VERSITY LIBRARI ISE IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII III II 31293 0089 III This is to certify that the dissertation entitled Use of the Elaboration Likelihood Model to Predict Outcomes of Inpatient Chemical Dependency Treatment presented by Kenneth Gerard Dugan has been accepted towards fulfillment of the requirements for Doctoral Philosophy degree in IZ‘ (ii vi A: } 71d" M ajor professor Date February 18, 1993 MSU is an Affirmative Action /Equai Opportunity Insmution 0—12771 LIBRARY Mlchlgen State UnIversIty PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or beIore date due. DATE DUE DATE DUE DATE DUE II T" 73E 7 a: 1..“- :0. 47 “I MSU Is An Affirmative Action/Equal Opportunity Institution QWWJ .._’_._.—-, USE OF THE ELABORATION LIKELIHOOD MODEL TO PREDICT OUTCOMES OF INPATIENT CHEMICAL DEPENDENCY TREATMENT BY Kenneth Gerard Dugan A DISSERTATION Submitted to Michigan Sate University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Counseling, Educational Psychology, and Special Education 1993 ABSTRACT USE OF THE ELABORATION LIKELIHOOD MODEL T0 PREDICT OUTCOMES OF INPATIENT CHEMICAL DEPENDENCY TREATMENT by Kenneth Gerard Dugan One-hundred subjects were recruited from two inpatient chemical dependency treatment centers. Sixty-seven subjects completed treatment and 33 subjects dropped out of treatment. All subjects completed demographic, state motivation (for treatment), need for cognition, social desirability, attitude, and symptom-severity surveys. Subjects who completed treatment also responded to questionnaires which assessed their: confidence in treatment, perceptions of treatment as applicable to their problems, perceptions of their ability to understand therapeutic messages, judgements about the frequency of exposure to therapeutic messages, ratings of counselor characteristics, and post-treatment attitudes and symptom-severity. Data from treatment completers were used in bivariate correlations and multiple regressions on post-treatment recovery attitudes, behavioral intentions, and symptom severity using . Data from all subjects were entered into a discriminant function analysis to predict termination status. Message evaluations (confidence and applicability ratings), counselor characteristics (expertness, attractiveness, and trustworthiness), and frequency of exposure to therapeutic messages were correlated with post-treatment recovery oriented attitudes and behavioral intentions. Multiple regression analysis revealed that message evaluations and a two way interaction between need for cognition and counselor trustworthiness explained variance in post-treatment attitudes after controlling pre-treatment attitudes. Message frequency explained variance in behavioral intentions. After pretreatment distress was controlled, state motivation and need for cognition explained variation in post-treatment distress. The discriminant function analysis did not aid in the differentiation between treatment completers and premature terminators. These results supported hypotheses about expected patterns of correlation in the prediction of attitudes and revealed that although variables of interest did not collectively predict treatment outcomes, some variables of interest did explain variance in treatment outcomes. Copyright by KENNETH GERARD DUGAN 1993 This work is dedicated to ... all who strive and persevere. ACKNOWLEDGEMENTS I wish to acknowledge my appreciation to all of the individuals who assisted in the completion of this project. I am grateful to my advisor Dr. Robert W. Lent, who assisted in the conceptualization, methodological design, and statistical analyses of the study. He also provided valuable editorial assistance. I am also thankful to Dr. Frederick Lopez, Dr. Richard Prawat, and Dr. Robbie Steward who served on my guidance committee. They provided valuable commentary which aided in the development of this study, as well as, in the creation of the final draft of this dissertation. They helped me to consider information that I had not previously considered and to look at the data from different perspectives. The comments of Lesley Jones, who served on my guidance committee during the proposal stage were also valued. I also wish to thank Dawn Rodmann, Linda Hill, and Judi Ponzer who assisted in data collection. Thanks also goes to the administration and staff of the treatment centers were the data was collected. They graciously allowed me to come into their work environment to collect the data. Steve Raudenbush provided statistical consultation during the prOposal stage of this project and Dr. Randy Fotiu provided consultation regarding statistics and the use of the SAS computer system. I am indebted to numerous friends who listened as I ventilated feelings of frustration and distress through out the many months it took to complete this challenging task. My appreciation also goes to Terry Tomlinson for typesetting the figures. Finally, much debt goes to the many subjects in this study whose recovery I hope is rewarding and lifelong. vi Table Table Table Table Table Table Table Table Table Table Table Table Table LIST OF TABLES Postulates of the Elaboration Likelihood Model of Persuasion Means and Standard Deviations of Demographic Data Frequencies of Nominal Level Demographic Data Pre and Post Treatment Surveys Demographic Frequencies For Total Sample & Completers by Location Means and Standard Deviations by Site, All Subjects Means and Standard Deviations by Collection Site and for Both Sites Combined for Subjects Who Completed Treatment Correlations of Demographic, Predictor, and Dependent Variables Regression Analysis Predicting Post-Treatment Attitude Regression Analysis Predicting Post—Treatment Intentions Regression Analysis Predicting Post-Treatment Distress Discriminant Analysis of Termination Status Means & Standard Deviations of Message Evaluations by Treatment Mode vii 41 42 55 60 61 62 64 67 70 7O 71 74 LIST OF FIGURES Figure 2-1. This diagram of the central and peripheral routes depicts potential outcomes of exposure to a persuasive message. 10 Figure 4-1. Regression lines predicting post-treatment attitudes as a function of the interaction between ratings of counselor trustworthiness and Need-for-Cognition scores. 68 viii TABLE OF CONTENTS List of Tables List of Figures Chapter 1: Introduction Chapter 2: Review of Literature Models of Attitude Change The Elaboration Likelihood Model Experimental Evidence on the ELM Motivation and Ability to Elaborate in an Objective Manner Issue Involvement Need-for-cognition Message repetition Message comprehensibleness Motivation and Ability to Elaborate in a Biased Manner Peripheral Cues Versus Objective Or Biased Elaboration Issue-involvement and peripheral cues Need-for-cognition and peripheral cues Prior-knowledge and peripheral cues Consequences of Central Route Processing Temporal endurance of attitudes Prediction of behavior Need-for-cognition and attitude-behavior consistency Resistance to change Utilizing the ELM to Understand Counseling ix xi xii 11 11 13 15 15 16 17 17 18 18 18 19 20 20 21 22 22 Theoretical Extensions of the ELM to Counseling Empirical Applications Analogue and experimental studies. Pretreatment intervention studies. Actual-counseling studies. Attitude Change and Addictions Treatment Models of addiction. Motivation to overcome addiction. Research Questions and Hypotheses Chapter Summary Chapter 3: Methodology Subjects Subjects completing treatment Premature terminators Measures of the Independent Variables Demographic Data Sheet (DDS) The Counselor Rating Form-Short (CRF—S). Treatment Evaluation Questionnaire (TEQ) Marlowe-Crowne Social Desirability Scale Short Form Need for Treatment Survey (NTS) Need for Cognition Scale (NCS) Thoughts and Feelings List (TEL) Therapist's Evaluation Form (TEF) Measures of Dependent Variables Judgements of Drug Use and Recovery-Oriented Behavior Statement of Behavioral Expectations (SBE). Brief Symptom Inventory (BSI). Completion of Treatment Form. x 23 27 27 30 31 33 34 35 36 38 39 39 4O 40 41 41 41 43 45 45 47 48 49 50 50 51 52 53 Procedures Data Analysis Chapter 4: Results Comparison of Data Collection Sites Social Desirability Correlations of Predictor Variables with Treatment Outcomes Prediction of Post-Treatment Recovery-Oriented Attitude Prediction of Post-Treatment Behavioral Intentions Prediction of Post-Treatment Distress Prediction of Termination Status Supplemental Analysis: Correlations Among Precitor Variables Supplemental Analysis: Comparison of Treatment Evaluations Across Modes. Chapter 5: Discussion Bivariate Correlation Analyses Prediction of Post-treatment Attitudes Using Multiple Regression Prediction of Behavioral Intentions Using Multiple Regression Prediction of Post-treatment Distress Using Multiple Regression Prediction of Termination Status Using Discriminant Analyses Discussion of Supplemental Analyses Conclusions Limitations Implications for Future Research Implications for Counseling Practice Appendix A: Demographic Data Sheet Appendix B: Counselor Rating Form xi 53 56 59 59 63 63 65 69 69 70 71 71 75 77 79 82 82 83 83 85 86 88 92 94 95 Appendix C: Appendix D: Appendix E: Appendix F: Appendix G: Appendix H: Appendix I: Appendix J: Appendix K: Appendix L: Appendix M: Treatment Evaluation Questionnaire Marlow-Crowne Social Desirability Scale - Short Form Need For Treatment Survey Need For Cognition Scale Thoughts & Feelings List Therapist's Evaluation Form Judgements of Drug Using and Recovery Oriented Behaviors Statement of Behavioral Expectations Completion of Treatment Form Consent To Participate in Research Consent to Permit Therapist to Review Thought List LIST OF REFERENCES xii 96 100 101 106 108 110 112 115 117 118 119 120 The ELM & Treatment Outcomes CHAPTER 1: INTRODUCTION In the twenty-five years since Strong (1968) conceptualized counseling as an interpersonal influence process, more than 100 research articles have appeared in the counseling literature which have been guided by his and other social psychological models of attitude change. Numerous theoretical papers addressing this topic have also been published. Despite the popularity of this line of investigation, it does not appear to have substantially enhanced our understanding of how change occurs in therapy. In fact, reviews of the research in this area paint a complex and seemingly paradoxical picture of how therapeutic persuasion occurs (see Corrigan, Dell, Lewis, & Schmidt, 1980; Heppner & Claiborn, 1989; Heppner & Dixon, 1981). Heesacker (1986a) suggested that this situation is due to the preponderance of conflicting findings and to a reliance on social psychological theories of attitude change having poor generalizability to counseling. Another factor which has limited the utility of this research is that the vast majority of studies have examined only therapist or source characteristics (e.g., counselor credibility, attractiveness). However, interpersonal influence has long been recognized as also being dependent upon message variables and recipient characteristics. Heesacker (1986a) concluded that, in order to move beyond the present state of affairs, it will be necessary for investigators to use more sophisticated theories and research designs. Research on attitude change has progressed considerably beyond Strong's (1968) classic framework. One recent theory which has resulted from this work, the Elaboration Likelihood Model (ELM; Petty & Cacioppo, 1981),&provides a relatively systematic account of the role of source, The ELM & Treatment Outcomes 2 message, and recipient factors and their interactions in effecting attitude change. The model provides a clear theoretical foundation for describing and explaining conditions that facilitate or inhibit attitude change, and it also offers guidelines for the systematic investigation of relevant variables-(e.g., characteristics of the counselor, client, and therapeutic intervention). The most basic theoretical statement of this model is that individuals will experience a greater degree of attitude and behavior change under conditions in which the likelihood that they will cognitively elaborate persuasive communications is high than under conditions in which cognitive elaboration likelihood is low (Petty & Cacioppo, 1981). Elaboration likelihood is considered to be high when an individual is both motivated and able to process a given message. When this is the case, individuals are said to rely on a thoughtful consideration of issue-relevant arguments in forming their attitudes. Elaboration likelihood is viewed as being low when an individual is either unable or unmotivated to process a message. Under these circumstances individuals tend to rely on situational cues or simple decision rules in formulating their opinions. These two sets of circumstances are postulated to result in two distinct routes to persuasion, the central route (associated with extensive cognitive elaboration) and the peripheral route (associated with minimal cognitive elaboration). The central route is hypothesized to produce more enduring change (Petty & Cacioppo, 1981). The architects of the ELM have explicitly hypothesized that the model could be relevant to counseling (Petty, Cacioppo, & Heesacker, 1984). They also noted discrepancies which have arisen in research The ELM & Treatment Outcomes 3 based on Strong's (1968) application of Festinger's (1957) theory of cognitive dissonance to counseling and described how the ELM can be used to explain these discrepancies. The purpose of this investigation was to determine if predictions derived from the ELM and from social and counseling psychological research could be applied in a real-life counseling context. Four questions are of interest. First, can post-treatment attitude toward drug use and recovery behaviors be predicted from a combination of recipient, message, and source characteristics? Second, can post-treatment behavioral intentions be predicted from these same variables? Third, can post-treatment distress, i.e., symptom severity, be predicted from these variables? Fourth, can termination status (sucessful versus premature) be predicted using these variables? Although answers to these questions will not demonstrate causal relationships, they will provide evidence that either supports or challenges the ELM and its use in predicting treatment outcomes. The real-life counseling setting selected for examination of social influence and ELM variables was inpatient chemical dependency treatment. Two sites were utilized to collect data. Both sites were approximately 15 bed programs; one was located in the Midwest and the other was located in the Northeast. Both sites are described further in Chapter three. The rationale for selecting in-patient chemical dependency units lies in the conceptual compatibility between the ELM variables and variables important in chemical dependency treatment. For example, the ELM emphasizes the role of motivation, and motivation has long been recognized as a crucial element of the treatment of substance abuse. Moreover, the ELM highlights the cognitive processing of The ELM & Treatment Outcomes 4 information related to attitudes and behavior: inpatient chemical dependency treatment helps patients to examine information pertinent to their drug-related attitudes and behaviors. The