— -——_. . _. v. ‘v—. _ NEED TO INFLUENCE: A PREDICTIVE MEASURE. 0F 7 PERSUASIVE BEHAVIOR? ' Thesis for the Degree of M. A MTCHTGAN STATE UNIVERSITY ANN CHRISTIANSEN 1968 mull Rf: -"":-.. 1" .- 3 K 3‘. .5 ; I; - E: I. ‘ o' (5.: . i :I ”:3.” 52.; ..., «name a. ‘6 37-“ “'3 *- wu‘ ’ L RETURNING MATERIALS: )V153I_] 1 Place in book drop to IJBRARJES remove this checkout from .‘l-IEIIIIL your record. FINES will be charged if book is returned after the date stamped below. NEED TO INFLUENCE: A PREDICTIVE MEASURE OF PERSUASIVE BEHAVIOR? By Ann Christiansen A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1968 ABSTRACT NEED TO INFLUENCE: A PREDICTIVE MEASURE 0F PERSUASIVE BEHAVIOR? by Ann Christiansen Communication research has studied the personality correlates of attitude change rather extensively. A complimentary focus studies the personality characteristics of the persuador, or influence agent. The motivational state of Elnfluence is one such characteristic that is of interest to the communication scholar. Knowing the strength of an individual‘s Need to Influence should enable the researcher to predict an individual's propensity to engage in persuasive behavior. Such predictions require some measure of Elnfluence and a demonstration that Elnfluence is related to persuasive behavior. The present study attempted to validate the ability of Uleman's (1965) Elnfluence measure to predict certain aspects of persuasive behavior. The eXperimental situation consisted of a persuasive task. Subjects were instructed to attempt to persuade a trained confederate to the subject's point of view on fluoridation of public water supplies. Two independent variables were employed: High and Low Elnfluence and High and Low External Motivation. The dependent variables were measures of the performance times of individuals, the reported degree Ann Christiansen of enjoyment and difficulty of the task, and the reported willingness of the subjects to participate in such a study again. The hypotheses predicted: 1) an interaction between Elnfluence and External Motivation on subjects' performance time such that the High Elnfluence: High External Motivation condition would spend more time trying to persuade the confederate than the High nlnfluence: Low External Motivation or the Low nlnfluence: High External Motivation conditions, and all three conditions would spend more time on the task than the Low nlnfluence: Low External Motivation condition; and 2) three Elnfluence main effects, such that when compared with the Low Elnfluence condition, the High Elnfluence condition would a) enjoy the task more, b) find the task less difficult, and c) be more willing to participate in such a task again. The results of the study do not provide any evidence to validate Uleman's nlnfluence measure., No significant differences were found on any of the dependent measures. However, when the Elnfluence measure was trichotomized an interaction between nInfluence and External Motivation on the performance time emerged (R(.3.O). The form taken by this interaction, however, was not as predicted. These findings suggest not only that the Elnfluence measure may be sensitive enough to detect differences in the motivational states of individuals, but also that motivation and behavior are completely related to each other. Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree. MW Kflfifiw ’Eirector of Thesis ACKNOWLEDGMENTS I would like to thank Dr. Gerald R. Miller for chairing my thesis committee and directing my thesis study. Thanks also go to the rest of my committee, Dr. Randall P. Harrison and Dr. Lawrence SarbaUgh, for their constructive evaluations and criticisms. As always, a thesis is more of a group effort than the work of one individual. My thanks go to the coders, Karen Pawlovich and Nancy Piet; to the researchers, Dave Beatty, Elizabeth Hanney, Betsy Niman, Gus Alexander, and John Frahm; and to the confederates, Jan Hyatt, Sandy Schaeffer, Jane White and Christy Tabb. Also many thanks to my typists, Shirley Sherman and Claudia Gean. ii TABLE OF CHAPTER I INTRODUCTION . . . . II METHODS . . . . . . III RESULTS . . . . . . IV DISCUSSION . . . . . BIBLIOGRAPHY . . . . . . . . [-10 Ho CONTENTS [-3. 15 21 36 1&3 TABLE l. 10. 11. LIST OF TABLES Page Means and Two-way Analysis of Variance of Responses to Item 5: "To what extent did you succeed in persuadin this person of your point of view?" . . . . . . . . . . Means and Two-way Analysis of Variance of ReSponses to Item 9: "How 'persuasible' (or susceptible to influence do you think the person you spoke to is?" . . . . . . . Means and Two-way Analysis of Variance for Change Scores derived from Items 6 and 7: "What do you think this person's attitude about fluoridation was when you first went in? (Item 6)...was at the end of the time you spent Withhim (her)? (Item7)" 00000000000000 Means on Item 6 (pre-persuasion) and Item 7 (post- pemuasion) C O O C C C C C O O C 0 O O C C O O C O 0 Means and One-way Analysis of Variance on Average Time EaCh confederate Spent With SUbjeCtS o o o o o o o o 0 Distribution of Subjects/Treatments to Confederates . . Means and Two-way Analysis of Variance for Subjects' Performance Time (in minutes) . . . . . . . . . . . . . Means and Two-way Analysis of Variance on Subjects' Performance Time with nInfluence Trichotomized . . . . Means and Two-way Analysis of Variance for Subjects' Performance Time deleting the Middle Third of the BInfluenceVariable.................. Contingency Tables for Frequency of Subjects Ranked High or Low by Time Spent with Confederates . . . . . . Contingency Table and Chi-Square for the Relationship Between Subject's Sex and Classification on the EInfluenceMeaSure0.0000000000000000 g 23 ) 21+ 25 26 26 28 29 30 31 32 35 LIST OF APPENDICES Page APPENDIX A: Pre- and Post-Experimental Questionnaires............. 