data. The following four tables, Tables l2, 13, 14, and 15 present the sat- isfaction groupings and the appropriate Z value for significance of dif- ference. A negative value for Z indicates that the direction is opposite of that predicted. In addition, Table 16 presents the mean satisfaction scores for each of the groupings of the previous four tables as a mani- pulation check. All the groups differ significantly on satisfaction by a t-test for the difference between independent sample means (:1 = .001, two-tailed). In fact, the means of the low group in this study compare 75 Table 12. Satisfaction Groups for Total Sample with H and W Controlled. Partial r Lg H_1 Z rCOMHxD’DPW -.553 .227 -2.09* rCOMWxD’DPH -.327 .298 -1.58 rCOMHxDPW‘D —.140 -.102 -0.09 rCOMWxDPH,D .401 —.377 2.01* *Indicates significance ato(= .05, two—tailed test. Table 13. Satisfaction Groups for Total Sample with H and W Randomized. Partial r £9- 51 .Z_ rCOMAxD’DPB --523 .211 —2.31* rCOMBxD'DPA “-420 .230 -1.67** rCOMAxDPB-D -350 -.340 l.76** rCOMBxDPA‘D ~153 --234 0.96 *Indicates significance atcx.= .05, two-tailed test. **Indicates significance at c( =.10, two-tailed test. Table 14. Satisfaction Groups for Extreme Scores with H and W Controlled. Partial r gg_(N=11) Hi_(N=l2) §_ - .. ** rCOMHxD’DPW .499 .340 1.74 rCOMHxDPW°D -.205 -.260 0.11 **Indicates significance at = .10, two-tailed test. 76 Table 15. Satisfaction Groups for Extreme Scores with H and W Randomized. Partial r Lg (N=11) ii. (N=12) g rCOMAxD-DPB -.381 -.021 -0.76 rCOMBxD-DPA -.309 .210 -1.06 PCOMAxDPB-D .218 -.208 0.87 Table 16. Mean Satisfaction Scores for High and Low Groups of Tables 12, 13, 14, and 15. Mean :11 1°. 1:. 51?. Table 12 123.37 79.41 -7.52 30 Table 13 123.37 79.41 -7.52 30 Table 14 126.25 69.54 -9.85 21 Table 15 126.25 71.50 -9.36 22 favorably with the mean adjustment score for the maladjusted sample of couples reported by Hobert and Klausner (1959, p. 260) as 71.17). Simi- larly, the mean adjustment score for the well-adjusted couples was reported at 135.9, only slightly higher than the mean adjustment score in the high groups in this sample. Data Display for Prediction Groupings In this section, the data relevant to testing the second derived hypothesis (i.e., concerning marital longevity) is presented. The order of presentation, division among the groupings, and statistics presented and tested are identical to that of the previous section. The following four tables, Tables l7, 18, 19, and 20 present the Prediction groupings and the Z statistic for the significance of difference 77 Table 17. Prediction Groups for the Total Sample with H and W Controlled. Partial r Lg H_i_ _Z_ rCOMWxD’DPH -.590 —.221 -l.ll rCOMHxDPW'D .040 .275 0.79 rCOMWxDPH'D -.151 —.256 0.27 Table 18. Prediction Groups for the Total Sample with H and W Randomized. Partial r _L_o_ EL _Z_ rCOMAxD°DPB -.047 -.252 +0.51 PCOMBxDPA'D -.l76 -.226 0.13 Table 19. Prediction Groups for Extreme Scores with H and W Controlled. Partial r _L_o_ (N=ll) H_i_ (N=12) _Z_ rCOMHxD'DPW —.201 -.119 -0.16 rCOMWxD°DPH -.147 —.l96 0.09 rCOMWxDPH'D -.197 -.638 1.04 rCOMHxDPW’D .027 ~.280 0.59 78 Table 20. Prediction Groups for Extreme Scores with H and W Randomized. Partial r Lg_(N=12) §i_(N=l2) g_ rCOMAxD,DPB -.213 -.237 0.50 rCOMBxD’DPA -.180 -.011 —0.34 rCOMAxDPB.D .001 -.541 1.22 .031 -.548 1.30 rCOMBxDPA‘D between high and low groups. None of the differences are significant and only r d in Table 20 begin to approach significance. rCOMBxDPA°D As with the satisfaction groups in the previous section, the predic— COMAxDPB'D an tion groups are significantly different on the prediction measures. The data on the appr0priate means will be found in Table 21. All are signi- ficant (c1 = .001) by t-test for difference between independent sample means. Table 21. Mean Prediction Scores for High and Low Groups of Tables 17, 18, 19, and 20. Mean 14'. 1:9. I. if. Table 17 376.78 279.16 -7.82 30 Table 18 376.78 279.16 -7.82 30 Table 19 392.14 266.68 -8.70 20 CHAPTER IV INTERPRETATION OF RESULTS, CONCLUSIONS, AND IMPLICATIONS The primary purpose of this chapter is to interpret and evaluate the model of Chapter I in light of the empirical results of the previous chapter. The interpretations and evaluations offered will aim to modify the model when empirical results so dictate, to draw out from both the data and model the implications fOr interpersonal communication, and to provide direction for future research on modification, and extension of the model. Validity gf_the Sample and Instruments Although we have no direct evidence for the representativeness of the sample or validity of the instruments, several data reported in the previous chapter offer indirect evidence supporting the representativeness of the sample and the validity of the test instruments. The correlation between satisfaction and prediction measures (.64) is only slighly higher than that reported by Locke and Burgess (1959, p. 261) (.47) between the same two measures on a much larger sample (N=236). Furthermore, the means for the low and high satisfaction groups (see Table 16, page 76) are very similar to the means for the maladjusted and adjusted groups in the Locke and Burgess study. Together, these results indicate that the sample used in this study is not severely biased relative to the much larger Locke and Burgess sample on satisfaction and prediction measures. Also, the 79 80 high correlation between the husbands' and wives' estimates of their satis- faction (.74) suggests that dividing couples into high and low groups by their mean score on this measure is a sound estimate of the couple's sat— isfaction. The same is not true of the prediction measure since the corre- lation between husbands' and wives' prediction scores is low enough to be non-significant (.34, cX_= .05). Also on the negative side, the extremely high correlation between husband's and wife's communication (.90) must be suspected, since such a correlation may be the result of an artifact of measurement or of the index construction procedure. Conceptually, the amount of communication generated by A and that generated by B are dis- tinct. But one interpretation of the high observed correlation is that the communication measure is tapping the cogple's communication together and not the individual's communication to the other. Unfortunately, there appears to be no means of distinguishing this interpretation from the one advocating