second chapter will review: (a) research on the central and peripheral routes to persuasion, (b) research exploring the interaction of source, message, and recipient variables, and initial applications of the ELM to counseling, and (c) relevant findings from the addictions literature. The methodology (including subject selection, instruments used, and procedures) is described in Chapter three. Chapter four is comprised of the statistical analyses and results of this investigation. Chapter five contains a discussion of the results and suggestions for further study. The ELM & Treatment Outcomes CHAPTER 2: REVIEW OF LITERATURE The main purpose of this chapter is to review social and counseling psychological findings on the Elaboration Likelihood Model(ELM). This model, developed by Petty and Cacioppo (1981, 1986), is a social psychological model that integrates knowledge about attitude change processes and outcomes from a variety of social psychological models of persuasion. The advantage of this model is its ability to account for conflicting interpretations among other theories in a parsimonious manner. This chapter begins by identifying the theories of attitude change which predominated the field prior to the ELM. Concepts from several of these theories are integrated into the ELM. The description of the pre-ELM models is followed by a discussion of the ELM and the empirical data supporting its postulates, including work relevant to counseling psychology. Finally, relevant findings from the addictions literature will be reviewed, given the nature of the subject population employed in this study. Thus, this chapter is divided into three major sections: (a) a review of ELM theory and research, (b) a review of the theoretical and empirical applications of the model to counseling, and (c) a review of the pertinent substance abuse literature. Models of Attitude Change Petty and Cacioppo (1981) categorized the diverse models of attitude change into seven primary approaches. First, they discussed theories which emphasized basic principles of learning, such as conditioning and modeling (e.g., Bostrom, Vlandis, & Rosenbaum, 1961; Cialdini & Insko, 1969; Miller & Dollard, 1941: Rosenbaum & Tucker, 1962; Scott, 1957: Staats & Staats, 1957, 1958). These approaches The ELM & Treatment Outcomes 6 stress the manner in which targets of persuasion are either directly or vicariously reinforced for attitude change. Second, Petty and Cacioppo (1981) outlined the verbal-learning approach of Hovland, Janis, and Kelly (1953) which highlights the variables that influence a person's attention to, understanding of, yielding to, and memory of the arguments in a persuasive communication. A third set of approaches are the perceptual-judgmental theories (e.g., Ostrom & Upshaw, 1968: Sherif & Hovland, 1961) which emphasize how individuals perceive messages, and how opinions are influenced or distorted by past experiences. A fourth area of focus involved the impact of motivation and cognitive consistency on attitude change (e.g., Festinger, 1957). Fifth are information processing approaches (e.g., Bem, 1967, 1972) which focus on the attributions individuals make about their own behavior or about a communicator's actions. A sixth class of theories is composed of probabilistic models of how individuals receive, evaluate, and integrate information into attitudes (e.g., Anderson, 1971: McGuire, 1960). The seventh approach discussed by Petty and Caccioppo (1981) underscores information that people generate on their own, either upon exposure to a persuasive communication or in the absence of one. These models are called self-persuasion approaches (e.g., Janis, 1959, 1968; Tesser, 1978). The Elaboration Likelihood Model Unfortunately, the findings of various attitude change theorists regularly contradicted each other. Further, "because proponents of each approach attempted to explain a wide variety of phenomena, the different approaches often provided competing interpretations for the results of a particular experiment" (Petty & Caccioppo, 1981, p. 255). Also, results The ELM & Treatment Outcomes 7 within a paticular line of research were frequently inconsistent. In response to this disarray, Petty and Caccioppo (1981, 1986) developed a metatheorectical model of attidue change, called the "Elaboration Likelihood Model." The model was named for its basic tenets which are: (a) the degree and direction of attitude change that individuals experience depends on the amount, direction, and type of cognitive elaboration that people undergo when confronted by an attempt to persuade them: and (b) the probability that an individual will engage in cognitive elaboration of messages contained in a persuasive attempt depends on the individual's level of motivation and ability to process the information. After reviewing the prominent social-psychological theories of attitude change, Petty and Cacioppo (1981) asserted that, although each contributed to a full understanding of persuasion, none served as an adequate integrative theory. The ELM was created to provide a general, unifying framework for comprehending attitude change processes. "Elaboration-likelihood" is the probability that an individual who is the target of a persuasive attempt will cognitively elaborate, i.e., think about, examine, agree with, or dispute, the messages that are intended to cause attitude change. According to the ELM, persuasion will sometimes result from a careful examination of the issue-relevant arguments (i.e., as perceived by the message recipient), and sometimes from the presence of issue-irrelevant positive or negative cues (for example, association with an attractive message source). The former is called the central route to persuasion and the latter is called the peripheral route to persuasion. The postulates of the ELM are listed in Table 2-1. The central and peripheral routes to persuasion, and their potential outcomes, are depicted in Figure 2—1. The ELM & Treatment Outcomes Table 2-1. Postulates of the Elaboration Likelihood Model of Persuasion 1. People are motivated to hold correct attitudes. 2. Although people want to hold correct attitudes, the amount and nature of issue-relevant elaboration in which they are willing or able to engage to evaluate a message vary with individual and situational factors. 3. Variables can affect the amount and direction of attitude change by (a) serving as persuasive arguments, (b) serving as peripheral cues, and/or (c) affecting the extent or direction of issue and argument elaboration. 4. Variables affecting motivation and/or ability to process a message in a relatively objective manner can do so by either enhancing or reducing argument scrutiny. 5. Variables affecting message processing in a relatively biased manner can produce either a positive (favorable) or negative (unfavorable) motivational and/or ability bias to issue-relevant thoughts. 6. As motivation and/or ability to process arguments is decreased, peripheral cues become relatively more important determinants of persuasion. Conversely, as argument scrutiny is increased, peripheral cues become relatively less important determinants of persuasion. 7. Attitude changes that result mostly from processing issue-relevant arguments (central route) will show greater temporal persistence, greater prediction of behavior, and greater resistance to counter- persuasion than attitude changes that result mostly from peripheral cues O 1From Communication and Persuasion: Central and Peripheral Routes to Attitude Change (p.5) by R. E. Petty and J. T. Cacioppo, 1986, New York: Springer-Verlag. Copyright 1986 by Springer-Verlag New York, Inc. Reprinted by permission. The ELM & Treatment Outcomes 9 As shown in Figure 2-1, a potential attitude change situation is initiated when an individual is presented with a message or series of messages intended to alter a person's affective evaluation of an issue or topic. The model predicts that central route processing (i.e., cognitive elaboration) will occur only when the person is both motivated and able to examine the arguments presented. The motivation to process a message stems from such factors as an individual's need for cognition, issue involvement (which is determined by the extent to which accepting the argument will result in favorable or unfavorable consequences to the individual), commitment to change or not change, and cognitive dissonance stimulated by the message (Petty, Cacioppo, & Heesacker, 1984). Additionally, Petty and Cacioppo (1986) assumed that motivation stems, in part, from the natural tendency individuals have to strive to hold correct attitudes. Although Petty and Cacioppo (1986) do not define "correct" attitudes, it appears that they mean people are driven by a psychological need to hold Opinions which can be justified on a rational or pseudorational basis to self and others. The ability to examine or process arguments depends on such variables as the presence or absence of environmental distractions, message comprehensibility, and recipient characteristics, for example, the recipient's familiarity with the issue, and his or her intelligence (Petty & Cacioppo, 1986: Petty et al., 1984). If either the motivation or ability to engage in a careful examination of the issue-relevant arguments is absent, then a temporary shift in attitude is predicted to occur if a persuasive cue is present. If no cue is present, the subject will retain his or her initial attitude. When motivation and ability are both present the outcome of the persuasive attempt depends on the nature of the cognitive elaboration, The ELM & Treatment Outcomes 10 I PERSUASIVE COMMUNICATION I r _____ ‘ J Temporary | J attitude shirt L ————— J MOTIVATEO TO PROCESS? issue involvement, relevance YES commitment. dissonance arousal. NO need Ior cognition. etc. \ PERSUASION CUE PRESENT? YESl Sell-presentation motives. demand NO characteristics. evaluation apprehen- AIILITY TO PROCESS? sion. source characteristics. etc. Distraction. message comprehensibility, 1 issue iamihrity. appropriate schema. I tear arousal, etc vss[ NO NATURE OF COGNITIVE PROCESSING? (Initial attitude. argument quality. etc) Favorable Unfavorable Neither or droughts thoughts neutral predominate predominate predominate V " I' TTTTT ‘I I I I | Retain or regain | initial altitude COGNITIVE STRUCTURE CHANGE L ..... .I Are new cognitions adopted and NO stored in memory? Are diiierent responses made salient than previously? Lites Lvss (favorable) (uniavorabie) Enduring positive Enduring negative ' attitude change attitude change (persuasion) (boomerang) Figure 2-1. This diagram of the central and peripheral routes depicts potential outcomes of exposure to a persuasive message. Reprinted by permission from Communication and Persuasion: Central and Peripheral Routes to Attitude Chaege (p.5) by R. E. Petty and J. T. Cacioppo, 1986, New York: Springer-Verlag. Copyright 1986 by Springer-Verlag New York, Inc. The ELM & Treatment Outcomes 11 which is influenced by the subject's initial attitude, the quality of the argument, credibility of the source, and other variables. The thoughts resulting from cognitive elaboration will be predominately favorable, unfavorable, or neutral. If favorable thoughts predominate, then attitude change will occur in the direction of the advocated position. If unfavorable thoughts predominate, the recipient's initial attitude will be retained or attitude change will occur in a direction antagonistic to the advocated position. If neutral thoughts predominate, the outcome will depend on the presence of persuasive cues. Experimental Evidence on the ELM An extensive amount of research has been conducted to validate and extend the ELM (e.g., Cacioppo & Petty, 1982: Cacioppo, Petty & Morris, 1983; Cacioppo, Petty & Sidera, 1982; Petty & Cacioppo, 1979; 1984; Petty, Cacioppo, & Goldman, 1981). This section will review the research based on postulates 4 through 7 of the ELM (postulates 1 through 3 are axiomatic, therefore, they have not been empirically tested). Material relevant to postulates 4 through 7 is presented according to specific ELM hypotheses. Moreover, only those variables of direct interest to this study are discussed at length. Motivation and Ability to Elaborate in an Objective Manner The fourth postulate of the ELM indicates that variables which affect an individual's motivation or ability to engage in cognitive elaboration in an objective manner either enhance or reduce examination of messages. Cognitive processing occurs in a relatively objective fashion when the individual strives to remain impartial and is guided by the facts presented in an argument; the individual is both motivated to seek the truth and has the necessary information and opportunity to The ELM & Treatment Outcomes 12 examine data without bias. In contrast, biased elaboration occurs when circumstances lead the individual to favor or oppose a position on the basis of a pre-existing schema or when his or her knowledge slants the interpretation of arguments. Petty and Cacioppo (1986) identified four variables which enhance motivation to elaborate in an objective manner: issue involvement, need for cognition, personal responsibility, and multiple message sources. Only the first two of these four variables will be expanded upon in this review because they are variables investigated in this study; the latter two variables were not included. The personal responsibility and multiple message sources variables were excluded because this study was non-experimental (in contrast to prior ELM studies in which these variables were manipulated experimentally) and because less emphasis has been placed on these variables in the ELM literature. Another reason for not investigating the effects of multiple message sources was that it was expected that subjects would be exposed to approximately the same number of sources via hospital treatment procedures. Additionally, four main variables have been hypothesized to influence the ability to engage in relatively objective cognitive processing: message repetition, message comprehensibleness/complexity, message distraction, and recipient physical posture (standing versus reclining). Again, only the first two of these four variables are discussed below, because they are variables included in this study. Message distraction is a variable which previously has been studied only as an independent variable, and it was not included in this study because of the absence of experimental control. Recipient physical posture was not included because subjects‘ physical posture (i.e., sitting during treatment) was not expected to vary. The ELM & Treatment Outcomes 13 Issue involvement. This variable is defined as the extent to which a message has personal relevance or intrinsic significance for individuals, as when message recipients expect an advocated position to have salient repercussions upon their lives (Petty & Cacioppo, 1986). One early hypothesis was that increased issue involvement could enhance persuasion for pro-attitudinal advocacy (i.e., when the message advocates change in the direction with which the subject already agrees) and reduce it for counter-attitudinal advocacy (Petty & Cacioppo, 1979). Two studies have examined the hypothesis that high issue involvement enhances cognitive