4” APPENDIX B: Experimental InStPUCtiOnS o o o o o o o 58 CHAPTER I INTRODUCTION Many recent communication studies have focussed on attitude change. These studies have manipulated such variables as speaker's credibility, message elements, and the psychological states of the individuals to determine the conditions which Optimize attitude change. Attitude change is one element of a process termed persuasion, and relates to this process primarily from the message receiver's point of view. But source characteristics are also of interest in this process. One variable that enables the communication scholar to look at the source in the persuasion process is nInfluence, or Need to Influence. The process of persuasion can be defined somewhat syno- nymously with the power relationship: A exercises power over (or persuades) B if A gets B to do something which B would not have done if A were not present. Power is a more inclusive concept than persuasion in terms of the means control generally used to accomplish a given effect. But what is important when power and persuasion are defined the same way is the relationship between an influence agent and some receiver. Nevertheless, such a definition says nothing about the internal motivational states of the influence agent or the persuador. When we turn to the notion of nPower or E_Influence, we are no longer talking about the relationship between A and B which leads to a given effect. Rather we are now considering which of A's personality characteristics or needs make him the persuador in the process rather than the persuadee. Holding constant differences in status or power that derive from situational variables, what characteristics lead an individual to assume the influential role? Thus, gInfluence is a motivational state, rather than an existing power structure. It is of interest because it is, by definition, interpersonal in nature, and it is concerned with predicting communication behavior, notably propensity to engage in persuasive activities. Much of the research on the variable REESE has emphasized environmental or structural variables that determine the kinds of interactions that take place between individuals. A conceptualization of nPower or nInfluence as a motivational date, however, necessitates a measure of individual differences. One approach to measuring the strength of this motivational state follows the lead of David McClelland and his associates. McClelland developed a content analytic scheme for determining the strength of Need for Achievement from stories written in response to various stimulus pictures. Joseph Veroff was the first to apply such a method to measure EPower. In 1965, James Uleman modified Veroff's categorization scheme. We are primarily concerned with Uleman's measure. Conceptually, nPower and nInfluence are quite similar. Veroff (1958) defines EPower in this way: ...[S]ocia1 power is a dimension to be considered in the analysis of interpersonal relationships. In [this] sense the power motive will be considered that disposition directing behavior toward satisfactions contingent upon the control of the means of influencing another person(s). (p.105) Uleman's (1965) concept of nInfluence is synonymous with Murray's EDominance: Epominance includes the desires 'to control one's human environment. To influence or direct the behavior of Os [i.e., others] by suggestion, seduction, persuasion, or command. To dissuade, restrain, or prohibit. To induce an 0 to an act in a way which accords with one's senti- ments and needs. To get Os to cooPerate. To convince an 0 of the "rightness" of one's opinion.) (p.10) Uleman frequently substitutes the term nPower for the term EInfluence, and thus uses the two interchangeably. This practice has been adopted in the present discussion. The question of what kinds of behaviors a measure of nPower or nInfluence might predict is relevant here. Veroff (1958) states that "before one can make these predictions about the relationship of power motivation to behavior, a means of measuring the strength of the motive has to be established" (p. 106). But Veroff does assume that there is a positive relationship between seeking elective office and the strength of the power motive. Uleman's (1965) description of the High EInfluence individual differs somewhat: Those high in gPower show only slight tendencies to seek elective office. However, they are rated as more dominant by their peers; they score higher on a well validated scale of dominance; they influence others more when given the Opportunity to exercise power; and they report more comfort in positions of power. Generally they seem to in- fluence others for a wide variety of reasons, with or without recognition, and do it for its own sake. (p. 219) Neither Veroff nor Uleman used these measures as independent variables in developing them. For this reason, any demonstration of the validity Of such scales depends on using the measures as predictor variables. Three studies employing Veroff's measure of pPower are relevant to the exercise of power, particularly to persuasive attempts and to interpersonal perception in terms of power. The first study is by W. J. McKeachie (1961). McKeachie's interest in nPower is primarily its relation to learning in the academic situation. He defines nPower as Veroff does. One of McKeachie's major hypotheses was tested in a naturalistic classroom situation. His supporting rationale and hypothesis were as follows: It seemed likely that a class in which the teacher encouraged students to volunteer their ideas freely [as Opposed to ideas elicited by questions from the teacher] would be one which would allow the student with high power motive to attempt to convince others of his point Of view, and thus one in which he would be motivated. We hypothesized (weakly) that the student high in power moti- vation would achieve well in classes characterized by a high proportion Of student volunteering. (p. 129) The hypothesis was tested by use of a modified Bales interaction coding system deve10ped by Mann (1959). Two observers in the class recorded student and teacher responses in eight categories: positive reactions, tension release, suggestions, orientation-opinion, asking for orientation or Opinion, asking for support, tension, and negative tensions. In addition, behaviors related to power motivation were noted. However, McKeachie does not specify what these behaviors were. The criterion for achievement was the course grade. McKeachie's findings support the hypothesis: "Males high in the power motive do get better grades in classes in which student volunterring is prevalent than in classes in which it is uncommon. The converse is true for males low in power motive" (p. 129). However, these findings do not generalize to females. McKeachie Speculates that "power motivation is not emphasized as an important aspect of socialization of gins, or perhaps assertiveness is not a behavior instrumental to satisfaction of power motivation for women" (p. 130). The second study focussed on an aSpect of interpersonal re- lations, generally labelled interpersonal inference. D. E. Berlew's (1960) study concerns three motivational variables: pAchievement, EAffiliation and pPower, and their relationship to accuracy in the process of interpersonal inference. His main hypotheses state that: [First]...persons are able to judge others more accurately relative to content areas