that the correlation is a "true" measure of the conceptually distinct variables. In the absence of such distinguishing data, the cor- relation must be treated as valid. Several of the correlations of Table 9 (page 72) can also be compared to correlations between conceptually similar (but by no means identical) pairs of variables often measured in marital studies on communication and empathy. For example, average satisfaction and average communication correlate at .71 as compared to a correlation of .91 between verbal com- munication and satisfaction and a correlation of .66 between non—verbal communication and satisfaction, as reported by Navran (1966, p. 178) using the same instrument to measure communication as in this study. The posi— ‘tive correlations between average satisfaction and husband's and wife's jpereeived consensus are also to be expected from the literature (for 81 example, Levinger and Breedlove, p. 370). In addition, we should expect that the husband's perceived consensus explains more of the variance in satisfaction than the wife's (Levinger and Breedlove, 1966, p. 369), and this is the case here as well (38% versus 17%). Finally, this sample shows a somewhat stronger relationship between satisfaction and system consensus (approximately .40) than is usually reported. In fact, Hobart and Klausner (1959, p. 259) report an essentially negligible correlation between empathy and adjustment. However, this is an exceptional result. Although empathy and system consensus are not strictly comparable, they are conceptually similar and should exhibit similar direction and roughly similar magnitudes. Together the above results provide additional indirect evidence that the sample self-selected for this study is at least comparable to samples used in studies employing conceptually similar variables. Although we have no direct validity check on our operationalizations (other than satisfac- tion), the above comparisons do indicate results comparable to studies employing conceptually similar variables and, hence, a certain minimum convergent validity. Testing the Underlying Propositions: Estimates 2£_aij The estimates of the aij coefficients presented in Table 10 (page 73) have been shown to be the best estimates of the path coefficients between variables i and j in the model of Figure 6. In Chapter I, it was argued that the amount of communication should exhibit no significant positive or negative relationship to system consensus or to the perceived individual consensus variables across the entire sample. This reasoning produced Propositions 4', 5', and 6'. Now while three of the four correlations 82 which can test this reasoning are not significant (a 5, and a of 36 is, all indicate a slight positive relationship 24’ al4 Table 10) and only a 26 between communication and the various consensus measures. Taking the small sample size into account, the safest conclusion to be drawn from these data is that the reasoning which led to Propositions 4', 5', and 6' should be modified slightly to allow for small, positive correlations between communication and the various consensus measures across the entire sample. HOwever, it is clear from the results of Tables 13 and 15 that the correlations between communication and consensus measures can take on a range of values from positive to negative depending on the level of satisfaction. We can safely conclude that propositions 4', 5', and 6' are sound in their implication that the effects of communication should differ from level to level. Part of the reason that this small, positive relationship has been consistently observed in the data could be a result of the measure of communication itself. Many of the questions do not isolate the sheer amount of verbal or nonverbal interaction, but actually measure the quality of that interaction. Questions 5, 9, and 13 of Section V of the questionnaire (see Appendix C) are flagrant examples. When the transfer of symbolic information is accurate, a strong, positive relation- ship between communication and the various consensus measures should be expected. It is only the sheer amount, undifferentiated as to quality, which should exhibit no relationship to the consensus measures over the range of the entire sample. The estimates of a and a3 displayed in Table 10 (page 73) are 56 4 opposite in direction to those hypothesized in Propositions 2' and 3'. Also, the correlations are not insignificant and, so cannot be explained 83 away by appealing to sampling error. Propositions 2' and 3' were derived from Newcomb's A—B—X model and the extension to higher order perceptions which was developed here. In addition, there is some empirical support for these propositions reported by Schachter (1951, p. 202) and Festinger and Thibaut (1951, p. 96). They indicate that more communication was directed to another when the speaker felt that there existed a discrepancy or disagreement between himself and the other. While on the face of it these results seem to be directly opposite to the results of this study, let us look deeper. One of the crucial characteristics separating successful fron unsuc- cessful attempts at theory construction is casting the variables of the theory at the same level of abstraction. This is especially true of causal theories, such as the one developed here. Although the "level of abstrac- tion" principle is a sound one, simple rules for satisfying it are diffi- cult to develop.13 I believe that this is the kind of problem which has occurred here. The predictions of Prepositions 2' and 3' are probably valid when there exists perceived dissensus on a topic of immediate rele— vance to the completion of the task at hand (as is the case in the Festinger and Thibaut, and Schachter studies cited). Thus, at a migpg_level of abstraction the propositions probably do characterize interaction. How- ever, in this study, general patterns of communication activity and overall perceived consensus on rules were the measures. These represent a higher level of abstraction. Thus, these findings are probably valid at a macro lanr a general discussion of the abstraction problem and an attempt to develop guidelines, see H. M. Blalock's Theory Construction (Prentice- Hall: Englewood Cliffs, N. J., 1969). 