elaboration of the content of a persuasive message. In the first study (Petty G Cacioppo, 1979), issue involvement (high or low) was crossed with attitudinal congruence (pro or counter) regarding a persuasive message. When issue involvement was high, attitude change increased with the presentation of pro-attitudinal messages and decreased with the presentation of counter-attitudinal messages. It is possible these findings resulted from assimilation and contrast effects (cf. Pallak, Mueller, Dollar, & Pallak, 1972). That is, individuals may simply be more likely to accept (assimilate) proattitudinal messages and reject (contrast) counter-attitudinal positions (Petty & Cacioppo, 1986). Thus, it was necessary to conduct a more stringent test of the hypothesis. In Petty and Cacioppo's (1979) second study, issue involvement was again manipulated but both messages were counter-attitudinal. Half of the subjects were exposed to a message containing strong or cogent arguments: the remaining subjects were exposed to a message containing weak or specious arguments. Cogent arguments elicited positive responses, and specious arguments elicited counter-arguments. The The ELM & Treatment Outcomes 14 combined results of these two studies were interpreted to indicate that message position (counter- or proattitudinal) can bias cognitive elaboration while issue involvement per se influences message processing in an essentially objective manner. A corollary finding was that, under conditions of low issue involvement and counter-attitudinal advocacy, weak arguments were about as likely to produce attitude change as were strong ones (Petty & Cacioppo, 1984: Petty, Cacioppo, & Goldman, 1981). These investigations, which demonstrated the interaction between personal relevance and argument quality, support the hypothesis that as issue involvement increases individuals are increasingly motivated to evaluate the issue-relevant arguments that are presented in persuasive communications. Petty and Cacioppo (1986) have shown that "... this increased processing can result in people showing a greater appreciation for the strengths of cogent arguments and the flaws of specious ones. Several cautions are in order, however, concerning the possible limitations of the effect" (p. 87). First, Petty and Cacioppo believed that in some situations personal interests are so intense that elaboration will be biased in preservation of the ego, or elaboration may cease in the interest of self-protection. Second, the pro or counter-attitudinal direction of a message may motivate acceptance or rejection given that all other variables are essentially equal. Third, in real life situations, issue involvement is likely to be confounded with prior knowledge, i.e., individuals will presumably be more knowledgeable and thus more able to process messages pertaining to personally relevant material. Need-for-cognition. An additional implication of the model is that people with a high "need-for-cognition" should be more motivated to The ELM & Treatment Outcomes 15 examine the issue relevant material contained in persuasive messages than would be individuals with a low need-for-cognition. Need for cognition refers to a psychological need to understand or make sense of one's world or to engage in and enjoy effortful mentation or problem solving. Cacioppo and Petty (1982) described a series of four studies in which they constructed and validated the Need for Cognition Scale (NCS). Cacioppo, Petty, and Morris (1983) provided evidence that individuals with a high need-for-cognition were more persuaded to accept counterattitudninal messages than individuals with a low need-for-cognition when cogent arguments were used. However, they were less persuaded when specious arguments were used. Petty and Cacioppo (1986) indicated that the effect of need-for-cognition is maintained when intellectual ability is controlled and that low need-for-cognition individuals are apparently more conservative in their willingness to put forth cognitive effort in processing persuasive communications. Messege repetition. Moderate message repetition is described as resulting in a two-phase process of attitude modification. In the initial phase, repetition of persuasive messages gives recipients an Opportunity to examine the merits of an advocacy in a relatively unbiased manner (Petty & Cacioppo, 1986). When an individual has adequately contemplated the associations and implications of the communication, the second phase is entered. In this phase affective evaluations are driven by boredom or reactance; both tend to lead to decreased acceptance of advocated positions either by biasing cognitive elaboration in a negative direction or by serving as a simple negative emotional cue (Petty & Cacioppo, 1986). The distinction between moderate and excessive repetition will vary depending on several The ELM & Treatment Outcomes 16 factors, such as recipient familiarity with the topic, and length and complexity of the message. In one investigation Cacioppo and Petty (1980a) found that message repetition enhanced recall of both strong and weak arguments, and that strong arguments resulted in more favorable thoughts. Moreover an argument quality x repetition interaction existed in which positive attitudes were associated with moderate repetition of strong arguments and negative attitudes were associated with moderate repetition of weak arguments. Cacioppo and Petty (1985) replicated these findings using different experimental tasks. According to Petty and Cacioppo (1986) the interaction of argument quality and message repetition is incongruent with alternate social influence models, for example, the verbal-learning approach (Hovland, Janis, & Kelly, 1953). Meseege comprehensibleness. Petty and Cacioppo (1986) have identified message complexity/comprehensibleness as another variable that has relatively objective effects on message elaboration. Viewed from the ELM perspective, "the strengths of cogent arguments and the flaws in specious ones should become more apparent as complexity is reduced and comprehensibility is increased" (Petty and Cacioppo, 1986, p. 76). This should hold true only when motivation to process an argument is high. When motivation is low, message complexity/ comprehensibility may act a cue to the validity of a position via the peripheral route. For example, a complex message may be associated with authority and act as a positive cue or it may be so complex as to cause confusion and induce a negative response. Motivation and Ability to Elaborate in a Biased Manner The fifth ELM postulate asserts the existence of variables which affect cognitive elaboration by producing a positive or negative The ELM & Treatment Outcomes 17 motivational or ability bias. Petty and Cacioppo (1986) identified several variables which act as motivational biases on elaboration. These variables include: forewarning, bogus personality feedback, and a high level of message repetition. Prior issue-relevant knowledge is the only variable identified which increases ability to cognitively process persuasive messages in a biased manner. These variables and biased cognitive elaboration were not investigated in this study and, therefore, are not discussed further. The effect of forewarning was not investigated because it was assumed that all subjects were forewarned. That is, since all subjects were recently admitted to an in-patient chemical dependency treatment unit, one may assume that they were aware that attempts would be made to influence their beliefs and opinions about the use of drugs. Bogus personality feedback was not studied because it contradicts ethical standards of treatment. Peripheral Cues Versus Objective Or Biased Elaboration The sixth ELM postulate assumes the existence of a tradeoff between cognitive elaboration and peripheral cues. Petty and CaciOppo (1986) suggested that the tradeoff between cognitive elaboration (objective or biased) and the influence of peripheral cues depends on the level of elaboration likelihood. On the basis of postulate six one would hypothesize that attitudes will be determined primarily by peripheral cues when elaboration likelihood is low. The impact of peripheral cues has been investigated in relation to subjects' degree of issue-involvement, need-for-cognition, and prior-knowledge. Issue-involvement and peripheral cues. Petty, Cacioppo, and Goldman (1981) carried out a 2 (issue involvement: high or low) x 2 (argument quality: weak or strong) x 2 (source expertise: inexpert or The ELM & Treatment Outcomes 18 expert) between-subjects factorial experiment. Source expertise was used as a peripheral cue based on earlier research (Hoveland, Janis, & Kelly, 1953) which suggested that the apparent expertise of a message source acts as a cue to the validity of a message. Petty et a1. (1981) found subjects to report more positive attitudes when argument quality was strong and when the message source was expert. They also found a personal relevance by argument quality interaction and a personal relevance by source expertise interaction which revealed that source expertise was a more important determinant of attitudes when personal relevance was low. This study was replicated using source likability as a cue, in place of source expertise (Petty, Cacioppo, & Schumann, 1983). An identical pattern of results occurred, including the personal relevance by cue (source likability) interaction. Need-for-cognition and peripheral cues. An additional prediction is that subjects high in need for cognition will rely less on peripheral cues than will subjects low in need for cognition. Cacioppo, Petty, Kao, and Hargitt (cited in Cacioppo & Petty, 1984b) manipulated the attractiveness of a source (by varying the makeup and clothing of the source presented on videotape) and found that subjects low in need for cognition were more affected by attractiveness in forming their opinion than were subjects high in need for cognition. Prior knowledge:and peripheral cues. The ELM tradeoff hypothesis has been supported in regard to subjects' prior knowledge of a topic (Cacioppo & Petty, 1980b: Wood, 1982). Peripheral cues have less impact when prior knowledge is high, but they enhance persuasion when prior knowledge is low. For example, Cacioppo and Petty (1980b) found that females (a) expressed more agreement than men both to accurate and The ELM & Treatment Outcomes 19 inaccurate evaluations of football tackles, and (b) showed less acceptance of inaccurate evaluations of fashions. They concluded that a lack of issue-relevant knowledge leads women to rely on simple decision rules "(e.g., 'As a woman, I should maintain harmony')" (p. 166). Perceptions or judgements of subjects about their own behavior have also been shown to act as peripheral cues. Wood (1982) concluded that individuals‘ perceptions of their own behavior had a more powerful influence on attitudes when individuals were less knowledgeable about and less able to evaluate a topic. Wood measured the knowledge of subjects regarding environmental conservation. At a later date she asked subjects to deliver a pro-environment message to two fellow students and to ask the fellow students to sign a petition. Half the subjects were provided an incentive (five dollars) and the other half were given no incentive. The proattitudinal position of low knowledge subjects decreased when they were given an incentive, while the attitudes of high knowledge subjects remained positive regardless of the incentive. Low knowledge subjects apparently used their behavior (in the context of the incentive) as a behavioral cue to infer that they did not fully agree with the proattitudinal position and were simply acting on the basis of the reward (Petty & Cacioppo, 1986). Consequences of Central Route Processing The seventh ELM postulate describes the consequences of persuasion occurring under high elaboration likelihood conditions. Specifically, central route processing is expected to result in greater temporal endurance of changed attitudes, a higher level of consistency between attitudes and behavior, and more resistance to change than peripheral route processing. Support for these predictions is discussed below. The ELM & Treatment Outcomes 20 Temporal endurance of attitudes. Petty and Cacioppo (1986) reasoned that, if extensive cognitive processing enhances the endurance of attitude change, then circumstances which promote cognitive elaboration ought to yield more attitudinal endurance than circumstances which reduce elaboration. This hypothesis was tested in a 2 (immediate or delayed assessment) x 2 (message type: negative cue plus weak arguments or positive cue plus strong arguments) x 2 (high or low issue-involvement) mixed model factorial design. The investigators induced essentially equivalent attitude changes via either the peripheral and central routes: the temporal persistence of attitudes was then measured (Petty, Cacioppo, Haugtvedt, & Heesacker, 1986; experiment 1). Opinions changed under central route conditions endured longer than opinions changed under peripheral route conditions. A message type x time-of-measurement interaction was found for low-involvement subjects. These individuals experienced a decay of the attitudes they had expressed immediately after message exposure; i.e., subjects who heard positive messages became less favorable and subjects exposed to negative messages became more favorable toward the advocated position at the time of delayed measurement. Petty et a1. (1986) interpreted their findings as supporting the ELM hypothesis about temporal persistence of attitude change under central versus peripheral processing conditions. Prediction of behavior. Petty and Cacioppo (1986) reviewed research supporting the hypothesis that opinions changed via the central route predict actual behavior better than opinions changed via the peripheral route. For example, Petty, Cacioppo, and Schumann (1983) found that subjects exposed to phony magazine advertisements for a disposable razor had more positive attitudes toward the razor and The ELM & Treatment Outcomes 21 expressed intentions to purchase the product more frequently when personal involvement (i.e., motivation) was high than when it was low. Also, Schumann, Petty, and Cacioppo (1986) increased subjects' ability to elaborate by allowing them more frequent opportunities to examine the issue-relevant tenets of a message endorsing the merits of using a particular ink-pen. Subjects reported engaging in more thought about the pen and stronger intentions to purchase it as message exposure increased. Petty and Cacioppo (1986) concluded that the data support the hypothesis that attitudes changed under conditions of high elaboration likelihood (i.e., high motivation and ability to process messages) are associated with increased behavioral intentions. Need-for-cognition and attitude-behavior consistency. Another relationship expected in the framework of the ELM is a higher consistency between attitudes and behaviors for people who are high in need-for-cognition (HNC) than for people low (LNC) in this need (Petty & Cacioppo, 1986). To test this hypothesis a survey was administered to 300 students eight weeks prior to the 1984 elections; attitudes toward, reported thoughts regarding, and knowledge about both presidential