that are moderately salient to them than content areas that are either highly salient or not at all salient to them. [And secondly,] the more salient a particular area of social interaction for a subject, the more he will tend to use categories relative to that area in classifying information about other peOple. (p. 20) Following a group problem solving situation, coded using Bales' Interaction Analysis, subjects were asked to volunteer three adjectives that described each of the members Of the group, including themselves. These adjectives were sorted according to their relevance to power, achievement and affiliation. Pre-measures on TAT instruments of power, achievement, and affiliation were taken as well as rankings of social values. These were then correlated with the frequency of use of power, achievement or affiliation adjectives. Results for the power measures were as follows: ...The Median Test indicated the presence of a negative relationship between power-motivation and power free- adjective scores.... The data, then, demonstrated a relationship exactly the reverse of what was predicted. Subjects with strong power-motivation tended to use fewer rather than more power related categories in classifying group members. The correlation with value scores with the free- adjective distribution scores revealed only one Of the predicted positive relationships: a correlation...between power value scores and power free-adjective distribution scores. Subjects who place a high value on power tend to use power-related categories to the exclusion of cate- gories related to achievement and affiliation, just the reverse Of subjects who have a strong power motive. (p. 70) Finally, Martin L. Hoffman (1963) studied.pPower in the parent- child relationship, notably in the quality of the disciplinary actions taken by the parent toward the child. He defines power "as the potential to compel others to act contrary to their desires" (p. 869). He posits that in a situation where one party is in a position to influence another, the influence methods chosen will differ as the strength of the power motive varies. That is: The greater one's power, then, the more one's own needs and values can determine the kinds of control techniques used. Thus, it is with high power persons that we can most confidently expect to find relations between personality and control techniques. (p. 869) Hoffman suggests two variables which determine choice of influence method: authoritarianism and need for power. The hypotheses studied were 3 O... ...The parent's authoritarianism and power need related positively to the three following indices Of power as- sertion: the frequent use Of unqualified power assertion as an initial technique,_the frequent use of unqualified power assertion in reaponse to the child's non-compliance to a prior technique; and the frequent attemptuto control the child's behavior, apart from the particular technique A fourth measure was the child's behavior in reSponse to the parents' power assertion. Specifically, it was hypothesized that the greater the parental attempts at control, the greater the child's resistance to peer group control attempts. The measure of power needs was Veroff's TAT projective technique. The_findings show an interaction between economic class and between the sexes. For middle class mothers, there was a significant relation between authoritarianism and 1) power orientation and 2) child's resistance to peer group influence. Middle class fathers are significantly lower than lower class fathers on unqualified power and reactive unqualified power assertion. For lower class fathers, significant relationships were found between authoritarianism and 1) initial power assertion, 2) reactive power assertion, and 3) resistance by child to peer influence. For both grOUps of lower class parents the relation between power need and power assertion tended to be negative... . Though non-significant, they cannot be ignored in view of the positive authoritarianism findings Obtained for these subjects and the fact that only in this group is there a negative relation between authoritarianism and power need.... (p. 876) The conclusions drawn by Hoffman suggest that "while the projective measure may be valid in the middle class, it may not be valid in the lower class" (p. 876). However one chooses to interpret the findings of all three studies, at least one generalization is possible. Measures of nPower are complexly related to the actual exercise of power. They are complicated by sex differences and economic class. The findings in reverse of the hypotheses suggest one of two things: 1) if the measures of nPower are validly measuring a power motive then those concerned with power in their fantasy, and thus those who seem to want to exercise power, may be those who have no Opportunity or are inhibited in the exercise of power in their normal relationships. This may account for their preoccupation with power; 2) the measures of nPower are not walidly measuring the propensity to actually engage in influence behaviors. At this point, the better route to follow is to question the validity of the scales used before making the sorts of generalizations required by the first alternative. This is, in fact, what Uleman (1965) has done. The core Of Uleman's measure of'plnfluence is the use of a projective technique, namely responses to a set of TAT stimulus pictures (See Appendix A). His measure is a modification of Veroff's nPower. Uleman's modification of Veroff's measure arose out of a dissatisfaction with the behaviors which the measure predicted. As Uleman puts it, Veroff's scheme involves at least three theoretic problems: The first problem [involves] the role of status or prestige 'in the need for power.’ The second prOblem concerns the role of defensiveness and fear of being influenced by others. And the third problem centers around the means-end relationships. Is the 'need for power' independent of the ends or goals for which that power is used? (p. 3%) Relative to the second problem Uleman says: Veroff's measure...seems to get at the need to attain power which can be used to defend against others' influence at- tempts. Here, power is primarily a static institutionalized condition which one can attain, rather than being an effect upon others. It is used largely in self-defense. (p. 210) Thus Uleman implies that the asPects of EPower measured by Veroff do not adequately correspond to behaviors that current defi- nitions of pPower describe, however vaguely. Namely, Uleman's pPower focusses on that individual who attempts influence, not the individual who defensively counteracts the influence attempts of others. Uleman has tried to Specify the elements likely to predict the strength of Need to Influence. His measure consists of ten categories, each of which is scored for presence or absence in a TAT story. The first category (Power Imagery) is a discriminating category which determines whether or not the story deals with power. If this category is not scored, no other categories can be scored. If, however, power imagery is present, then there are nine other categories which can be scored for the story. The first five of these emphasis modifiers concern the parties in the action. They serve to identify certain aspects Of the parties: 1) prestige, in terms of social status, wealth, organizational position, special honors, or aspiration to high status; 2) participation of an lO organization itself, or a representative of an organization in the influence situation; 3) lack of self-deprication, or belittling of the parties; 4) absence of reminiscence, or non-functional daydreaming in the story; and 5) absence of dread, doubt, or apprehension, or absence of criticism of a future plan or course of action. The last four categories qualify the nature of the action- reaction sequence. These are: 6) counter-reaction, or the presence of an overt, distinctly identifiable reaction by one party to the action of the other party; 7) consultation, or an action situation involving the seeking or giving of advice, or mutual planning of a future event; 8) threat involving an action by one party which threatens some Object of importance to the other, and a reaction by that other party to neutralize the effect of the first party's action; and finally 9) separation, or the voluntary or forced departure of one party from the other. (See Uleman, 1965, Appendix J) These categories are not theoretically gmunded in the sense of being developed out of a theory of power. Uleman employed a gambling exPeriment to arouse the plnfluence motive in one group; another group formed the non-aroused group. This gambling eXperiment involved giving the subject control over the rewards of the other person in the situation (the confederate). The TAT was then administered to this group and the control group. A series of content analytic schemes were devised to differentiate the scores of the experimental group from those of the control group. As such, the pInfluence measure evolved primarily in the manner of a dependent variable, and has not been used as a pre- dictor variable. 11 One study that has used nInfluence as an independent variable is C. R. Berger's dissertation (1968). His study shows a relationship between nlnfluence and sensitivity to amount of external monetary rewards for persuasive behavior and a relationship between nInfluence and sensitivity to positive or negative feedback on the persuasive attempt. These results derive mainly from manipulation checks, rather than tests of the main hypotheses of the study. As such, the study bypasses a number of issues dealing with the persuasive attempt itself, and has not assessed the validity of the Elnfluence measure in predicting communication behavior. The first question is whether or not the nInfluence measure predicts propensity to engage in influence attempts. That is, is the nInfluence measure valid? From this follow other questions. They are questions of theory and operationalization. The notion Of nlnfluence as a motivational state deserves more attention. Uleman considers nInfluence a motivational state, if not {Jrimary or instinctive, then at least learned or conditioned. But regardless of the origin of the motivational Sate, nInfluence could be defined as an intrinsic source, rather than extrinsic somuce, of motivation for a given behavior. Exercise of the behavior itself, then, is rewarding for the individual. It is Obvious, however, that extrinsic sources of motivation, that is, sources of motivation independent of nlnfluence can elicit the same behavior. Therefore the study of nlnfluence must discriminate if possible behaviors which are unique to this particular motivational state. For this reason, external sources 12 of motivation must be explicitly manipulated and differences in behavior explored. Theoretically, what is the relationship between intrinsic and extrinsic sources of motivation? Do they operate independently or do they interact? What aspect of the influence attempt is rewarding-- the attempt, or the success of the attempt? If it is assumed that the attempt is rewarding, then expenditure of effort itself is rewarding. This may be a contradiction Of certain aspects of learning theories. The next question is, logically, what is the effect of feedback--success or failure of a persuasive attempt-- on the behaviors of High or Low pgnfluence individuals? Under conditions of success, the attempt may be intrinsically rewarding. But under conditions of failure, what happens? Qperationally, the same problems arise. What is an apprOpriate measure Of the presence of a state of motivation? It must be assumed, at this point at least, that the apprOpriate measure is some expenditure of energy. These questions are dealt with in the present study. First, the study sought to demonstrate the validity of nInfluence measure in predicting the persuasive behavior of an individual. In order to do this, two independent variables were employed: nInfluence and the presence or absence of External.Sources of Motivation. In order to separate the persuasive attempt from the elements of success or failure, feedback was held constantly negative for all communicators. The major 13 criterion variable was a measure of effort: specifically, the time spent performing a persuasive task. Therefore, we might suggest that this study sought to determine in part which element of the persuasive task is rewarding-- the attempt regardless of feedback, or success in the attempt. Additional dependent measures were reports Of enjoyment, difficulty in performing the task, and willingness to participate in such a study again. The major assumption of this study was that the nInfluence measure validly predicts propensity to engage in persuasive behavior. This assumption was tested via two theoretic hypotheses. Based on the definitions of the High nInfluence individual that characterize him as eager to exercise influence where he can and as finding such exercise of influence rewarding (Uleman, 1965), the first theoretic hypothesis was as follows: H1: a) When compared to Low nInfluence persons, High nInfluence persons will persist longer at an influence attempt regardless of success or failure. b) When compared to Low nInfluence persons, High nInfluence persons will enjoy the influence at- tempt more regardless of success or failure. c) Under a condition of failure to persuade, High nInfluence persons will be more willing to make 3 persuasive attempt another time than Low nInfluence persons. d) When compared to Low nInfluence persons, High nInfluence persons will find a persuasive attempt less difficult regardless of success or failure. 