84 level of abstraction. Intuitively, the results are plausible, since we would expect spouses who perceive themselves to be more in agreement with their mate and who perceive the other to be more accurate on the rules governing their interaction, to engage in more communicative interchange than those perceiving disagreement and inaccuracy. Furthermore, it is fairly well-established that high disagreement and low communication are characteristic of maladjusted marriages and that low disagreement and high communication characterize adjusted marriages. These propositions imply that we should not expect the overall patterns of interaction for long-term relationships to be as hypothesized in Propositions 2' and 3'. However, this does not mean that Propositions 2' and 3' are false in other contexts, such as those described in the Festinger and Thibaut, and Schachter studies. In sum, Propositions 2' and 3', while valid in certain contexts, are reversed in this one because of the error involved in measuring the propositions at a high' level of abstraction but casting them at a low level of abstraction. Changing the sign in Propositions 2' and 3' and combining them, they now read Proposition 2": The greater the degree of perceived individual con- sensus on rules for interaction, the greater the amount of communication. But this single alteration has profound effects on the predictions from the model. In order to obtain the derived hypotheses, it was necessary to assume the validity of the direction (or sign) of the underlying propositions since the sign in turn determined whether each ti (see page 37) was greater or less than zero. Now that analysis must be altered to take into account the fact that as” and a56 are positive rather than 85 negative coefficients. Before presenting re-analysis of the conditions for flexibility-stability, it would be wise to evaluate certain other assump— tions which underpinned the analysis of Chapter I. Evaluatipg_the Derived Hypotheses As we have indicated, the data for the satisfaction and prediction groupings was presented both with sex controlled and sex randomized, be- cause sex differences had been anticipated. If sex differences were present, they would show up as patterns of difference between a2” and and r ), and a rC0MHxD°DPw COMWxD-DPH 54 rCOMHxDPW-D and rCOMWxDPH'D)' For the prediction groups (Tables 17 and 19, page 77), no such patterns are present and, hence, sex differences 6 (that is, and a36 (that 13, a2 are negligible. However, for the satisfaction groups (Tables 12 and 14, page 75), a slight but consistent pattern exhibits itself in the low groups. These differences are summarized in Table 22. For the low satisfaction Table 22. Patterns of Sex Differences for Low Satisfaction Groups on a and a26, and a and a . 24 54 36 Coefficient Direction Coefficient Table 12 PCOMHxD'DPW (-.55) <1 rCOMWxD.DPH {-.33) 'rable 1” rCOMHxD-DPw ('°5°) <: rCOMWxD-DPH ('°32) 'Table 12 rCOMHxDPw-D (“'1”) <: rCOMWXDPH-D ('”0) 'Table 1” rCOMHxDPw-D ('°21) ‘<' PCOMWxDPH-D ('29) groups, it appears that the husband's communication activity is more posi- O O Q l” O O ‘tively related to system consensus than is the wife's. More surprisingly, 1”Remember that D, DPW, and DPH are dissensus measures. 0 , 86 the wife's communication activity is positively related to the husband's perceptions of dissensus while the husband's communication shows a slight positive relationship to the wife's perceptions of consensus. If the lan— guage of causality can be temporarily assumed, husband's communication produces slightly more system consensus than the wife's. iFurthermore, (for the low satisfaction groups only) the wife's communication tends to cause more perceived disagreement and inaccuracy for the husband than his does for her. Although these observations cannot be explained within the context of the model, it might be that the effectiveness of communication is cor- related with marital satisfaction. That is; we observe differences between the effects of husband's communication and wife's communication in the low satisfied group and not in the high satisfaction groups. It might be that a relationship is perceived as more satisfying when the efforts at persua- sion, understanding, and discussion are equally efficacious rather than imbalanced in favor of the husband or wife. However, this can only be a tentative hypothesis since the differences on which it is based are not strong. Once husbands and wives are randomly assigned to the A and B roles (Tables 13, 15, 18, and 20), even slight patterns of difference disappear, as should be expected. That is, with random assignment to the A and B should equal a should equal a This is the opera- roles, 6’ and a a24 2 54 36 ' tional equivalent of rejecting situations 6 through 10 of Table 3 (page 39). It had been argued in Chapter I that situations 6 through 10 represented empirically implausible styles of interaction. However, the more com- pelling argument is that any style is possible for a particular couple but random assignment across the sample will insure that as” = a36 and a2”: 87 a To take situations 6 through 10 into account, the model would need 26' to include some exogeneous correlates of sex whose effects would produce differences between 354 and a36’ and between a2” and a26. It had also been argued that situation 3 (see Table 3, page 39) rep- resented an empirically plausible style of interaction despite the fact that a5” # a36 and a # a But by the argument presented above, the 24 26' control of sex differences through randomization allows the plausibility of situation 3 as a style of interaction for a particular couple, but negates the possibility of situation 3 arising across a sample or a sub- group of a sample. Thus, it should not have been included as a possible style of interaction. It was also argued in Chapter I that styles 4 and 5 of Table 3 would not be expected to arise because it would be implausible to assume that the effect of communication on perceived consensus (a 36) would differ 54’ a markedly from its effects on system consensus (a ). It now appears 24’ a26 that this assumption is also erroneous. A quick glance at the Z values in Tables 13 and 15 (pages 75 and 76 ) show that the differences between low and high groups on the estimators of a2” and a26 are opposite in direction to the differences for as” and a36. If situations 4 and 5 were truly impossible styles across the sample, then such stark reversals as found in Tables 13 and 15 could not be present. Although the results for the randomized prediction groupings (Tables 18 and 20 ) are not as striking, there is a