tickets were obtained (Cacioppo, Petty, Kao, & Rodriguez, 1986). A subset of HNC (n=61) and LNC (n=108) subjects was then selected. A larger percentage of HNC than LNC reported having voted, but the difference was not significant. In a test of the major prediction that attitudes are better predictors of behavior for HNC rather than LNC persons separate correlations between a preference index and reported voting behavior were calculated for each group. Analyses of the data for subjects who actually voted indicated that preelection attitudes predicted voting behavior (i.e., the candidate for whom the subject The ELM & Treatment Outcomes 22 voted) to a greater degree for subjects high (r=.86, _;40) as opposed to low in need for cognition (r=.41, e§4l), z=3.71, p(.Ol. Resistance to chaee_. A third hypothesis derived from postulate seven pertains to the differential resistance-to-change of opinions formed under central versus peripheral route conditions. Opinions formed through the central route are expected to withstand (or resist) counter-persuasion to a greater degree than Opinions formed through the peripheral route (Petty & Cacioppo, 1986). Specifically, the ELM predicts that individuals who form an attitude based on a peripheral cue (e.g., source attractiveness) will be less capable of defending and, therefore, maintaining their opinion than will individuals who form their attitude based on a thoughtful consideration of the merits of an advocated position. Petty et a1. (1986; experiment 2) found support for this hypothesis: subjects exposed to conditions that facilitate a high degree of elaboration were more resistant to counter persuasion than were subjects who initially demonstrated a similar amount of attitude change yet had not been exposed to conditions facilitative of cognitive elaboration. Utilizing the ELM to Understand Counseling Shortly after the ELM was first posited, its potential for illuminating the social influence process of counseling was recognized (Petty et al., 1984). Several authors (Cacioppo, Petty, & Stoltenberg, 1985; Heesacker, 1986a; 1986b: Heppner & Claiborn, 1989; McNeill & Stoltenberg, 1989: Stoltenberg, 1986) have advocated the use of the model to account for the apparent inconsistencies of previous research based on Strong's (1968) interpersonal influence model of counseling (see reviews by Corrigan et al., 1980; Heppner & Dixon, 1981; Heppner & The ELM & Treatment Outcomes 23 Claiborn, 1989). The ELM is likely to account for inconsistencies because it provides a more systematic explanation of the interactions of source, message, and recipient variables than did earlier theories of social influence. The counseling psychology literature pertaining to the ELM may be divided into theoretical treatises and empirical applications. Theoretical Extensions of the ELM to Counseling In a theoretical discussion of the relevance of the ELM for counseling Stoltenberg and McNeill (1984b) concluded on the basis of existing data that: (a) it is hazardous for counselors to rely on source cues to cause attitude change in the absence of issue-involvement because any changes would probably be transitory and (b) the quality of of counseling interventions (i.e., change messages) is crucial. Stoltenberg, Bratt, and McNeill (1985) elucidated the role that affect plays in facilitating or inhibiting cognitive processing. Dysphoric moods usually interfere with recall and processing. However, depressed individuals have been found to process and recall negative feedback and memories more efficiently than do happy individuals (Stoltenberg et al., 1985). Ellis (1983) suggested that depression has two primary effects on memory: (a) it yields a differential allocation of cognitive resources, and (b) it reduces the amount of task-relevant elabortation engaged in by a subject. Stoltenberg et al. (1985) identified depression as one type Of affective schemata. They hypothesized that when an affective schema is activated it triggers the subjective experience of the emotion and the activation of cognitive networks with which it is affiliated. Thus, happy moods are expected to facilitate processing of positive information and sad moods to The ELM & Treatment Outcomes 24 facilitate processing Of negative information. Stoltenberg et al. (1985) also hypothesized a curvilinear relationship between the ability or motivation to process data centrally and depression. Specifically, they suggested that clients are unlikely to be motivated to process information if they are not depressed; that motivation would increase as depression increased up to a given level; and that, as depression continues to increase, motivation and ability to process relevant therapeutic information is expected to decline. Heesacker (1986b) suggested that ELM concepts can be extrapolated to explain counseling processes. "The ELM holds promise in helping to resolve three existing problems in social influence research, namely, inconsistent findings regarding counselor characteristics, little behavior change resulting from client attitude change, and difficulty generalizing from social influence research in the lab to actual counseling" (Heesacker, 1986, p. 50). From an ELM perspective, source cues will fail to elicit attitude change consistently when (a) subjects process information via the central route and thus are not influenced by source traits or (b) subjects process data via the peripheral route but the source does not act as a noticeable attitude cue (Heesacker, 1986b). Heesacker (1986) favored the first (central route) explanation, and stated that there are at least two ways to understand the inconsistent findings using the central route explanation: inconsistencies may result from either (a) uncontrolled differences in subjects' thoughts toward the topic and toward persuasive arguments regarding the topic or (b) differences in level of subjects' issue involvement. Heesacker (1986b) suggested that the lack of connection between The ELM & Treatment Outcomes 25 attitude and behavior in traditional social influence research exists because investigators have not demonstrated that subjects were motivated or able to process information; nor have they shown manipulations of independent variables to produce clearly favorable or clearly unfavorable issue-relevant cognitions. Heesacker (1986b) concluded that it is likely central route processing did not occur and therefore behavior change did not coincide with opinion change in most counseling studies of social influence variables. Heesacker (1986b) also suggested that in order to enhance the generalizability (i.e., external validity) Of laboratory investigations, experimental designs must match the route of persuasion (central versus peripheral) which exists in the field. Other conditions needing to be met to improve generalizability include subject: motivation to change, and personal investment in the topic/attitude of study. Stoltenberg (1986) raised the question of how counselors motivate their clients. He identified a subject's "degree of decisiveness" on a given issue as an important aspect of motivation to examine issue-relevant arguments. A client who has already selected a course of action is less likely to be motivated to examine a topic than is an individual who is struggling to make a decision. Further, it may be assumed that clients are more likely to engage in a high level of cognitive processing when they view the issues in counseling as important or personally relevant. Perceived relevance can be enhanced by pointing out the (negative) consequences of continuing current dysfunctional behaviors compared to the expected (positive) consequences of alternative behaviors (Stoltenberg, 1986). Stoltenberg (1986) described three attempts (Dixon & Claiborn, 1981; Heppner & Dixon, 1978; The ELM & Treatment Outcomes 26 Heppner & Heesacker, 1982) to examine the influence of issue-involvement by measuring clients' perceived need for help and motivation for counseling: no differential effects of perceived need for help were discovered. Stoltenberg (1986) argued that counseling motivation or perceived need for counseling may not be a specific enough measure of issues that are personally relevant to clients in resolving their problems. For instance, some clients may be motivated to attend sessions regularly but not to examine issue-relevant arguments. Instead, Stoltenberg (1986) viewed Dixon and Claiborn's (1981) notion of a client's "commitment to change" as commensurate with personal relevance. Premature termination and a failure of attitude change to result in behavior change were seen as by-products of low motivation to cognitively elaborate therapeutic messages (Stoltenberg, 1986). Stoltenberg and McNeill (1987) suggested that when counselors present evidence regarding the potential benefits of change, address client values, and confront counter-productive beliefs, then client resistance can be minimized. They also noted that issue-involvement is topic-specific, thus a clients's degree of involvement can shift a great deal during a single session, depending on the extent of topic variation. Finally, they recognized that a client's level of intelligence, psychological mindedness, or both may affect his or her ability to process information in counseling (Stoltenberg & McNeill, 1987). McNeill and Stoltenberg (1989) compared Strong's (1968; Strong & Matross, 1973) Social Influence Model to the ELM and pointed out advantages of the ELM. Strong (1968) hypothesized that counselor power, The ELM & Treatment Outcomes 27 which is derived from credibility (a combination of expertness and trustworthiness), is the primary determinant of attitude change. In contrast, the ELM hypothesizes that message quality is the primary determinant of attitude change when clients are motivated and able to process information: peripheral cues, such as counselor expertness, are the primary determinant of attitudes when clients are unmotivated or unable to engage in cognitive elaboration. McNeill and Stoltenberg (1989) suggested that the ELM explains the impact of different depths of cognitive processing on attitude change, while Strong's (1968) interpersonal influence model only explains the effect of peripheral cues. Thus, Strong's model can be subsumed within the ELM. Empirical Applications A growing body of empirically based applications of the ELM reveals the utility Of differentiating between central and peripheral information processing in order to describe, predict, and experimentally control interpersonal influence variables within counseling. This body of literature consists of analogue studies and lab experiments (e.g., McNeill 8 Stoltenberg, 1988; Stoltenberg & Davis, 1989: Stoltenberg & McNeill, 1984), pretreatment intervention studies (e.g., Heesacker, 1986a), and investigations using actual clients (e.g., Heppner & Heesacker, 1988). Analggge and experimental studies. Stoltenberg and McNeill (1984a) examined the effects of issue-involvement and counselor expertise on perceptions of counseling, finding that subjects' level of involvement did have differential effects on perceptions of a counseling interview. Specifically subjects who were highly involved and who were exposed to a counselor of high expertise listed fewer negative thoughts The ELM & Treatment Outcomes 28 and expressed more agreement with the counselor's statements than did subjects who were highly involved but exposed to a counselor of moderate expertise. Additionally, subjects having low issue-involvement who were exposed to the high expertise counselor expressed less agreement than low issue-involvement subjects exposed to the moderately expert counselor. Finally, high issue-involvement subjects paid more attention to the interview and they agreed more with the high expertise counselor than did low issue—involvement subjects. These results were interpreted as tentatively supporting ELM predictions (Stoltenberg & McNeill, 1984a). McNeill and Stoltenberg (1988) used a 2 (high vs. low message quality) x 2 (high vs. low counselor credibility) x 2 (high vs. low issue involvement) factorial design to investigate: (a) favorableness of subjects' thoughts in response to an audiotaped counseling interview, (b) subject's agreement with the recommendations made in the session, and (c) their intentions to seek counseling. Statistical analyses indicated that subjects were influenced primarily by high-quality messages. Significant main effects of message quality were discovered in relation to all of the dependent variables. There were no main effects for credibility or issue-involvement. The presence of main effects for message quality suggested that central route processing had occurred. Significant source credibility x message quality and source credibility x level of issue involvement interactions were also found. The first interaction was characterized by a tendency for subjects to generate a larger number of favorable thoughts in response to a moderately credible source as opposed to a highly credible source, under conditions of high message quality. Also, subjects exposed to the The ELM & Treatment Outcomes 29 highly credible source and a high quality message listed a larger number of favorable thoughts than did subjects exposed to the moderately or highly credible source under low message-quality conditions. The second interaction was characterized by a tendency for low-involvement subjects who were exposed to the highly credible source to agree more with the way the session was conducted than did high-involvement subjects exposed to the highly credible source or low-involvement subjects exposed to the moderately credible source. The first interaction effect was interpreted by McNeill and Stoltenberg (1988) to be inconsistent with ELM predictions. This is a potentially erroneous interpretation, however, because the ELM suggests that source credibility is irrelevant when message quality and issue involvement are high, and it is not clear what level of issue involvement was present for the subjects whose data contributed to the interaction. Additionally, this study differed from the typical ELM lab study in that sources of high and moderate credibility were used as opposed to sources of high and low credibility. The ELM does not make specific predictions about the effects of moderate source credibility. The second interaction supported ELM predictions. Stoltenberg and Davis (1989) studied attitudes toward career and study skills information and found counselors of low credibility to be more persuasive than counselors of high credibility when issue involvement was low and cogent arguments were used. It was believed that more attention was paid to argument quality than source credibility cues. Finally, McNeill and Stoltenberg (1989) cited two investigations (Bratt & Stoltenberg, 1987; Stoltenberg, Leach, & Bratt 1988) which suggested that dysphoric moods increased central route processing The ELM & Treatment Outcomes 30 relative to euphoric moods when the issue was mood-congruent (i.e., counter-attitudinal for a negative mood). Pretreatment intervention studies. Heesacker (1986a) attempted to influence subjects to adopt more favorable attitudes toward counseling in a pre-counseling intervention. Undergraduates (N=254) categorized as relatively high or low in