14 These hypotheses are based on the assumption that the rewards for the High nInfluence person making a persuasive attempt derive from the exercise of influence itself. Theoretically Low nInfluence persons do not derive the same rewards. Nevertheless, it is probable that some external sources of motivation can produce behaviors similar to that predicted by the state of nInfluence. We eXpect persons under a condition of High External Motivation to exert a higher degree Of energy in the persuasive attempt than those under a condition of Low External Motivation. But under conditions of failure in the attempt, we do not expect conditions of external motivation to account for differences in perception of difficulty, enjoyment, or willingness to attempt a persuasive task again. Differences on these variables should be a product of the intrinsic rewards involved in making an influence attempt. The second hypothesis parallels H1: a: H2: When compared to persons under a condition of Low External Motivation, persons under a condition of High External Motivation will persist longer at an influence attempt regardless of the success or failure. In effect, the prediction suggests an interaction between intrinsic and extrinsic motivation on persistence of the persuasive attempt. Main effects for the nInfluence variable are predicted for enjoyment, difficulty and willingness to attempt the influence again. Thus, this study seeks evidence to validate the nInfluence measure as a predictor of communication behavior. Such a demonstration is necessary before the measure can be extensively used in the area of communication research. CHAPTER II METHODS Subjects: Sewere selected from freshman and SOphomore classes in psychology, advertising, and physical science at Michigan State University. In all, 81 Se were originally pretested. Of these, 72 percent participated on a required or extra credit basis. The other 28 percent participated on a voluntary basis. The final design called for 60 Se; however, only 49 participated in the second part of the study. Of these, 8H percent participated on a required or extra credit basis. Of the #9 who participated in the whole study, 19 were male and 30 were female. Qperationalization of Independent and Dependent Variables: nInfluence: For each S) the measure of nInfluence was the cumulative score on four TAT stories. Uleman's coding scheme was used and provided a possible range from zero to 36 for each S, A dichotomization into High and Low nInfluence was made on the basis of the median score of the obtained range on the 60 SS scored. 15 16 External Motivation: External motivation was manipulated by a Time: The Enjoyment, message given to half the Ss. The determination of which gs received the message (High) and which did not (Low) was done by randomly assigning Se to High and Low External Motivation conditions at each of the two levels of nInfluence. The manipulation consisted of the following statement delivered individually to the §_during the experimental situation: Your instructor has indicated a great interest in your performance on this task. He has asked us to report to him how well you do in trying to persuade this person. amount of time Spent on the persuasive task was measured by the researcher, noting the time that the individual began the interview and the time that he emerged from the interview. The unit of measure was total number of minutes spent with the confederate. Difficulty and Willingness to Participate Again: Measures of these three variables were obtained from Ss' responses on a post-questionnaire to the following questions: 1. How much did you like or dislike the task you just performed? ReSponses were made on a five-point Likert-type scale ranging from "liked a great deal" to "disliked a great deal." 2. 17 How difficult did you find this task to be? Responses were made on a six-point scale ranging from "extremely difficult" to "extremely easy". The scale had no neutral position. 3. Would you participate in a similar study again if given the Opportunity? Responses were made on a five-point Likert-type scale ranging from "definitely yes" to "definitely no." Manipulation Checks: The post-questionnaire also contained items which enabled the researcher to determine if the _S_ perceived the confederate's behavior in the intended manner. These questions are: 1. To what extent did you succeed in persuading this person of your point of view? The response scale had four points ranging from "not at all" to "a very great extent." How "persuasible" (or susceptible to influence) do you think the person you spoke to is? Responses were made to a six-point scale ranging from "extremely persuasible" to "extremely un- persuasible." The scale had no neutral point. The third item consisted of two questions: a) What do you think this person's attitude about fluoridation of public water supplies was when you first went in? b) What do you think this person's attitude about fluoridation of public water supplies was at the end of the time you spent with him (her)? The response scales for both questions consisted of six points ranging from "extremely Opposed" to "extremely favorable." The scale had no neutral position. A change score was then obtained by finding the differences between the two scales. This change score was the reSponse used for this check. 18 Procedures: Pretest: Approximately two and one-half weeks prior to the experiment, a pretest was administered. The pretest consisted of the nInfluence measure (called the "Test of Creative Imagination" for purposes of disguise) and an attitude measure of six topics (See Appendix A). From these topics, one was selected which showed least variation across subjects: fluoridation of public water supplies. The lack of variation was desirable for the mechanics of the study rather than for theoretic considerations. Namely, it made the training of confederates easier, since they had to learn only one role. Assigning Conditions: §s were first assigned to High or Low nInfluence conditions on the basis of their scores on the TAT measure. Within these conditions, Se were then randomly assigned to High or Low Ex- ternal Motivation treatments. The EXperiment: The eXperiment designed to test the theoretic hypotheses involved a persuasive task. The §_was asked to report to a particular room where he was told that the §_was investigating the notion of persuasibility or susceptibility to influence. The §_was told that in an adjoining room there was another student (the trained confederate) who had completed the same pretests as the S, It was stated that this person's position on fluoridation of public water supplies was already known and was the Opposite of the st position. Each §_was told that his task was to try to persuade the person to accept st point of view on fluoridation of public water supplies. 