similar effect here, especially with the extreme soore group (Table 20). What is implied by the differential results between the effects of communication on actual consensus (a 4’ a26) and the effects 2 of communication on perceived consensus (a54’ a ) is that styles 4 and 5 36 88 cannot be rejected out of hand as implausible nor controlled as sex dif- ferences were. In other words, the rejection of these styles cannot be upheld in the face of the results obtained. In light of the failure of Propositions 2' and 3', and in light of the problems encountered in accepting style 3 which should have been re— jected, and rejecting styles 4 and 5 which should have been included, we would expect our derived hypotheses to fail. And indeed they do. Although as” and a36 are in the predicted direction in all cases, a2” and a26 are not. Now this would be an acceptable result if there was essentially no difference between the low and high groups on a2” and a26, for it could be argued that the system would still exhibit stable-positive characteristics under those conditions. However, although this is almost the case for the total—sample, randomized-prediction groups (Table 18), it is clearly 223_ the case for the randomized satisfaction groupings (Tables 13 and 15). In fact, as pointed out above, there is a fairly strong opposite effect in the satisfaction groups for a than had been predicted. In the extreme 26’ a24 score prediction groups, the differences from low to high on a2“ and a26 are essentially negligible. Because of these clearly unexpected results and failure of the model at a few key junctures, it is imperative that those failures be repaired and we attempt to explain the observed data with a reformulated theory. We turn to this task in the following section. Egplanation and Reformulation With the failure of Propositions 2' and 3', the conditions for stabil- ity and growth must be recalculated. We begin with a new table of possible communication styles akin to Table 3 of Chapter I. Table 23 takes into 89 Table 23. A Reformulated Set of Communication Styles. 1 2 3 4 a24 I ' ' a26 + ’ ' a54 + - + - a36 + - + - account the discussion and reasoning of the previous section to limit the number of different styles.- It should be noted that style 3 of Table 22 corresponds to style 4 of Table 3 and 4 above corresponds to 5 in Table 3. Styles 1 and 2 are the same in both. Now, as in Chapter I, if we assume that the relationships specified in the propositions hold as modified, then 1' if a24’ a26’ a54’ a36 It2 + t3 + t” + t5 + t6! ) 0, then "aij" > 0 if and only if Itll> 2. 1f a2”, a26, as”, a t3+tul>1t2+t 5,, < 0: then "aij" > 0 if and only if Itl + 5+t6l 3. if a2”, a26‘( O, and as”, a36 )' 0, then “aifil>' 0 if and only 1f ltl + t3 + t” + t5 + tel) [1:21 4. if a2”, a26 > O, and as”, a36< 0, then ”aij" ) 0 if and only if fltl+t5+t6I>It2+t3+tul . Applying the same reasoning as in Chapter I, we interpret situation 1 as the maximally unstable, growth-oriented case, situation 3 is the maximally stable case, and situations 2 and 4 represent the cases optimal between flexibility and stability. In the present analysis, if we were to develop new derived hypotheses, then, using the flexibility-stability criterion (Proposition 9') as before, 90 we would expect the data to basically conform to either situation 2 or situation 4. That is, we would expect our results to be of the form of Table 24a or Table 24b for both satisfaction and prediction groupings. However, neither set of new predictions reproduces the results of Tables 13 and 15. In fact, situation 4 is exactly opposite in direction to that reported in the results. The situation which d2§§_best reproduce the ob- tained results is the maximum stability condition, situation 3. Table 24. Expected Results for the Reformulated Flexibility- Stability Cases. a. Situation 2 Low High est a2“ >. est a2“ est a26 )’ est a26 est a2S ) est a54 est a36 > est a36 b. Situation 4 Low High est a24 <. est 824 est a26 < est a26 est as” '> est a5,+ est a36 > est a36 Here, the greater correlations between communication and system consensus should be obtained in the low groups and the greater correlations between communication and perceived individual consensus should be Observed in the high groups. This does seem to be the pattern of results, more signi— ficantly so in the satisfaction than in the prediction groupings. 91 Of course, such a result offers damaging evidence to a central argu- ment of this thesis, namely that stability alone is insufficient to insure system satisfaction and longevity. What we seem to have found in our pg§t_ hgg_analyses is that stability alone is the better predictor of marital behavior. Although it represents a distasteful set of results to this author, it seems that the more satisfied and adjusted couples seek to maintain the status quo. They react to situations which might perturb the current relationship with a communication style which is stability oriented. It should be carefully noted that the more satisfied group has higher mean values on communication, system consensus, A's perceived con- sensus, and B's perceived consensus than the low satisfaction group. These facts are meant to indicate that the stability-oriented styles of the more satisfied couples is pat associated with less overall communication or con- sensus but with differential effects of that communication on consensus across levels of satisfaction. Those stability-oriented styles can be summarized as follows: The symbolic interaction of less satisfied couples tends to produce actual consensus on rules but perceptions of the lack of consensus. The symbolic interaction of more satisfied couples results in the percgption of consensus but actually produces dissensus. The impli- cation of the above statements is that the style of interaction which is stability oriented and, hence, associated with satisfaction is one which seeks to produce the guise of similarity, agreement, and understanding. It need not necessarily result in actual similarity, agreement, and understanding. Of course, the implications of the above results for counseling married couples on their techniques of communication are profound. To 92 achieve satisfaction and happiness, styles increasing the perception of agreement and understanding would be advocated and taught. Because such a guideline for counseling is contrary to some current approaches (Miller, Nunnally, and Wackman, 