ego-involvement in regard to either social skills or career concerns participated in a pretreatment intervention comprised of strong or weak arguments and presented by a counselor of high or low credibility who promoted participation in either a social skill or career decision making group. Analyses indicated that favorable attitudes towards counseling increased as ego involvement increased and when intervention quality was high. Another result was a two-way interaction of intervention quality and counselor credibility. Subjects exposed to messages from a counselor low in credibility distinguished between weak and strong argument better than did subjects exposed to a counselor high in credibility (regardless of level of ego-involvement). Heesacker (1986a) claimed that these results can be more parsimoniously explained by the ELM than by other theories of social influence because the ELM predicts interaction effects for credibility and intervention quality while other theories, for example, Strong's (dissonance) theory, predict only main effects. In a similar study, Neimeyer, Guy, and Metzler (1989) examined ELM variables in a pre-treatment study of attitudes toward a cognitive restructuring technique aimed at the treatment of disordered eating. Subjects consisted of 107 undergraduate students who took part in a 2 (involvement: excess restraint of eating vs. low restraint of eating) x 2 (intervention quality: strong vs. weak) x 2 (source credibility: high The ELM & Treatment Outcomes 31 vs. low) between-subjects factorial design. Analyses of the number of positive and negative thoughts listed in response to the intervention revealed significant main effects for intervention quality. Subjects exposed to the high quality intervention listed a significantly larger number of positive thoughts than did subjects exposed to the low quality intervention. Also, subjects exposed to the low quality intervention listed a significantly larger number of negative thoughts than did subjects exposed to the high quality intervention. Analysis of subjects' attitudes toward the cognitive restructuring technique indicated only a main effect for intervention quality. As predicted the high quality intervention was judged to be more effective. In an attempt to determine the impact of the manipulations on behavior, a subset of subjects was instructed to list food-relevant thoughts over a one-week period and to engage in rational restructuring of the thoughts when appropriate. Two dependent measures were used: the percentage of appropriate applications and the range of days on which the technique was applied. Analysis of the percentage of appropriate applications revealed a significant main effect for intervention quality and an interaction between level of involvement and source credibility. As expected, the high quality intervention was associated with a higher percentage of relevant applications and source credibility affected the percentage of appropriate applications only when issue involvement was low. This same pattern was found in analysis of the range of days across which the technique was applied. Actual-counseling studies. Heesacker, Heppner, and Shaw (1988) studied the relationship between client motivation for counseling, client expectations about counselor acceptance, expertness, The ELM & Treatment Outcomes 32 attractiveness, and trustworthiness, and premature termination from counseling. No main effects for motivation or expectations on termination were found. However, a significant interaction effect occurred, revealing lower expectations to be related to premature termination in 2312: psychotherapy while higher expectations were related to premature termination in longer psychotherapy. A similar, though statistically insignificant, trend regarding motivation was also found (Heesacker et al., 1988). A relevant study was conducted by Grimes and Murdock (1989) who, although not relying on ELM concepts, investigated the relationship of social influence variables to change in distress level (measured with the Brief Symptom Inventory; Derogatis & Spencer, 1982) and premature termination from counseling. Specifically, these authors utilized Strong's (1968) model and assessed the relationship of client perceptions of expertness, attractiveness, and trustworthiness (with the CRF-Short Form) to client symptom reduction and premature termination. Twenty-nine Of 51 subjects completed at least four sessions of counseling; the 22 subjects who attended four or fewer sessions subjects were classified as premature terminators. Multiple regression analysis revealed that all three social influence variables were significantly related to symptom reduction, as measured by the Global Severity Index of the BSI. Termination status was also negatively correlated with judgments of counselor expertness, trustworthiness, and a composite counselor rating score. Pierce and Stoltenberg (1990) examined the effects of an ELM-based intervention on attrition rates in behavior education (BE) classes conducted within a weight loss program. A motivation enhancement The ELM & Treatment Outcomes 33 program (MEP) was compared to 3 control conditions: a standard BE class aimed at changing negative eating habits, a time-attention placebo group, and a group of clients who had declined to participate in the MEP. The MEP was designed to instill positive attitudes and behavioral intentions toward regular attendance of the BE class. The dependent measure was the number of BE classes attended over a 12 week period. Results revealed that MEP clients attended classes on significantly more consecutive weeks than did clients in the control conditions. On balance the accumulated social and counseling psychology data warrant continued use of the ELM to investigate counseling processes and outcomes. The theory is a major step beyond more simplistic conceptualizations of attitude change, and may assist practitioners and scientists to clarify the means by which counseling changes attitudes and behaviors. Social psychological variables (source, message, recipient) are paralleled by counseling variables of therapist, intervention, and client. As in social psychology laboratory studies, message quality appears to interact with personal relevance (i.e., motivation) in determining recipient/client attitudes. Unfortunately, the variable of recipient ability to cognitively elaborate therapeutic statements has not been investigated. This is a major short coming in the counseling-psychological applications of the model because ability is equal in importance to motivation withinin the ELM. Attitude Change and Addictions Treatment The investigation of substance abuse or chemical dependency has not been approached from a social influence perspective. An extensive computer-based literature search revealed no investigations that examined social influence variables in the context of addictions The ELM & Treatment Outcomes 34 treatment. Only one article was located that linked the two topics. Lacrosse (1980) utilized a group of thirty-six subjects undergoing outpatient treatment of drug addiction in a validation study of the Counselor Rating Form (CRF: a measure of client perceptions of counselor expertness, attractiveness, and trust-worthiness). The CRF variables, particularly expertness, were found to have predictive utility relative to therapeutic outcomes. NO specific mention was made about the nature of the sample in the results or discussion section of this article. Although little overlap exists between addiction and attitude-change literature, discussion of addictions treatment is relevant because of the population under study in this investigation. Moreover, the extension of the ELM to the explication Of chemical dependency treatment is relevant because the treatment of addictions entails a great deal of persuasion and also relies heavily on cognitive interventions. A brief discussion of different theories of addictions treatment follows, and the chapter will conclude with a discussion of client motivation within the context of chemical dependency treatment. Models of addiction. Miller and Hester (1989) identified eleven conceptual models of addictions treatment: Moral, Temperance, American Disease, Education, Characterological, Classical/Operant Conditioning, Biological, Social Learning, General Systems, Sociocultural, and Public Health. They then proposed a twelfth approach in which they encouraged the use of an "informed eclecticism." Miller and Hester's (1989) Informed Eclecticism emphasizes the importance of health factors and three additional concepts: (a) no single approach is superior to all others in treating all individuals, (b) different people respond better to different forms of treatment, and (c) treatment effectiveness and The ELM & Treatment Outcomes 35 efficiency can be maximized by matching individuals with their optimal treatment. The American Disease Model is more restrictive: it attributes the cause of addiction to physical, psychological, and spiritual conditions within the addict that make him or her incapable of drinking or using drugs in moderation. This model has been embraced by Alcoholic Anonymous and is the most widely utilized model of inpatient and outpatient chemical dependency treatment. It is also the model used at each of the data collection sites in this study. Motivation to overcome addiction. Though a wide range of differences exist among and within various professions involved in treating addiction, all agree that motivation of the client is crucial (Miller, 1989). Miller (1989) encouraged a shift in our definition of motivation away from something the client has (i.e., motivation as a trait) to something the client does. "It [motivation] involves recognizing a problem, searching for a way to change, and complying with that change strategy" (Miller, 1989, p. 69). Miller (1989) has identified several strategies which can be used to enhance the motivation of individuals who need chemical dependency treatment. One method is to advise clients of the need to change and to provide specific and sound recommendations. A second method entails removal of practical barriers such as excess cost, inadequate child care, and inaccessibility of treatment facilities. A third approach involves providing clients with alternatives and choices. Individuals are more likely to comply with an intervention they have helped create. Choice has been found to reduce dropout rates and resistance (Miller, 1989). The ELM & Treatment Outcomes 36 A fourth commonly used method is decreasing the attractiveness of a behavior. This entails helping a client to become more aware of the negative consequences of their continued substance abuse. Methods used in the ELM, e.g., presenting the possibility of an unwanted consequence like a requirement to take comprehensive exams, are similar. A fifth method used to enhance motivation to participate in treatment is external contingencies. This refers to the use of pressure, for example, when an individual will lose his/her employment if he/she does not participate in treatment. A sixth means to increase motivation is the provision of individualized feedback. Most people respond well to specific feedback regarding the way their drug use harms themselves and others. Assisting clients to set clear, measurable, and realistic goals is a seventh method which increases motivation. An eighth medium is labeled the "helping attitude." Miller (1989) describes it as combination of the core therapeutic conditions of empathy, genuineness, and unconditional positive regard, and of therapist optimism. Miller (1983, 1989) weaves these methods into a confrontational process named "motivational interviewing." The interview is intended to increase a client's insight into and responsibility for his or her addiction and to instill a commitment to change. "The underlying strategy ... is to create a dissonance or discrepancy between the person's current behavior and important personal goals" (Miller, 1989, p. 74). In ELM terminology the interview can thus be used to help clients engage in cognitive elaboration Of addiction-relevant information. Research Questions, Hypotheses, and Statistical Analyses The primary question investigated in this study was whether recipient (e.g., client), message (e.g., substance abuse treatment), and The ELM & Treatment Outcomes 37 source (e.g., therapists) variables which have been hypothesized by the ELM to be relevant to interpersonal influence processes in counseling can be used to predict inpatient chemical dependency treatment outcomes. The four treatment outcomes studied included: post-treatment attitudes toward drug using and recovery-oriented behaviors, recovery-oriented behavioral intentions, symptom severity, and termination status (i.e., successful vs. premature termination). It was hypothesized that these outcomes could be predicted from a combination of ELM variables. Specifically, on the basis of ELM postulate four, it was hypothesized that (a) measures of client motivation and ability to examine therapeutic messages should be predictive of each of the above treatment outcomes. Based on investigations of the need-for-cognition construct (e.g., Cacioppo, et al., 1983), it was also hypothesized that (b) a measure of need-for-cognition should predict treatment outcomes. On the basis of Petty and Cacioppo's (1981, 1986) assertion that the nature Of recipient elaboration (i.e., favorable or unfavorable evaluations of persuasive arguments) determines the amount and direction of attitude change, it was hypothesized that (c) measures of client evaluations of therapeutic messages should predict treatment outcomes. Finally, based on the identification of counselor characteristics (McNeill & Stoltenberg, 1989; Petty & Cacioppo, 1986) as peripheral cues, it was hypothesized that (d) a measure of client perceptions of counselor expertness, attractiveness, and trustworthiness should also predict treatment outcomes. Bivariate correlations and multiple regressions were used to explore the simple and multivariate relations of the theoretical predictors to each of the outcome variables. In addition to these hypotheses, which assume that social The ELM & Treatment Outcomes 38 influence variables may individually and collectively predict various treatment outcomes, it is possible that interaction effects could exist. For example, because the ELM posits that both motivation and ability are necessary for cognitive elaboration, it is likely that the effects of clients' motivation to examine therapy-relevant information will depend on the level of their ability to comprehend therapeutic messages. It is also possible that motivation and ability will interact with perceptions of counselor characteristics. That is, in keeping with ELM theory, counselor characteristics (peripheral cues) may be unassociated with outcomes when motivation and ability are high, but positively associated with outcomes when motivation and ability are low. These and other interaction effects were explored within the statistical analyses. Chapter Summagy This chapter began with a description of the ELM and the seven postulates on which it is based. This was followed by a review of research on the theory, with particular emphasis on those studies examining variables to be