19 In order to use time as a dependent variable, each §_was told that no one would interrupt him once he began his task, that it was entirely up to him to decide how long to stay. His decision, it was suggested, should be dependent upon his judgment of how likely the person was to change his opinion. If §_was up against a hopeless case, he could leave; if he felt he could succeed given a longer time, he could stay as long as he wished. The S was then escorted individually to a room where a confederate waited. During this time the researcher again emphasized that the §_ would not be interrupted and that he would have to decide for himself when to leave. The manipulation of external motivation was also done at this time. The researcher delivered the message concerning the report §_would make to‘st instructor on st performance on the task to High External Motivation §s but not to Low External Motivation Es. Timing began as soon as the door was closed and ended when the S came out of the room. Also, the conversations were tape-recorded. Originally, this was for purposes of testing a hypothesis about differing styles of persuasion. But incomplete data, due to technical difficulties, necessitated the elimination of this hypothesis. If the §_persisted in the task for thirty minutes, he was interrupted by the E_on grounds of a problem in the time schedule. Following the persuasive attempts, §_completed a questionnaire on the task and on the behavior of the confederate. The confederate also completed a brief questionnaire on the st performance. This completed the study. 20 Training of the Confederates: Four female confederates were hired. They were approximately the same age range as the gs, and were all women so as to hold constant the sex variable, at least to some extent. They were trained prior to the study to respond negatively to all influence attempts by the gs. This was done in a role play- ing situation where the researcher could comment on their performance. Instructions to the confederates limited the amount and kind of statements they were allowed to make to the S3 A uniform script of potential feedback was necessary to restrict variation across the confederates as much as possible (See Appendix B). Theuanipulation of feedback required that: l. The feedback could not indicate any agreement with the position of the S; that is, all feedback must indicate some degree Of disagreement. 2. No feedback could be given unless requested by the S3 so that the S was always the individual who initiated the influence attempt. 3. The confederate could not elicit by questions any state— ments from the Ss; that is, all feedback must be in the form of statements. u. The length of feedback should be uniform as much as possible from confederate to confederate to avoid biases in the time measure. Therefore, confederates were limited to replying with a) yes or no statements, where possible, and "I don't agree," "I don't think so," etc.; and b) content statements of no more than three to four senten- ces in length. Because the situation was unstructured to a great extent, these last three requirements became in effect recommendations. Adherence to these guidelines was checked by Ss' reports of the confederates' behavior. CHAPTER III RESULTS Coding and Reliability on the nInfluence Measure: Two coders trained for previous studies coded the nInfluence measure. Each of the two coders had correlated significantly with coders in these other studies, but their scoring had not previously been coded with each other. All of the TAT measures for those 60 subjects who had agreed to participate in part two of the study were coded. Not all of these subjects actually participated, but the Split into High and Low nInfluence was based on all 60 scores. A Pearson Product Moment Correlation for 28 measures done by both coders resulted in an r_of .58. Deleting the one worst pair of scores yielded an r_of .75. Given the scores on the nInfluence measure, which are prdbably ordinal rather than interval, a more appropriate reliability figure is obtained by using a Spearman Rank Order Correlation Coefficient. Rho on 28 pairs of scores was .57. Again deleting the one worst pair of scores yielded a Rho of .73. While an r_or Rho of .80 or better was desirable, the reliability coefficient obtained was judged sufficient for the study. A further check on the discrepancies in coding was made. Out of 28 cases coded by both coders, a total of eight disagreed on whether 21 22 a subject would be classified High or Low on nInfluence. Four were coded High by Coder l and Low by Coder 2, and four were coded Low by Coder l and High by Coder 2. This does not suggest any bias in the classification of subjects. The median score for the 28 nInfluence measures was 16 for Coder l and 12 for Coder 2. This discrepancy disappears when a median split is made for each range of scores. As already reported, the classifi- cation of subjects is unbiased. The median for the second coder. . however, on the other 32 scores was 15. This suggests that the coding sOheme as used by both coders was applied similarly, despite the discrepancies in medians on the original 28 scores. But the dis- crepancy does not influence the Split Of the total range of 60 scores into High and Low nInfluence. 