1971), contrary to strong intuitive views, and based on one study's p2§t_hgg analyses, it cannot be strongly trusted. But also because it is counter—intuitive, it deserves further attention. In sum, we can be relatively satisfied with the quality of the data generated by the instruments used in this study, and while the results are contrary to our initial hypotheses, we were able to reformulate the model of Figure 6 and the possible styles of interaction which could accompany that model. In this sense, the results have corrected our errors rather than fundamentally invalidated our propositions. On the other hand, the applicability of Proposition 9' of Chapter I, which argued the necessity of flexibility as well as stability, to marital communication systems must be called into question. While no one set of empirical results can in- validate such a postulate, that interpretation of the results cannot be lightly brushed aside. Conclusions, Directions Aside from the particular model and the results pertaining to it, several other positive results have accrued from the work in this thesis in the areas of explanation and cybernetics (we have already spoken of the developments for theory construction). From the point of view of explanation, the techniques of theory construction advocated above result in what might be called a system's explanation (Meehan, 1968; Monge, 1972). That which permits one to understand the relationships derived are the propositions and the 93 connection among propositions which undergird the derived relationships. Without the underlying propositions and their interconnection, the derived hypotheses cannot be explained. Indeed, the derived hypotheses would represent nothing but isolated relationships whose significance or in- significance was indeterminant and whose basis was inscrutable. On a less abstract level, we have showed that it is possible under certain conditions to employ a cybernetic logic with its purposive charac— teristics in a causal framework. Although in retrospect the simple trans- formation of variables, which permitted this, seems almost trivial, the effects are significant. Rather than speaking of growth in terms of mere change (as Maruyama does), we are now capable of speaking of growth rela- tive to some goal state. The emphasis changes from mere alteration to directed development or decay. Although the model that was formulated from the above developments was found to be erroneous at several points, its reformulated version still merits the attention of students of interpersonal communication. Any attempts to replicate this study based on predictions from the reform- ulated model ought to include the following modifications: First, the communication index ought to consist of items tapping both qualitative and sheer quantitive dimensions of communication activity. In this way, Propositions 4', 5', and 6' can be more adequately tested. Second, the prediction measure ought to be dropped or replaced by a similar measure, whose validity is well—established or whose validity is established independently of the satisfaction measure. Third, the number of items used as objects of coorientation ought to be increased to permit the use of Wackman's (1969) proposed measure of accuracy without a concomitant 94 large standard error of estimate. Fourth, measures of coordination ought to be obtained so that estimates of a12, al3’ and a15 can be obtained. With these modifications and a slightly larger sample size (about N=50), more precise estimates for stability and growth conditions can be obtained and, hence, a more complete understanding of the roles of stability, on the one hand, and flexibility, on the other, can be obtained. BIBLIOGRAPHY Allport, F. "A Structuronomic Conception of Behavior: Individual and Collective." J. of Abnormal and Social Psychology, LXIV (1962), 3-30. Appelbaum, R. A. Theories of Social Change. Chicago: Markham, 1970. Bach, G. A. and P. wyden. The Intimate Enemy. New York: William Morrow, 1970. Bailey, K. D. "Evaluating Axiomatic Theories." In Sociological Meth- odology, edited by E. Borgotta and G. A. Bohrnsted. San Francisco: Jossey-Bass, 1971, Pp. 48—71. Behrens, M. L., St: al. "The Henry Ihelson Center Family Interaction Scales." GenetiE Psychology Monographs, LXXX (1969), 203—295. Berlo, D. K., gt? El: "Supervisor-Subordinate Communication Relations." Unpublished report of Department of Communication. East Lansing: Michigan State University, 1971. Blalock, H. M. Theory Construction: From Verbal to Mathematical Formu- lations. Englewood Cliffs, New Jersey: Prentice-Hall, 1969. . Social Statistics. Second edition. New York: McGraw- Hill, 1972. Buckley, W. J. Sociology and Modern Systems Theory. Englewood Cliffs, New Jersey: Prentice-Hall, 1969. 111 . Modern Systems Research for the Behavioral Scientist. Chicago: Alwood PubliShing, 1968. Burgess, E. W. and P. Wallin. Engagement and Marriage. Philadelphia: J. B. Lippincott, 1953. Campbell, D. T. and J. C. Stanley. Experimental and Quasi-Experimental Design fOr Research. Chicago: Rand McNally, 1963. Chaffee, S. "Pseudo-data in Communication Research." Paper presented to Association for Education in Journalism Convention. Columbia, South Carolina : 1971 . , and J. McLeod. "Sensitization in a Panel Design: A Coorientational Experiment." Journalism Quarterly, XLV (1968), 95 96 , , and J. L. Guerrero. "Origins and Implica- tions of the Coorientational Approach in Communication Research." Paper presented to Association for Education in Journalism Conven- tion. Berkeley: August, 1969. Coser, L. A. The Functions of Social Conflict. New York: The Free Press, 1956. Costner, H. L. and R. K. Leik. "Deductions from Axiomatic Theory." American Sociological Review, XXIX (1964), 813-835. Cronbach, L. J. "Processes Affecting Scores of 'Understanding of Others' and 'Assumed Similarity.'" Psychological Bulletin, LII (1955), 177-194. Cushman, D. P. and G. Whiting. "An Approach to Communication Theory: Toward Consensus on Rules." Journal of Communication, XXII (September, 1972), 217-238. Festinger, L. "Informal Social Communication." Psychological Review, XVII (1950), 271-282. , and J. Thibaut. "Interpersonal Communication in Small Groups." Journal of Abnormal and Social Psychology, XLIV (1951), 92-99. Gessous, Muhammed. "A General Critique of Equilibrium Theory." In Foundations for a General Theorygof Social Changg, edited by C. E. Russett. ’New Haven, Yale UniverSIty