included in the present study. Theoretical extensions and empirical applications of the ELM to counseling were also presented. Various models of addiction treatment were then identified and the importance of client motivation for addictions treatment was discussed. Finally, the research questions of this investigation were enumerated. The ELM & Treatment Outcomes CHAPTER 3: METHODOLOGY Subjects Participants in this study were 100 adult inpatients on the chemical dependency units of two psychiatric hospitals. Seven referral sources were identified: self (2313), family member (es37), friend (2317), outpatient therapist (g?15), doctor (g=2), and court order (eél6). They ranged in age from 18 to 61: mean=32.40, S2?10.00. All participants were volunteers. There were 57 males and 43 females. The average number Of years of schooling was 11.83 (§2=2.04). Subjects indicated an average of 5.71 (§2=7.43) biological relatives who also had substance abuse problems. Most subjects were unemployed (ee64); thirty-four subjects indicated that they were currently employed and two subjects did not report their employment status. Specific data on the socioeconomic status of subjects was not gathered. However, it is likely that most subjects in this study were from low to lower-middle class backgrounds given the large number of unemployed subjects in the sample. Also, most subjects reported that they worked or had previously worked in unskilled or semiskilled occupations e.g., dishwasher, janitor, hair-stylist, assembly line worker. The marital status of subjects included: never married (3:42), married (2:31), separated (e=12), divorced (eél3), and unreported (2:2). The number of prior outpatient chemical dependency treatments which subjects had undergone ranged from O to 6; fifty-three reported having never had outpatient treatment, 31 reported having gone through outpatient treatment once, and 13 subjects indicated they had gone trough outpatient treatment 2 or 39 The ELM & Treatment Outcomes 40 more times. The number of previous inpatient chemical dependency treatments ranged from O to 8: fifty-five subjects had never previously received inpatient care, 24 subjects had gone trough inpatient treatment once before, and 18 subjects reported 2 or more inpatient treatments (3 subjects did not indicate whether they had had inpatient care). The two sites from which data were collected were both inpatient chemical dependency units housed within the mental health departments of private hospitals. One site was located in a moderately sized urban community in the Midwest (e=52); the other site was in a small urban community in the Northeast (2?48). The primary treatment orientation of both was based on the American Disease Model; each provided individual, group, family, and activity therapy, and also used education sessions and A. A. meetings. Both treatment sites were unwilling to share specific diagnostic information about the subjects, however, each site reported that they treated the following types of chemical dependencies (listed in descending order of frequency): alcohol, polysubstance - usually alcohol and cannabis, cannabis, cocaine, and other. Racial minorities comprised less than ten percent of the individuals receiving treatment at the Midwest site and zero percent of the Northeastern site. See Tables 3-1 and 3-2 for further clarification of the demographic data. Subjects completing treatment. Sixty-seven subjects completed treatment (males: e§37, females: e=30): they ranged in age from 18 to 56, mean=32.51, §2=9.86. Premature terminators. Thirty-three subjects terminated treatment prematurely against medical advice (males: e=22, females: 2:11): they ranged in age from 19 to 61, mean=32.27, §2=10.63. The ELM & Treatment Outcomes 41 Table 3-1 Means and Standard Deviations of Demographic Data Successful Premature Total Completers Terminators Sample (E=67) (N=33) (NelOO) Variable M §2 M §2 ‘M §2 Age 32.51 9.86 32.41 10.63 32.40 10.00 Years of School 11.94 2.09 11.58 1.95 11.83 2.04 1 Prior or Hosp's* .88 1.48 .65 1.23 .81 1.40 # Prior IP Hosp's* .94 1.65 .90 1.54 .93 1.60 # Bio-Relatives* 6.32 8.70 4.37 3.22 5.71 7.43 # NonBio-Relatives* .23 .52 .13 .35 .21 .48 *These variables represent the number of: previous outpatient (0?) treatments for addiction, previous inpatient (IP) hospitalizations for addiction, and biological and non-biological relatives identified by the subject as having a substance abuse problem. Measures of the Independent Variables Demographic Data Sheet (DDS). The DDS is a 9-item survey used to obtain information about subjects' age, gender, marital status, employment status, number of previous inpatient and outpatient treatments for substance abuse, and number of relatives with substance abuse problems (see Appendix A). No explicit relationship between these variables and attitude change is postulated by the ELM: however, they were included in order to examine, and control for, their possible relation to the treatment outcomes. The Counselor Rating Form-Short(CRF-S). Corrigan and Schmidt (1983) developed a 12-item abbreviated version of the CRF (Barak & LaCrosse, 1975) and demonstrated that the reliability of the scale was comparable to that of the longer version (see Appendix B). The CRF is a The ELM & Treatment Outcomes 42 Table 3-2 Frequencies of Nominal Level Demographic Data Successful Premature Total Completers Terminators Sample (N=67) (N=33) (§?100) Variable Frequency Frequency Frequency Gender Male 37 22 59 Female 39 ll 31 Total 67 33 100 Marital Status Never Married 31 11 42 Married 19 12 31 Separated 10 2 12 Divorced l .9 12 Total 67 31* 98* Employment Status Part-Time 3 2 5 Full-Time 22 7 29 Unemplgyed fig 2g 83 Total 7 31* 98* Level of Education Junior High 14 9 23 G.E.D. 11 5 16 High School 21 10 31 Some College 14 4 18 Associates Degree 2 2 4 Bachelors Degree _2 _l .2 Total 67 31* 98* Admission Status Voluntary 57 27 84 Court Ordered 19 e .18 Total 67 33 100 *This figure is not equivalent to the sample size e_because of missing data. The ELM & Treatment Outcomes 43 36-item scale which was found through principal components factor analysis to have three primary factors tapping client perceptions of counselor expertness, attractiveness, and trustworthiness (Barak & LaCrosse, 1975). Items which had consistently shown high factor loadings across most analyses were selected for the CRF-S. Inter—item reliabilities of the 4-item expertness, attractiveness, and trustworthiness scales in the CRF-S were .85, .91, and .91, respectively, using a sample of community mental health center clients (Corrigan & Schmidt, 1983). Validation of the CRF-S factor structure using confirmatory factor analysis indicated that a 3-factor oblique model could best account for the results, thereby replicating and extending Barak's and LaCrosse's work. The CRF-S served to Operationalize source characteristics in this study. In this sample the Cronbach's alpha coefficients were .90 for the whole survey, .79 for the expertness scale, .82 for the attractiveness scale, and .77 for the trustworthiness scale. Each scale was composed of four 5-point Likert scale items. Expertness, attractiveness, and trustworthiness scores were derived from the calculation of the average of the four items in the scale. Scores could potentially range from 1 to 5; higher scores indicate subject perceptions of a higher degree of counselor expertness, attractiveness, and trustworthiness. Treatment Evaluation Questionnaire (TEQ). This 24-item scale was designed by the principal investigator to reflect subjects' perceptions of different qualities of therapeutic messages to which they had been exposed during treatment (see Appendix C). Subjects used a 5-point Likert scale to evaluate six modalities of treatment: (a) individual therapy. (b) family therapy, (c) group therapy: (d) activity therapy, The ELM & Treatment Outcomes 44 (e) education sessions, and (f) A.A.lN.A. meetings. The subjects were asked to indicate their perceptions of how frequently they where exposed to arguments intended to change their opinions about drug use, how epplicable suggestions made to them in each modality were to their drug problem, how confident they were that the suggestions made to them in each modality would help them stay sober, and the degree to which they were able to understand the jargon and suggestions to which they were exposed in treatment. Items were combined over treatment modalities to form four scales. Each scale contained six items and scores were determined by calculating the average ratings. Scores could range from 1 to 5 with higher scores reflecting perceptions of a higher degree of the variable of interest. The first scale is a measure of the perceived frequency (TEQ-PF) of message exposure (Cronbach's alpha coefficient = .83); it was intended to approximate the variable of message repetition used in ELM laboratory studies. The second is a measure of the perceived applicability (TEQ—PA) of therapeutic suggestions to the subjects' addiction (Cronbach's alpha coefficient = .77). The third scale is a measure of the perceived confidence (TEQ-PC) that therapeutic suggestions will assist the subject in maintaining sobriety (Cronbach's alpha coefficient = .77). The fourth was used as an approximation Of subjects' ability to comprehend (TEQ-MC) and therefore engage in cognitive elaboration of therapeutic messages. A Cronbach's alpha coefficient of .84 was obtained for this scale. Each of the scales is a face valid measure and no other validity data is available for any of the four scales. The ELM & Treatment Outcomes 45 Marlowe-Crowne Social Desirability Scale, Short Form. This scale, called the Personal Reaction Inventory (PRI), was developed by Reynolds (1982) who selected 13 items from the original (33-item) Marlowe-Crowne Social Desirability Scale (see Appendix D). Subject's respond true or false to items such as, "I sometimes feel resentful when I don't get my way". The social desirability score equaled the number of items endorsed in a direction reflecting a socially desirable response. Validity was established by correlating the PRI to the standard Marlowe-Crowne scale, and to the Edwards Social Desirability Scale. These coefficients were .93 and .41, respectively. Reynolds (1982) found an internal consistency of .76 using the Kuder-Richardson formula-20 reliability coefficient. This instrument was used to estimate the extent to which subjects' responses to the various self-report measures in this study may have been affected by the tendency to respond in a socially desirable manner. The Cronbach's alpha coefficient of this scale for this study was .75. Need for Treatment Survey (NTS). This 21-item scale was constructed by the primary investigator to tap aspects of subjects' issue-involvement, i.e., motivation to listen to and engage in cognitive examination of the messages, suggestions, and recommendations presented in treatment (see Appendix E). It is intended to tap motivation for treatment in a general manner and to measure motivation to examine particular topics in treatment (e.g., spiritual well-being, emotional and physical health, and intimate relationships) that are relevant to recovery from addiction. It does not specifically measure subjects' motivation regarding such issues as participating in A.A. or abstaining from drugs. The ELM & Treatment Outcomes 46 The first two items asked subjects to indicate who, among several possible significant individuals, had suggested or demanded that they receive treatment, and to judge the extent to which their decision to seek treatment was based on this social pressure. Preliminary item analysis revealed that the vast majority of subjects responded identically to the second portion of both items (i.e., virtually all subjects indicated that they strongly disagreed that they had based their decision to seek treatment on the suggestions or demands of others). Thus, there was essentially no variance for either item and these items were dropped from further analyses. The remaining 19 items required subjects to rate their extent of agreement or disagreement with statements about their commitment to treatment, their expectations of positive consequences that would result from treatment, and their expectations of negative consequences should they not participate in treatment. All items used a 5-point Likert scale. Eighteen items were constructed so that a stronger degree of agreement reflected higher motivation. For one item, item 14, disagreement reflected higher motivation. These 19 items produced a Cronbach's alpha coefficient of .90. Items were designed to reflect face validity. Most items were worded in a manner which asked subjects judge their level of agreement with statements pertaining to their perception of the likelihood of avoiding negative consequences and attaining positive consequences as a result of listening to and seriously considering treatment recommendations. This wording specifically addresses the issue of subject motivation to cognitively examine treatment messages. NO other validity data is available. The ELM & Treatment Outcomes 47 The NTS is conceptualized as a measure of egege motivation. This method of determining subject motivation differs from the typical method used in laboratory investigation of ELM variables. In most ELM studies, motivation is a variable which is experimentally manipulated, i.e., subjects are exposed to conditions which either increase or decrease their issue-involvement. Because of the non-experimental nature of this study and the fact that it would be unethical to attempt to decrease patients' treatment motivation, it was necessary to assess levels of motivation among a sample in which all members were expected to be relatively high in motivation. It was expected that higher motivation within this restricted range would be more predictive of treatment success. Need for Cognition Scale (NCS). This scale was originally developed and validated as a 34-item instrument in a series of four studies (Cacioppo & Petty, 1982; see Appendix F). The scale was designed to measure the psychological need of an individual to engage in and enjoy thinking. It taps subjects' perceptions of their motivation to use their mental capacities to examine different aspects of problems and then to base solutions on these efforts. Cacioppo, Petty, and Kao (1984) reduced the NCS from a 34-item scale to an 18-item scale and concluded that the reliability was not sacrificed because the maximized Cronbach's alpha coefficient was .90 for the 18-item scale and .91 for the longer version. A principal components factor analysis yielded one dominant factor, thereby supporting the validity of the 18-item Need for Cognition Scale. The short version of the NCS was used in this study as a measure of gee}; motivation. A Cronbach's alpha coefficient calculated on this sample was .86. Responses to the eighteen 5-point The ELM & Treatment Outcomes 48 Likert scale items were averaged to yield the Need-for-cognition score. Higher scores reflect higher motivation. Thoughts and Feelings List (TFL). This instrument (see Appendix G), designed as a modified version of Cacioppo and Petty's (1981) thought-listing technique, was intended as an additional index of whether subjects had been eele to comprehend therapeutic messages