0n the total range, the median was 16. This median did not change the classification of subjects from the split Obtained from the two smaller distributions. Checks on the Confederates: Before considering the eXperimental findings, several questions must be raised. First, did the confederates convey and attitude of disagreement with the subjects? Or in other words, was the feedback given consistently negative? Secondly, were there biases or constant differences in the times that each confederate spent with the subjects? The answer to both questions is yes. Subjects were asked on the post-questionnaire to answer some questions about the confederates. These were questions about how successful they had been in persuading the confederate, whether the confederates had changed their positions from the beginning to the end of the persuasion attempt, and how persuasible the confederates appeared to be. The results on these checks are as follows: 23 The first check was the response to the question, "To what ex- tent did you succeed in persuading this person of your point of view? (Item 5)" The response scale ranged from four ("to a very great extent") to one ("not at all"). Using a two-way analysis of variance, there were no significant differences among the four groups. The results are shown in Table l. Table 1. Means and.Two-way Analysis of Variance of Responses to Item 5: "To what extent did you succeed in persuading this person of your point of view?" External Motivation High Low nInfluence High 1.60 1.uo Low 1.70 1.20 Sources of Variance SS df_ MS F p nInfluence . 025 1 .025 < 1 NS External Motivation 1.225 1 1.225 2.15 NS Interaction .225 1 .225 < 1 NS Within 20. 5 36 o 569 "- -- Total 21.97 39 -- -- -- The overall mean (§'= 1.H8, sd = .7u) on this question indicates a point on the response scale between "not at all" successful (1.00) and successful "to a Slight extent" (2.00). This is satisfactory evidence that the confederates performed as instructed. 24 The second manipulation check involved the reSponse to the question, "How 'persuasible" (or susceptable to influence) do you think the person you spoke to is? (Item 9)" The reSponse scale ranged from six ("extremely unpersuasible") to one("extremely persuasible"). A two-way analysis of variance yielded no significant differences among the four groups. The means for the four groups indicate that the subjects perceived the confederates generaUW'as "slightly" (4.00) to "very" (5.00) unpersuasible. The results are shown in Table 2. Tane2. Means and Two-way Analysis of Variance of Responses to Item 9: "How 'persuasible' (or susceptible to influence) do you think the person you spoke to is?" External Motivation High Low nInfluence High H.40 H.30 Low H.00 “.80 Sources of Variance SS ‘df. MS F p nInfluence .025 1 .025 ( 1 NS External Motivation 1.225 1 1.225 ( 1 NS Interaction 2.025 1 2.025 1.3“ NS Within 5n.10 36 1.502 -- -- Total 57.375 39 -- -- -- The third manipulation check was based on two questions. The subjects were asked to report what they thought the confederate's 25 position on fluoridation of public water supplies was at the beginning of the interview and at the end of the interview. The response scale for each question consisted of six points ranging from six ("extremely Opposed") to one ("extremely favorable") toward fluoridation of public water supplies. A change score was derived from the two questions. If change occurred, the subject thought he had succeeded to some degree in changing the confederate's Opinion. A two-way analysis of variance for the change scores revealed no significant differences among the four groups. In addition, the means of the groups suggest that virtually no change in the confederate's position was reported. The response means for each question are also reported. They demonstrate that the confederates were perceived as opposed to the issue of fluoridation. For the most part these means represent the confederate's position as "extremely" (6.00) to "very" (5.00) opposed to water fluoridation. These results are given in Tables 3 and 4. Table 3. Means and Two-way Analysis of Variance for Change Scores derived from Items 6 and 7: "What do you think this person's attitude about fluoridation was when you first went in? (Item 6)...was at the end of the time you spent with him (her)? (Item 7)" External Motivation High Low nInfluence High .40 .60 Low .75 .20 Sources of Variance SS df_ MS F p nInfluence .006 l .006 4 1 NS External Motivation . 306 l . 306 4 1 NS Interaction 1.40 l 1.40 2.19 NS Within 23.03 36 o 639 -- '- Total 24.74 39 -- -- -- 26 Table 4. Means on Item 6 (pre-persuasion) and Item 7 (post-persuasion. External Motivation High Low nInfluence High pre 5.70 5.70 post 5.30 5.10 Low pre 4.90 5.70 post 4.85 5.50 The results from these manipulation checks show that the con- federates did act in accordance with instructions. However, the dif- ferences in average time that each confederate spent with the sub— jects varied from 22.18 minutes to 12.09 minutes. These differences were significant beyond the .05 level as tested by a one-way analysis of variance, randomly dropping subjects to equalize cell size. The results are shown in Table 5. Table 5. Means and One-way Analysis of Variance on Average Time Each Confederate Spent with Subjects Confederates l 2 3 4 Time 22.18 18.00 12.91 12.09 Sources of Variance SS d£_ MS F p Confederates 733.70 3 244.57 3.73 (.05 Within 2619.45 40 65.48 -- -- 27 The £_tests among the pairs of confederates Show that Confederate l is significantly different from Confederates 3 (t_= 2.73, d£_= 20, p<.05) and u (I = 3.00, if = 20, p<.05). Confederates 2, 3, and u are not significantly different from one another, although the difference between Confederates 2 and 4 approaches significance (t_= 1.68, 51.: = 20, p (.10). Confederates 1 and 2 are not significantly different from one another. This finding caused the researcher to check back on the treat- ment groups' time measures for evidence of bias. Originally this bias was controlled by rotating subjects from each treatment group to each confederate. But the time spent by subjects differed, necessitating a shift in the subjects' assignments according to which confederates were available at any given time. After subjects had been randomly dr0pped to equalize cell size, the number of subjects per confederate for each treatment group was tabulated. Although the second confederate was not significantly different from Confederates 3 and 4, Confederates l and 2 were grouped as High time and Confederates 3 and 4 were grouped as Low time. If such a grouping does not unduly distort the data, it can be suggested that subjects were relatively equally distributed to confederates. A Chi Square for 2 by k cells yields a non-significant figure (X2 -‘ <1). The groupings are shown in Table 6. 