Press, 1968. Hays, W. L. Statistics for Psycholggists. New York: Holt, Rinehart, and Winston, 1963. Heise, D. A. "Causal Inferences from Panel Data." In Sociological Methodology, edited by E. Borgotta and G. Bohrnsted. San Francisco, Jossey-Bass, 1970, Pp. 3‘27- Jackson, D. "The Question of Family Homeostasis." Psychiatric Quarterly Supplement, XXXI (1957), 79-90. Kurtz, R. R. "A Comparison of Different Approaches to the Measurement of Counselor Empathy in Interpersonal Counseling." Unpublished doctoral dissertation, Michigan State University, 1972. Koopmans, T. C. "Identification Problems in Economic Model Construction." In Causal Models in the Social Sciences. Edited by H. M. Blalock. Chicago and New York: Aldine-Atherton, 1971, Pp. 159-178. Laing, R. D., H. Philipson, and A. R. Lee. Interpersonal Perception: A Theory and a Method of Research. New York: Springer, 1966. Levinger, G. and J. Breedlove. "Interpersonal Attraction and Agreement: A Study of Marriage Partners." Journal of Personality and Social Psychology, III (1966), 367-372. 97 Locke, H. J. and Sabaugh, and Thomes. "Correlates of Primary Communi- cation and Empathy." Research Studies of the State College of Washington, XXIV (1956), 116-124. Marcus, M. and H. Minc. Introduction to Linear Algebra. New York: MacMillan, 1965. Maruyama, M. "The Second Cybernetics: Deviation Amplifying Mutual Causal Processes." American Scientist, LI (1963), 164-179. McNemar, Q. Psychological Statistics. Fourth edition. New York: John Wiley 6 Sons, 1969. Meehan, E. J. Explanation in Social Science: Ay§ystem Paradigm. Homewood, Illinois: Dorsey Press, 1968. Miller, S., E. W. Nunnally, and D. B. Wackman. The Minnesota Couples Communication Training Program. Bloomington, Minnesota: Marital Resource Center, 1971. Monge, P. "The Study of Human Communication from Three Systems Paradigms." Unpublished doctoral dissertation, Michigan State University, 1972. Monge, P. "Theory Construction in the Study of Communication: The System Paradigm." Journal of Communication, XXIII (1973), 5—16. Navran, L. "Communication and Adjustment in Marriage." Family Process, VI (1967), 173-184. Newcomb, T. "An Approach to the Study of Communicative Acts." Psy- chological Review, LX (1953), 393-404. Petersen, D. M. "Husband-Wife Communication and Family Problems." Sociology and Social Research, LIII (1969), 375-384. Russell, H. M. "Coorientational Similarity toward Procedural Aspects of Communication." Unpublished doctoral dissertation, Michigan State University, 1972. Schachter, S. "Deviation, Rejection, and Communication." Journal of Abnormal and Social Psychology, XLVI (1951), 190-207. Scheff, T. "Toward a Sociological Model of Consensus." American Sociological Review, XXXII (1967), 32-46. Simon, H. Models of Man. New York: John Wiley, 1957. Speer, D. C. "Family Systems: Morphostasis and Morphogenesis or 'Is Homeostasis Enough'." Family Process, IX (1970), 259-278. 98 Toulmin, S. "Concepts and the Explanation of Human Behavior." In Human Action edited by Theodore Mischel. New York: Academic Press, 1969. Pp. 71-104. Udry, J. R., Nelson, and Nelson. "An Empirical Investigation of Some Widely Held Beliefs about Marital Interaction." Marriage and Family Living, XXIII (1961), 388-392. Wackman, D. "A Proposal for a New Measure of Coorientational Accuracy or Empathy." Paper presented to the Association for Education in Journalism Convention: Berkeley, California, 1969. , and D. Beatty. "A Comparison of Balance and Consensus Theories fOr Explaining Changes in ABX Systems." Paper presented to the International Communication Association Convention: Phoenix, 1971. Zetterberg, Hans. On Theory and Verification in Sociology, New York: Bedminster Press, 1965. APPENDICES APPENDIX A Vector notation, because of its unfamiliarity within the social sciences, can make interpretation difficult. However, a simple trans- formation should clear up this difficulty. A vector is nothing more than a directed line segment with magnitude (the length of the line) and orientation (the angle between the line and some arbitrary axis). Let us take the artificial vector defined as (ACCA, ACCB) = (P2A - PlB’ P2B - PlA) or system accuracy, and apply these criteria and interpret thenn Since this vector is two dimensional, the coordinate system is also two dimensional with the horizontal axis the P2A - PlB value (or A's accuracy) and the vertical axis the P2B - PlA value (or B's accuracy). The origin of this coordinate system is at the point (0,0) or perfect system accuracy. The distance between the point representing system accuracy, (P2A—P B, P B—PlA), and the origin is given by 1 2 2 2 1/2 ((P2A-P1B) + (P2B—P1A) ) which is nothing more than the amount of inaccuracy in the system (about a particular X). Furthermore, the point represented by (P2A-P B, P B-PlA) 1 2 can lie in any of the four quadrants labeled in Figure Al. The angle be- tween the line segment and the horizontal axis determines in which quadrant the vector lies. The angle provides infOrmation concerning the distribution of inaccuracy within the interpersonal system. If the angle is 0° or 180°, then all the inaccuracy is due to A. 100 101 II PaB‘P\A (P2A-P B,P B-PlA) 1 2 PA-PB III IV Figure Al. The Vector Describing System Accuracy. If the angle is 90° or 270°, then all the inaccuracy is due to B. If the angle is in the first quadrant, then both A and B tend to overestimate the other's orientation; if the angle is in the third quadrant, then both A and.B tend to underestimate the other's actual orientation; in quadrant two, A is underestimating and B is overestimating and in quadrant four, A is overestimating and B is underestimating. Similar interpretations can be made for other state vectors. APPENDIX B The theory and details of the methods for evaluating determinants need not be presented here since several mathematical and social scienti- fic texts (Blalock, 1969; Marcus and Minc, 1965) provide lucid discussions. The method used here is that of expanding the determinant by minors as follows: all al2 0 0 0 0 0 a22 0 a24 0 a26 aij = a31 0 a33 0 0 a36 0 0 a43 an” 0 0 a51 0 0 a54 ass 0 0 0 0 0 a65 a66 a22 0 a24 0 a26 0 a33 0 0 a36 = all 0 a43 a44 0 O 0 0 a5” a55 0 0 O 0 a65 a66 '0 0 a2” 0 a26 a31 a33 0 0 a36 + a12 0 a43 a44 0 0 a51 0 a54 ass 0 0 0 0 a65 a66 102 a22 a24 _ 0 a ’a11a33 44 0 a54 0 0 a31 a33 0 a a12a24 ”3 a51 0 0 0 a22 0 a11a33a44 0 a31 a ”al2a24a43 ‘ 51 0 = alla22a33a44a55a66 ‘ (a a a —a33auu 12 26 65 51 -aaaaaa 12 24 43 36 65 51 55 65 55 65 55 65 a 0 a 103 26 a22 -a a O 11 36 o 66 O 36 a31 0 ”312a26 a51 66 0 26 ’alla36a65 66 36 "a12a26a65 59 (a a a a ) a11322 36 65 54 43 ) (a a a a‘ ) a55a66 31 12 24 43 ”312a26a65a54a43331. 