and whether they had changed their attitudes toward drug use and recovery-oriented behaviors. When subjects in this study completed treatment they were asked to list all of the thoughts and feelings they had had about their chemical dependence and how they might best maintain their recovery. (It was planned that the TFL would later be shown to patients' primary therapists, who would rate the nature of the thoughts listed; see Appendix H - Therapist's Evaluation Form.) The method used here differs from Cacioppo and Petty's (1981) technique in that the present method does not request subjects to list thoughts immediately after being exposed to persuasive arguments in a single session, but rather is administered at the end of treatment. The technique used in this study was similar to Greenwald's (1968) procedure in which subjects are asked to list general thoughts on a topic. This adaptation was used because it would have been extremely cumbersome for subjects to list their thoughts after every therapy session or every time they were exposed to a persuasive communication during treatment. It lacks the experimental control of the lab, yet is more realistic in regard to tasks an actual patient would be asked to complete. A second difference is that subjects in this study were asked to list feelings in addition to thoughts. Subjects were asked to list feelings, because in therapy emphasis is often placed on feelings more than on thoughts, and The ELM & Treatment Outcomes 49 asking for feelings was a way to further reveal the subject's treatment-relevant cognitions (i.e., statements about feelings were defined as a subclass of subjects' thoughts). A third difference was that the thought lists in this investigation were intended to be ancillary measures used in conjunction with the Therapist's Evaluation Form. In the thought listing technique described by Cacioppo and Petty (1981), the number of favorable and unfavorable thoughts generated are used as dependent variables (e.g., Petty & Cacioppo, 1979, 1984). In order to minimize the possibility that subjects would alter their responses to impress their therapists, subjects were not informed that their therapist would evaluate their TFL until after they had completed the form. The use of deception was avoided by informing subjects that the experimenter wished to have their therapist review only their TFL (no other surveys would be seen by their therapist or other hospital staff members), and then requesting them to sign a form (see Appendix M) granting permission for their therapist to review the TFL. Therapist's Evaluation Form (TEF). This 8-item scale was developed by the principal investigator to provide concurrent evidence of the subjects' motivation and ability to engage in a thoughtful consideration of treatment-relevant issues (see Appendix H). The primary therapist for each subject was asked to evaluate the subject on dimensions related to his or her motivation, ability, and attitude change. The first item allows the therapist to identify the type and number of treatment modalities (individual, group, family, activity, educational, A. A. meetings) in which the subject participated. Items two and three ask the therapist to indicate his or her level of The ELM & Treatment Outcomes SO agreement with statements regarding the subject's motivation to engage in cognitive elaboration of treatment issues. Items four and five ask therapists to rate their level of agreement with statements regarding the subject's ability to engage in cognitive elaboration of therapeutic messages. Items six, seven, and eight were designed to be used in conjunction with the subject's completed "Thoughts and Feelings List". Specifically, the therapist was asked to examine the TFL of each subject who granted permission for his or her TFL to be reviewed, judging whether the thoughts listed demonstrated: (a) that the subject comprehended therapeutic messages, (b) change in attitudes toward drug use, and (c) change in attitudes toward behaviors needed to recover from addiction. Unfortunately, useful data were not obtained with either the TEF or the TLF. Only 28 TEFs were completed, all 28 were completed by therapists at the Midwest site, and these measures were collected only on the first 28 subjects at that site. This occurred because the therapists who had initially agreed to provide assistance stOpped completing the forms. The reason given for their discontinuation was that they felt overworked and did not have time to complete superfluous paperwork. Therapists at the Northeastern site had initially agreed to complete the forms, but no completed TEF's from this site were returned to the experimenter. Additionally, only 12 subjects (9 at the Midwestern site and 3 at the Northeastern site) gave permission for their therapist to review their TLF. Measures of Dependent Variables Juggements of Drug Use and Recovery—Oriented Behavior. This 10-item scale was developed by the principal investigator to assess subjects' attitudes regarding various drug using and recovery promoting The ELM & Treatment Outcomes 51 behaviors (see Appendix I). It is based on a method of measuring attitudes toward behaviors from the theory of reasoned action (Ajzen & Fisbein, 1980). Specifically, attitudes towards a given activity were assessed by asking respondents to rate their position (with a 9-point Likert scale) on four continua regarding the extent to which the activity is: harmful or beneficial, bad or good, unpleasant or pleasant, and punishing or rewarding. This measure was administered to each subject at the outset of treatment and again when they completed treatment. Cronbach's alpha coefficients of .97 occurred at both assessment intervals. A general recovery-oriented attitude score was created by determining the average value of each subject's ratings across all lO-items. Scores could have ranged from 1 to 9 with higher scores reflecting a greater degree of pro-recovery oriented attitudes. Items were designed as face valid measures of attitudes toward various recovery oriented behaviors, for example, remaining abstinent. The survey was not statistically validated. Statement of Behavioral Expectations (SBE): This 7-item face valid measure was constructed by the primary investigator to assess subjects' intentions to follow treatment recommendations to remain abstinent, join a 12-Step recovery peer support group (e.g., A.A.,N.A.,C.A.), and seek outpatient aftercare therapy (see Appendix J). Subjects were asked to predict the likelihood that they would actually follow through on their stated intentions using a 5-point Likert scale. Scores were created by calculating the average of each subject's responses to the seven items. A higher score reflected a higher degree of pro-recovery oriented intentions. A Cronbach's alpha coefficient of .84 was found for this measure in this sample. All items were face 'valid and no other form of validity data is available. The ELM & Treatment Outcomes 52 Brief Symptom Inventory (BSI). This is a 53-item inventory developed by Derogatis and Spencer (1982) which yields scores on 9 symptom dimensions (somatization, obsessive-compulsiveness, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychoticism) and three global indices of distress (global severity index, positive symptom distress index, and positive symptom total). A five-point Likert scale is used to assess level of distress for each symptom. Internal consistency coefficients were computed using Cronbach's coefficient alpha on a sample of 719 psychiatric outpatients. The alpha coefficients for the 9 symptom dimensions ranged from a high of .85 on Depression to a low Of .71 on Psychoticism. Test-retest coefficients across a two-week interval ranged from .91 on Phobic Anxiety to .68 on Somatization. Test-retest coefficients for the global severity index, positive symptom distress index, and positive symptom total were respectively .90, .87, and .80. The authors of the BSI found substantial convergent validity between the BSI and the MMPI. To assess internal structure and construct validity, a principal components factor analysis was conducted and a normal varimax rotation of the principal components indicated that there were 9 interpretable factors. The Global Severity Index (GSI), which reflects a combination of the positive symptom distress index and the positive symptom total, was used as the measure of symptomatology in this study. The GSI is calculated by dividing the sum Of all responses by 53 (the total number of? items on the inventory) and is then converted into a standardized TD—score using published norms (Derogatis & Spencer, 1982). The ELM & Treatment Outcomes 53 Completion of Treatment Form. This form (Appendix K) was used to identify whether patients prematurely terminated or successfully completed treatment. It was completed by either the principal investigator or a research assistant. Procedures On the first or second day following admission to the inpatient chemical dependency units, patients were approached and informed that F- the unit was providing assistance to a student who was investigating changes that people experience during inpatient treatment for addiction. Potential subjects were informed that their participation would involve Ia- responding to six brief surveys at the outset of treatment and six brief surveys on their second to last or last day of treatment. They were also informed that the total time commitment required for the study was no more than an hour at the beginning and approximately one-half hour at the end of treatment. All potential subjects were informed that their participation would be completely voluntary and that their consent to participate in the study could be withdrawn at any time. They were also told that their involvement or lack of involvement in the study would not affect their current or future services in the unit and that no staff member other than those persons directly involved in collecting the data would have access to their surveys. They were also told that their confidentiality would be protected because their names would not be attached to their surveys and any reports based on the study would identify only group trends, not the data of any specific individual. Two part-time staff members acted as research assistants at the liidwestern site (one was a psychology technician and the other was an .activity therapist): both were made familiar with the above procedure The ELM & Treatment Outcomes 54 for recruiting subjects and with the instruments. Both of these assistants were paid five dollars for each subject to whom they administered surveys. Subjects were not remunerated for their participation. At the Northeastern site, data were gathered either by the primary investigator or by a nurse who had been trained in the recruitment procedure and familiarized with the measures. Those who agreed to participate were required to read and sign a consent form (Appendix L) which described the study and informed them of their rights and duties as subjects. After signing the consent form subjects were given the Demographic Data Sheet, Personal Reaction Inventory, Brief Symptom Inventory, Judgements of Drug Using and Recovery-Oriented Behaviors, Need for Treatment Survey, and the Need for Cognition Scale (see Table 3-3). The instruments were self-administered: however, the primary investigator or a research assistant remained in the subject's presence while he or she completed the surveys, in order to clarify instructions, if necessary. Most subjects were administered the instruments in small groups: some subjects were individually administered the surveys. Group versus individual administration depended upon the number of subjects who had been admitted or discharged from the program on the days that on or very near the days that data was collected. Most subject completed the pre- treatment surveys in 30 to 45 minutes. All subjects then underwent the standard treatment provided by the Chemical Dependency Unit to which they were admitted. The typical length of stay at each unit was from 12 to 14 days. On their last or second to last day of treatment subjects completed the Brief Symptom Inventory, Judgements of Drug Using and Recovery-Oriented Behaviors, Statement of Behavioral Expectations, Counselor Rating Form-Short, The ELM & Treatment Outcomes 55 Table 3—3 Pre and Post Treatment Surveys 1. 2. 3. 4. Pretreatment Surveys Demographic Data Sheet Personal Reaction Inventory Brief Symptom Inventory Judgements of Drug Using and Recovery-Oriented Behaviors Need for Treatment Survey Need for Cognition Scale Post-Treatment Surveys Brief Symptom Inventory Judgements of Drug Using and Recovery-Oriented Behaviors Statement of Behavioral Expectations Counselor Rating Form-Short Version Treatment Evaluation Questionnaire TEQ-Message Comprehension TEQ—Message Frequency TEQ-Applicability TEQ-Confidence Thoughts and Feelings List & Therapist's Evaluation Form Construct Measured Demographic Information Social Desirability Distress (Symptom Severity) Attitudes State Motivation Trait Motivation Construct Measured Distress (Symptom Severity) Attitudes Behavioral Intentions Expertness, Attractiveness, & Trustworthiness Ability to Understand Messages Perceived Message Frequency Perceived Applicability of Tx Perceived Confidence in Tx Therapist Ratings of Subject Ability to Understand Messages Treatment Evaluation Questionnaire, and the Thoughts and Feelings List. Most subjects took from 20 to 30 minutes to complete the post treatment surveys. An attempt was made to administer post-treatment instruments to subjects who left treatment against medical advice before they had actually signed out and left the treatment unit. the premature terminators completed the post-treatment surveys. However, only one Of Data 'were collected continuously for a period of twelve months until a sufficient sample size (N=100) was obtained. At the Midwestern site, approximately twenty patients declined to participate: all indicated that they were not interested because study participation was not required as part of their treatment. ‘Northeastern, site twelve patients declined to participate. At the One of these individuals said he was unable to read; the remaining eleven The ELM & Treatment Outcomes 56 stated they were not interested in doing anything that was not a required part of treatment. Four subjects at the Midwestern site withdrew from the study without providing a reason for their withdrawal; two dropped out stating the surveys were too difficult to complete. At the Northeastern site two subjects who had initially agreed to participate withdrew after finding the surveys too hard to understand. They both indicated they were illiterate and had been embarrassed to admit this at first. One subject dropped out stating that the surveys involved too much work. Data Analysis The internal consistency of each instrument was first explored by calculating coefficient alpha reliabilities. Data from the two collection sites were compared using Chi-square analyses and multivariate analysis of variance. A correlation matrix was also computed, assessing relations among the individual theoretical predictors and treatment outcomes. After these preliminary analyses, the theoretical variables derived from the ELM were entered into