28 Table 6. Distribution of Subjects/ Treatments to Confederates Experimental Treatments Hi-Hi Hi-Lo Lo-Hi Lo-Lo Confederates' High 5 3 4 4 Time Low 5 7 6 6 Some bias is probably present, but not enough to account for the time differences among the groups. More will be said about this problem in the hypothesis testing section of this chapter. Test of Hypotheses: The main hypothesis predicts an interaction on the time measure between nInfluence and External Motivation, such that The High nInfluence: High External Motivation condition will Spend more time on the persuasive task than the High nInfluence: Low External Motivation or Low nInfluence: High External Moti- vation conditions, and all three conditions will be greater than the Low nInfluence: Low External Motivation condition. The statistical test of the hypothesis employed a two-way analysis of variance for groups of equal size. Because the cell sizes were unequal due to attrition, nine subjects were randomly drOpped from three cells to match the smallest cell. The n_for each cell was 10. Those subjects deleted for this analysis were also deleted for all subsequent analyses requiring cells of equal size. The analyses in- clude two subjects who suspected that the confederates had been coached. However, these subjects did not indicate any suspicion of the relation- ship of the task to the nInfluence measure. 29 The results of the analysis of variance do not support the hypothesis. The interaction was non-Significant, and there were no significant main effects due to either nInfluence or External Motivation. The results are shown in Table 7. Table 7. Means and Two-way Analysis of Variance for Subjects' Performance Time (in minutes) External Motivation High Low nInfluence High 17.60 13.90 Low 18.40 19.30 Sources of Variance SS df. MS F p nInfluence 96.10 1 96.10 1.11 NS External Motivation 19.50 1 19.60 <1 NS Interaction 52.90 1 52.90 (1 NS Total 3298.4 39 -- -- -- The lack of results suggests a reconsideration of the nInfluence variable. A study by C. R. Berger (1968) using nInfluence as a predictor variable in a dissonance study showed that a trichotomization of nInfluence into High, Medium and Low improved the predictive power of the variable. The trichotomization of the present data was accomplished by taking the upper, middle and lower thirds of the nInfluence scores for each condition of External Motivation. The statistical test used 30 to determine differences in mean times in the six conditions was a two-way analysis of variance. The results of this analysis are shown in Table 8. Table 8. Means and Two-way Analysis of Variance on Subjects' Performance Time with nInfluence Trichotomized External Motivation High Low nInfluence High 20.87 12.13 Medium 10.63 17.50 Low 19.38 18.75 Sources of Variance SS .d: MS F p nInfluence 200.04 2 100.02 1.29 NS External Motivation 8.33 l 8.33 (1 NS Interaction 488.52 2 244.26 3.17 NS Within 3315.02 43 77.09 -- -- Total 4011.92 47 -- -- -- The improvement in the interaction term, however ambiguous in interpretation, suggests that analyzing only the high and low thirds of the distribution might yield clearer results. The results of such an analysis are reported in Table 9. 31 Table 9. Means and Two-way Analysis of Variance for Subjects' Performance Time deleting the Middle Third of the nInfluence Vari ab 1e . nInfluence Sources of Variance ‘nInfluence External Motivation Interaction Within Total High Low SS 52.53 175.78 132.02 2153.14 2513.47 19.38 MS 52. 175. 132. 76. External Motivation High 20.88 Low 12.13 18.75 F p 53 (1 NS 78 2.29 NS 02 1.72 NS 89 -- -- As is apparent from the table the interaction term from the analysis of upper and lower thirds of the nInfluence variable is lower than the interaction term when all three thirds are included. This suggests that the middle third of the nInfluence variable is responsible for a considerable prOportion of the variance in performance time. Due to the lack of support for the hypothesis, one final analysis was carried out. The discrepancies in the average time each confederate :spent on the task have already been noted. been caused in part by these discrepancies. The lack of results may have Each subject's time was classified as High or Low according to the median time of the confederate he interviewed. A prediction paralleling the first hypothesis substitutes the variable number of subjects ranked High or Low for the variable time spent on the task: 32 When compared with Low nInfluence persons, High nInfluence persons will be ranked significantly higher on time spent with their respective confederates. This prediction was tested for each condition of External Motivation. The statistical test consisted in the combination of two Chi Squares from two fourfold contingency tables. (See McNemar, 1962, pp. 220-223) The results of the analysis are reported in Table 10. Table 10. Contingency Tables for Frequency of Subjects Ranked High or Low by Time Spent with Confederates Ranked by Time High Low External Motivation External Motivation High Low High Low nInfluence High 7 3 3 7 Low 6 5 4 S 2 2 X yc<1 xyc(1 The analysis does not show any relationship between the time spent, shown as High or Low time, and the Subject's treatment group. Therefore the time discrepancies among confederates do not account for the lack of significant findings on the major hypothesis. 33 The test of the remaining three hypotheses called for a test for differences between the means of two equal and independent groups. Five subjects were randomly eliminated from one group to equalize cell size. The first hypothesis predicts that: When compared to Low nInfluence persons, High nInfluence persons will enjoy the influence attempt more regardless of success or failure. The data were based on reSponses to the question, "How much did you like or dislike the task you just performed (Item 3)?" The response scale was a five-point Likert type scale ranging from five ("liked a great deal") to one ("disliked a great deal"). A preliminary two-way analysis of variance did not indicate the presence of an interaction between nInfluence and External Motivation. The results of the t test do not support the hypothesis (3 < 1, if = 42, NS). The individual means for the two groups are 3.50 (High nInfluence) and 3.09 (Low nInfluence). The overall mean on this question was 3.10 with a standard deviation of 1.30 (n=40). Regardless of the condition, subjects were relatively neutral toward the task they performed. The second hypothesis predicts that: Under a condition of failure to persuade, High nInfluence persons will be more willing to make a persuasive attempt another time than Low nInfluence persons. The data were based on responses to the question, "Would you par- ticipate in a similar study again if given the Opportunity? (Item 12)" The reSponses were scored on a five-point Likert-type scale ranging from five ("definitely yes") to one ("definitely no"). The possibility of an interaction was checked by a two-way analysis of variance. A 34 significant interaction did not emerge. The results of the t_test do not support the hypothesis (t