33 43 31 51 24 44 54 44 54 33 43 55 65 55 65 24 44 54 44 54 APPENDIX C 105 General Instructions In the following pages are a series of questions about your relation- ship with your spouse. Most of the questions can be answered very quickly. Others will require more thought -- about yourself and your spouse together. The purpose of these questions is to find out how typical couples, like yourselves, interact in marriage. The whole study is contained in this questionnaire. There will be no further interviews or questions. The questions which follow are divided into five sections. You should begin answering the first question in Section I and answer ell questions in the order presented. Once you have completed a page, go on to the next page but d2_pgt_gg_be£§p work as rapidly as is comfortable for you. Remember, some questions will require more thought than others. Your spouse's questionnaire is identical to yours, but it is very important to ANSWER THE QUESTIONS INDEPENDENTLY OF YOUR SPOUSE. One last point must be strongly emphasized: Because of the con- fidential nature of some of the questions, careful steps have been taken to guard your anonymity. No names, addresses, birthdates, respondent numbers, or identification of any sort will be found on the question- naires or the envelope. In this way, both the resporflents and the researcher are protected. There is no way of knowing which question- naires belong to which respondents. When you are ready, open the envelope, make sure that each of you takes one copy of the questionnaire, and begin to answer the questions, independently of each other. 106 Section I In this section you will be asked about certain aspects of your re- lationship with your spouse. Each question will have a series of state- ments describing your relationship. Next to these statements will be a vertical line. Make a slash on the vertical line nearest to the state- ment which most accurately describes your relationship. Questions 0a and 0b are examples of the kind of questions which follow. 0a. I think that ... The almost always have meals together. we usually have meals together. we have meals together about as often as not. we seldom have meals together. we almost never have meals together. and 0b. Suppose your spouse disagreed with you about the necessity of having meals together. How would you feel? 3. I'd feel strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It would bother me some, but not enough to tell him (her). It wouldn't bother me at all. _L All of the questions which follow will have this same fbrmat, so if you have any questions or difficulties now, please ask the administrator to help you befOre you go on. Be sure to answer the questions in terms of how you actually behave, not how you would like to behave nor what you think is acceptable behavior. 107 It is a well-known fact that couples fight, argue, disagree, squab- ble, and quarrel. It is also well-known that this need not be a bad thing. Questions 1 through 3 are about how you and your spouse fight, quarrel and argue. '__— 1a. In our bigger quarrels, I think that ... We usually stOp before the emotional hurts become intolerable to either of us. We usually stOp when the emotional hurts become intolerable to one of us. We sometimes stop when the emotional hurts become intolerable to one of us. We seldom st0p before the emotional hurts become intolerable to one of us. We almost never stop until one or both of us can no longer -- tolerate the emotional hurts. lb. Suppose your spouse disagreed with you about when to stop quarrels. How would you feel? '1!- I'd feel very strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It would bother me some, but not enough to tell him (her). It wouldn't bother me at all. _J- 2a. I think that ... TOur big quarrels almost always end with one or both of us trying to undo or repair damage done during the quarrel. Our big quarrels usually end with one or both of us trying to undo or repair damages done. Our big quarrels end with attempts at reconciliation and forgiveness as often as not. Our big quarrels usually end without any attempts to undo or repair damages done. Our big quarrels never end with attempts at repair and __ reconciliation. 108 2b. Suppose your spouse disagreed with you about how to end quarrels. HOw would you feel? ‘ Fl 'd feel very strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It would bother me some, but not enough to tell him (her). It wouldn't bother me at all. 3a. I think that ... ‘ F Our big quarrels are almost always directed at current actions and here and now situations. Our big quarrels are usually directed at here and now situations, but sometimes bring in sore points from past quarrels not relevant to the current situation. Our big quarrels usually bring sore points from past quarrels into the current situation. Our big quarrels almost always bring in sore points from past ' quarrels although they begin with here and now disagreements. 3b. Suppose your spouse disagreed with you about bringing up past quarrels in 1 here and now situations. How would you feel? I'd feel very strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It would bother me some, but not enough to tell him (her). It wouldn't bother me at all. d Questions 4, 5, and 6 are about other aspects of your relationship with your spouse. 109 4a. When we are alone and a particularly sensitive topic comes up, I think that ... 1— We We We We usually avoid directly discussing the problem. usually discuss the problem, but avoid exploring our deeper feelings about it. usually discuss the problem, stating how we feel but being careful not to get emotionally involved. usually discuss the problem, stating how we feel and not worrying if we get emotionally involved or not. 4b. Suppose your spouse disagreed with you about how to talk over sensitive topics. How would you feel? I'd feel very strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It ‘ would bother me some, but not enough to tell him (her). It wouldn't bother me at all. L 5. When we make important decisions affecting both of us, I think that ... 