multiple regression equations predicting post-treatment recovery-oriented attitudes, behavioral intentions, and symptom severity outcomes. Theoretical variables were grouped into sets according to the theoretical construct which they :measured. Specifically, state and trait motivation comprised one set (motivation), message comprehension a second (ability), treatment rapplicability and confidence in treatment a third (message evaluation), 'perceived message frequency a fourth (frequency), and counselor expertness, attractiveness, and trustworthiness a fifth set (counselor iratings). The rationale for grouping independent variables was to The ELM & Treatment Outcomes 57 conserve statistical power by maintaining an acceptable ratio of subjects to independent variables. Interaction effects were also examined using multiple regression analyses. In multiple regression two variables are said to interact in accounting for variance in a third variable when they have a joint effect over and above the combination of their separate main effects (Cohen & Cohen, 1983). Therefore, it is necessary to enter main effects prior to interaction effects in order to have the additive combination of main effects partitioned out so that the joint effect can be identified. Because of limitations in the SAS program it was necessary to conduct the regressions in a two-stage process. In the first stage the groups were entered into a stepwise analysis to determine main effects. (Stepwise selection involving forward entry followed by backward entry was used in all regressions). In the second stage the main effects identified in the first equation were hierarchically entered in a second equation followed by interaction effects entered in a step-wise manner. This was done to ensure that main effects would be entered prior to interaction effects. The SAS program is incapable of entering variable sets in a stepwise manner while simultaneously entering interactions after the sets. The interactions were entered into the regression equations in a step-wise fashion. The proposal for this research project had stipulated that rmultiple regression analyses would be used to predict changes in attitude and distress (i.e., symptom severity) because the ELM is a model of attitude change. However, because of difficulties in the interpretation of pre- post change scores, an alternate data analysis strategy was used: pretreatment attitude and distress were controlled by The ELM & Treatment Outcomes 58 entering each as the first variable in the regressions predicting, respectively, post-treatment attitude and distress. Finally, data pertaining to the prediction of termination status were analyzed using canonical discriminant function analysis. This analysis was conducted to explore potential differences between successful completers and premature terminators. Only pretreatment data were used in the discriminant function analysis because premature terminators did not complete post-treatment measures. Supplemental analyses not directly related to the purposes of this study were also carried out. One form of analysis consisted of comparing the significance of the differences between the correlations of central route variables (e.g., confidence in treatment) and peripheral route variables (e.g., perceptions of counselor expertise). A e-score formula described by Steiger (1980) for comparing the significance of difference between correlations from the sample was used. The method involves comparing the difference between the respective correlations of two independent variables with a third dependent variable. The correlations of the central and peripheral route variables and the differences between various pairs of correlations are discussed in the context of outcomes that could be expected on the basis of the ELM. A second supplemental analysis consisted of determining the means and standard deviations of treatment evaluations, i.e., judgements about message frequency, treatment applicability, confidence in treatment, and message comprehension, according to the six treatment modalities. This .analysis was conducted to explore variation among the six treatment modes. The ELM & Treatment Outcomes CHAPTER 4: RESULTS Comparison of Data Collection Sites Chi-square analyses revealed no significant differences between subjects from the two data collection sites in regard to marital status, level of education, employment status, whether or not they were court-ordered to undergo treatment, or termination status (successful vs. premature). This was true for Chi-squares conducted using all 100 subjects and also for Chi-squares conducted using only the 67 successful completers. A significant difference between the two samples was found in regard to gender balance, x2(1)=5.34, p<.05, gs1oo. The ratio of males to females was nearly equal (males=25, females=27) at the first collection site, but differed substantially at the second data collection site (males=34, females=l4). A gender balance difference also existed among those subjects who completed treatment, x2(1)=5.78, p<.05, es67. The ratio of males to females changed at the first data collection site (males=15, females=21) because of treatment dropouts, and continued to be unbalanced at the second site (males=22, females=9). See Table 4-1. After Chi-square analyses were conducted on nominal level demographic data, two one-way MANOVAs were conducted to check for differences between the data collection sites in terms of continuous level demographic and theoretically derived variables. The first MANOVA included 97 subjects (data from 3 subjects were eliminated because of Inissing data on some variables) and tested for pretreatment differences on six demographic variables (age, years of education, number of prior outpatient and inpatient treatments, number of chemically dependent biological relatives, and number of chemically dependent non-biological 59 The ELM & Treatment Outcomes 60 Table 4-1 Demoggephic Frequencies: Total Sample & Completers by Location All Subjects Completers Site Site Variable #1 #2 Total #1 #2 Total Gender Male 25 34 59 15 22 37 ___Fema1e a y; .21 2.1. .2 22 Total 52 48 100 36 31 67 Marital Status Never Married 19 23 42 l4 17 31 Married 20 11 31 11 7 18 Separated 5 7 12 5 5 10 Divorced _e _Z 12 _§ _2 _1 Total 50 48 8 35 31 66 Employment Status Part-Time 2 3 5 1 2 3 Full-Time 17 12 29 14 8 22 Unemployed 21 32' 85 29 21 £1 Total 50 48 98 35 31 66 Level of Education Junior High 12 ll 23 7 7 l4 G.E.D. 7 9 16 6 5 11 High School 11 20 31 7 13 20 Some College 12 6 18 9 5 14 Associates Degree 3 1 4 2 O 2 Bachelors Deggee Q; _l .9 _e _l _5 Total 50 48 98 5 31 66 Admission Status Voluntary 41 43 84 28 29 57 Court Ordered 11 _§ 16 _§ 2 10 Total * Chi-squares significant for all subjects and completers, p<.05. The ELM 8 Treatment Outcomes 61 Table 4-2 Means and Standard Deviations by Site. All Subjects Site I 1 Site 9 2 Combined 11 ER. .5 .§2 .! .EE .8222: Age 34.16 9.80 30.56 9.98 32.40 10.00 18.00-61.00 Years of School 12.22 2.10 11.42 1.91 11.83 2.04 6.00-17.00 0 Prior 0P Hosp's .72 1.47 .90 1.32 .81 1.40 0.00- 6.00 4 Prior 1P Hosp's .98 1.67 .88 1.55 .93 1.60 0.00- 8.00 r Bio-Relatives 4.98 4.96 6.46 9.30 5.71 7.43 0.00-25.00 l NonBio-Relatives .27 .53 .15 .41 .21 .48 0.00- 2.00 Social Desirabilitya 3.73 2.89 3.81 3.30 3.77 3.08 0.00-12.00 State Motivationb 3.97 .80 4.08 .70 4.02 .75 1.21- 5.00 Trait Motivationb 3.08 .60 3.08 .61 3.08 .60 1.17- 4.42 Attitudes Pretreatment‘ 7.14 1.49 7.27 1.51 7.20 1.49 3.20- 9.00 Distress Pretreatmentd 56.25 8.61 55.52 7.10 55.90 7.89 36.00-80.00 aThis variable was scored by summing the number of items endorsed in a direction reflecting a socially desirable presentation. Scores on these variables could range from 0 to 5, with higher scores reflecting stronger perceptions or feelings. cScores on this variable could range from O to 9: higher scores reflected increasingly positive attitudes towards recovery. This score reflects the average Global Severity Index T-Score; T-scores were obtained from the BSI normative tables for psychiatric inpatients. relatives) and all five theoretically derived pretreatment variables (pretreatment symptom severity, social desirability score, pretreatment attitudes, state-motivation for treatment, and trait-motivation to use cognition for problem solving). This MANOVA was nonsignificant §(ll,85)=l.33, 2:.2247. See Table 4-2. The second MANOVA, which used data from 65 subjects who completed treatment (data from 2 subjects were eliminated because of missing data on some variables), was conducted to determine if significant differences existed between treatment completers at the two data collection sites, in terms of six demographic variables and sixteen theoretically derived variables (See Table 4-3). This analysis was also nonsignificant 2(22, 42)=1.33, p=.l42. Because a significant difference between data collection sites was detected only on one variable (gender The ELM & Treatment Outcomes 62 Table 4-3 Means and Standard Deviations by Collection Site and for 80th Sites Combined for Subjects Hho Completed Treatment Site I 1 Site I 2 Combined (gf36) (2331) (2367) .! .EQ .! .EQ .! 3!! .3222: Age 33.92 9.18 30.87 10.50 32.51 9.86 18. 00 56. 00 Years of School 12.26 2.23 11.58 1.93 11.94 2.10 6.00- 17. 00 No. Prior DP Hosp's .91 1.69 .84 1.24 .88 1.48 0.00- 6. 00 No. Prior IP Hosp's 1.09 1.77 .77 1.52 .94 1.65 0. 00- 8. 00 No. Bio-Relatives 5.06 5.54 7.74 11.20 6.32 8.70 0. 00 25. 00 N0. NonBio-Relatives .26 .56 .19 .48 .23 .52 0.00- 2.00 No. Treatment Modesa 5.44 .56 5.55 .58 5.49 .56 4.00- 6.00 Social Desirabiiityb 4.17 2.55 3.58 3.39 3.89 2.96 0.00-11.00 State Motivationc 3.97 .81 4.19 .54 4.07 .70 1.21- 5.00 Trait Motivationc 3.13 .59 3.10 .70 3.12 .64 1.17- 4.56 Message Frequencyc 3.54 1.01 3.56 .96 3.55 .98 0.00- 5.00 Treatment Appiicabiiiyc 3.91 .88 3.82 .69 3.87 .79 0.00- 5.00 Confidence in Treatmentc 3.97 .87 3.98 .73 3.97 .80 0.00- 5.00 Message Comprehensionc 4.07 1.15 4.00 .82 4.04 1.00 0.00- 5.00 Counselor Expertnessc 4.69 .44 4.38 .95 4.55 .73 0.00- 5.00 Counselor Attractivenessc 4.64 .47 4.20 1.00 4.44 .79 0.00- 5.00 Counselor Trustworthinessc 4.70 .50 4.47 .92 4.59 .73 0.00- 5.00 CRF-Score (EAT Combined)c 4.68 .41 4.35 .92 4.53 .71 0.00- 5.00 Behavioral Intentionsc 4.20 .62 4.23 1.07 4.21 .85 0.00- 5.00 Attitudes Pretreatmentd 7.20 1.44 7.20 1.52 7.20 1.47 3.40- 9.00 Attitudes Post-Treatment 7.81 1.16 8.11 .99 7.95 1.09 4.40- 9.00 Distress Pretreatmente 56.56 8.49 55.94 7.21 56.27 7.87 40.00-80.00 Distress Post-Treatmente 45.75 9.51 47.52 6.85 46.57 8.37 30.00-80.00 aThis variable reflects the number of various treatment modalities that subjects experienced. Each collection site had the same six modes. i.e. individual. group, family. education. activity. and attendance of A.A. meetings. This variable was scored by summing the number of items endorsed in a direction reflecting a socially desirable presentation. cScores on these variables could range from 0 to 5, with higher scores reflecting stronger perceptions or feelings. Scores on these variables could range from 0 to 9, with higher scores reflecting increasingly positive attitudes towards recovery. eThese scores reflect the average Global Severity Index T-Scores; T-scores were obtained from the BSI normative tables for psychiatric inpatients. The ELM & Treatment Outcomes 63 balance), the data were collapsed over sites in subsequent statistical analyses. Social Desirability An abbreviated version of the Marlow—Crowne Social Desirability Scale was included in this study because of the possibility that responses to certain measures might be biased by a social desirability set. The correlation matrix presented in Table 4-4 reveals a moderate negative correlation between social desirability and pretreatment distress. Subjects with higher social desirability scores tended to report less pretreatment distress. Social desirability was not significantly correlated with post-treatment distress or any other variable included in the study. Social desirability was entered into the discriminant function analysis to identify the difference between the mean social desirability scores of successful completers and premature terminators and to determine if this difference contributed to the discrimination between these two groups. The social desirability score was entered as the first variable in the multiple regression on post treatment distress to control for its effect on pre-treatment distress. It was excluded from all other analyses because its lack of correlation with each of the variables derived from the ELM suggested that subjects' responses were not biased by a social desirability set. Correlations of Predictor Variables with Treatment Outcomes The correlations presented in Table 4-4 display the correlation of each ratio level demographic variable, independent variable, and dependent variable with every other variable. The following description of the results focuses on the correlations between the independent (demographic and theoretically derived) variables and the dependent The ELM & Treatment Outcomes 64 . Soomu .809. .303 -- u~.- mo.- c~.- o~.- "H.-ee~.- om. c~.- -.- so. u..- -.-umo. as.. we. use. can. u-.- so. uuuuuuuo xu-uuoa .o~ -- awn. 8mm. own. sum. as. 8mm. umw. amm. a”. mu. NH. uo. o~.- ow. oo.- me. No.- oH.- ueouueuueu nu-uuoa .6" -- new. emN. gun. a”. ums. uoa. umN. Ho. o~. use. no.. mo.- Ho.- me. n“. mo.- Hm. uuusu.uu< x»-uuoa .mfi -- uom. uoo. ow. H_. u". new. we. mc.- mH. emm.- om. mc.- ~o.- "a. ma.. oH. umoe.euuosuusuu .NH -- use. as. LN. 8mm. nun. Lo. mo.- so. uwm.- a". No.- ec.- NH. “H..uu~. uuueu>.uueuuu< .uu -- on. as. Hm. emu. no. mo. uoN. eu~.- ”N. No.- oo. o“. no.- u“. uuueutuoxu .mH -- uum. umm. uum. H". uoN. on. ao.- a". as. ac.- «a.. mo. mu. eo.ueu;ueoeou uuuuuo: .ufl -- use. uum. 6“. 8mm. umw. uo.- 6“. ma. __. ac. ac. om. eh e. uueuuuueou .mH -- ufia. mu. own. new. uo.- mu. we. no. He. NH. mm. x» co suupueuu.Poa< .Nfi -- on. so. «H. u_.- N". “a. uH.- -.- «H. Hm. sueusouuu umuuuu: .LH -- u". Ho. mo.- ~_. no. o~.- cu..umm. a". eo.uu>.uo= u.ucu .cfi -- on. 4“. No. mo. on. H”. NH. mo. eouuusuuo: uuuum .o -- m_.- H". me. u“. mo. ao.- no.- uuusu.uu< xuuua .m -- eam.- ac. use. umN. uo.- o~.- uuucuu.o xhuea .a -- so. no.. NH. "H.- a“. mu..*euu.uua Pu.uom .u -- ~o.- no. as.. so. uu>.uepu¢ uuuu.uu< .m -- umu. uo..- mu. ueu.ueoeu uo.ua .u -- uum.- ac. ueu.uuauac to.ea .m -- ~o.- .ooeum co uuuu> .N -- one .u ON a” «H as a” m" a“ m" NH "H on o m a m m e m N H u.eu.eo> mu—pm.ga> «cavemamo can .Louorumco .uvzamcmosoo mo meowunpocaou etc opaa» The ELM & Treatment Outcomes 65 variables, i.e, post-treatment attitudes, behavioral intentions, and distress. Post-treatment attitudes were more highly correlated with pretreatment attitudes (5?.44,‘p(.001) than any other variable. Post-treatment attitudes were also significantly correlated with subjects' confidence in treatment (5;.43, p<.OOl), perceptions of treatment as applicable to their problem (5?.40, 2K.OOl), ratings of message frequency (£3.28, pK.OS), and perceptions of counselor expertness (3?.34, p(.01), attractiveness (5?.25,‘p<305), and trustworthiness (5?.24, p<.05). Post-treatment recovery oriented behavioral intentions were significantly correlated with message frequency (5;.33,‘p<.01), applicability of treatment (ET'ZS’.E