1- We We We We almost always compete with each other fOr the dominant position in a hostile atmosphere. usually compete with each other for the dominant position. usually balance the dominant positions between us with only occasional competitiveness. almost always balance the dominant positions between us, and have little desire to compete with or dominate one another. 6a. When we make important decisions affecting both of us, I think that ... 1 FW e We We We cooperate very little, seldom helping one another or working together. cooperate about some decisions some of the time. cooperate on most decisions most of the time. cooperate almost always, helping each other verbally and in activities. 110 6b. Suppose your spouse disagreed with you about the way to reach decisions on important matters. How would you feel? 7I'd feel very strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It would bother me some, but not enough to tell him (her). It wouldn't bother me at all. _L The next question is a little different from the others. Below is a list of topics that most partners talk about at one time or another. Some topics are easier to talk about with your spouse than others because they are less sensitive. For each of the topics listed below, put a check in the column at the right which indicated how easy or difficult it is to bring the topic 2p_ip_conversation with your spouse. 7a. I think that the topic ... Very Easy "So—So" Diffi- Very Easy cult Diffi- cult "How to deal with inlaws" is "Right and proper conduct at parties" is "Who decides when to have sex relations" is "When and how to show affection" is "Deciding about matters of recreation" is "Handling family finan- ces" is "Which friends to visit or invite over" is 111 7b. Suppose your spouse disagreed with you about what can and cannot be talked over. How would you feel? WI'd feel very strongly, enough to try to change his (her) mind. I'd be bothered enough to tell him (her) so. It would bother me some, but not enough to tell him (her). It wouldn't bother me at all. all— Please go on to the Next Section Section II 1. On the line below, circle the dot which best describes the degree of happiness, everything considered, of your present marriage. The middle point, "happy," represents the degree of happiness which most people get from marriage, and the scale gradually ranges on one side to those few who are very unhappy in marriage and on the other, to those few who experience extreme happiness and joy in marriage. Very Happy Perfectly Unhappy Happy 112 Estimate the approximate extent of agreement or disagreement between you and your spouse on the following items 2 through 9. Please make one check for each item. Always Usually Occasion- Often Always Agree Agree ally Disagree Disagree Disagree 2. Handling family finances 3. Matters of recreation 4. Demonstrations of affection 5. Friends 6. Sex relations 7. Right, good or proper conduct 8. Philosophy of life 9. Ways of dealing with inlaws For questions 10 through 15 please check the one response which best answers the question. 10. When disagreements arise, they usually result in: Husband giving in Wife giving in Agreement by mutual give and take 11. Of your outside interests, how many do you engage in with your spouse? All of them Some of them Very few of them None of them 113 12. In leisure time do you generally prefer: To be on the go To stay at home Does your spouse generally prefer: To be on the go To stay at home 13. If you had your life to live over, do you think you would: Marry the same person Marry a different person Not marry at all 14. Do you ever wish that you had not married? Frequently Occasionally Rarely Never 15. How frequently do you confide in your spouse? Almost never Rarely Often Almost always Section III 1. Circle the number which represents the highest grade of schooling which you had completed at the time of your present marriage. 1 2 3 4 5 6 7 8 1 2 3 4 1 2 3 4 1 2 3 4 Grade School High School College Postgraduate For the remaining questions in this section, please check the one response which best answers the question. 2. Check the number which represents your age at the time of your present marriage. 19 and under 20 - 24 25 - 30 31 and over 114 How long did you "keep company" with your spouse before marriage? 1 to 3 months 3 to 6 months 6 months to 1 year 1 to 2 years 2 to 3 years 3 years or longer How long did you know your spouse at the time of marriage? to 3 months to 6 months months to 1 year to 2 years to 3 years to 5 years years or longer since childhood (DOME-'00)!“ My father and mother ... Both approved my present marriage Both disapproved Father disapproved Mother disapproved My childhood and adolescence, for the most part, were spent in: Open country A town of population 2,500 or under A city of 2,500 to 10,000 10,000 to 50,000 50,000 and over Did you every attend Sunday school or other religious school for children and young people? No Yes, if YES, at what age did you stOp attending such a school? Before 10 years old 11 to 18 years old 19 and over Still attending Religious activity at the time of your present marriage: Never attended church Attended less than once per month Once per month Twice per month Three times per month Four times More than four times 10. 11. 12. 13. 14. 115 Indicate the number of your friends of the same sex before your present marriage: Almost none A few Several Many Before marriage how much conflict was there between you and your father? None Very little Moderate A good deal Almost continuous Before marriage how much attachment was there between you and your father? None Very little Moderate A good deal Very close Before marriage how much conflict was there between you and your mother? None Very little Moderate A good deal Almost continuous Before marriage how much attachment was there between you and your mother? None very little Moderate A good deal Very close Give your appraisal of the happiness of your parents' marriage: Very happy Happy About averagely happy Unhappy Very unhappy 15. 16. 17. 18. 19. 20. 21. 116 My childhood on the whole was: Very happy Happy About averagely happy Unhappy Very unhappy In my childhood I was ... Punished severely for every little thing Punished frequently Occasionally punished Rarely punished Never punished In my childhood, the type of training in my home was ... Exceedingly strict Firm but not harsh Usually lax Always lax Irregular (sometimes strict, sometimes lax) What was your parents' attitude toward your early curiosities about birth and sex? Frad