1 Meta-Analysis and Interpersonal Communication: 3. Function and Applicability. Mike Allen and Raymond W.Preiss. 2 Meta-Analysis in Context: A Proto-Theory.
Interpersonal Communication Research: Advances Through Meta-Analysis
LEA’s COMMUNICATIONS SERIES Jennings Bryant and Dolf Zillmann, General Editors Selected titles in Interpersonal Communication (Rebecca G.Rubin, Advisory Editor) include: Allen/Preiss/Gayle/Burrell • Interpersonal Communication Research: Advances Through Meta-Analysis Cupach/Spitzberg • The Dark Side of Interpersonal Communication Daly/Wiemann • Strategic Interpersonal Interaction Hewes • Cognitive Bases for Interpersonal Communication Kalbfleisch/Cody • Gender, Power, and Communication in Human Relationships Petronio • Balancing the Secrets of Private Disclosures For a complete list of titles in LEA’s Communication Series, please contact Lawrence Erlbaum Associates, Publishers.
Interpersonal Communication Research: Advances Through Meta-Analysis
Edited by Mike Allen University of Wisconsin-Milwaukee Raymond W.Preiss University of Puget Sound Barbara Mae Gayle University of Portland Nancy A.Burrell University of Wisconsin—Milwaukee
LAWRENCE ERLBAUM ASSOCIATES, PUBLISHERS Mahwah, New Jersey London
This edition published in the Taylor & Francis e-Library, 2009. To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk. Copyright © 2002 by Lawrence Erlbaum Associates, Inc. All rights reserved. No part of this book may be reproduced in any form, by photostat, microform, retrieval system, or any other means, without prior written permission of the publisher.
Lawrence Erlbaum Associates, Inc., Publishers 10 Industrial Avenue Mahwah, NJ 07430
Cover design by Kathryn Houghtaling Lacey
Library of Congress Cataloging-in-Publication Data
Interpersonal communication research: advances through meta-analysis/edited by Mike Allen…[et al.]. p. cm. Includes bibliographical references and index. ISBN 0-8058-3131-2 (cloth: alk. paper) ISBN 0-8058-3132-0 (pbk.: alk. paper) 1. Interpersonal communication. 2. Interpersonal communication—Research. I. Allen, Mike, 1959– BF637.C45 I644 2001 153.6–dc21 00–067768 CIP ISBN 1-4106-0455-1 Master e-book ISBN
Preface: On Numbers, Narratives, and Insights Regarding Interpersonal Communication Raymond W.Preiss and Mike Allen
I: INTERPERSONAL COMMUNICATION RESEARCH AND META-ANALYSIS 1 Meta-Analysis and Interpersonal Communication: Function and Applicability Mike Allen and Raymond W.Preiss 2 Meta-Analysis in Context: A Proto-Theory of Interpersonal Communication Charles R.Berger
II: INDIVIDUAL ISSUES IN INTERPERSONAL COMMUNICATION 3 An Overview of Individual Processes in Interpersonal Communication Barbara Mae Gayle and Raymond W.Preiss
4 Sex Differences in Self-Esteem: A Meta-Analytic Assessment Erin Sahlstein and Mike Allen
5 Comparing the Production of Power in Language on the Basis of Gender Lindsay M.Timmerman
6 Social Skills and Communication Brian H.Spitzberg and James Price Dillard
III: DYADIC ISSUES IN INTERPERSONAL COMMUNICATION 7 An Overview of Dyadic Processes in Interpersonal Communication Barbara Mae Gayle and Raymond W.Preiss
8 Sexual Orientation of the Parent: The Impact on the Child Mike Allen and Nancy Burrell
9 Similarity and Attraction Kimo Ah Yun
10 Self-Disclosure Research: Knowledge Through Meta-Analysis Kathryn Dindia
11 The Effects of Situation on the Use or Suppression of Possible Compliance-Gaining Appeals Dale Hample and Judith M.Dallinger
IV: INTERACTIONAL ISSUES IN INTERPERSONAL COMMUNICATION 12 An Overview of Interactional Processes in Interpersonal Communication Barbara Mae Gayle and Raymond W.Preiss
13 A Synthesis and Extension of Constructivist Comforting Research Mike Allen
14 Divorce: How Spouses Seek Social Support Nancy A.Burrell
15 Couples Negotiating Safer Sex Behaviors: A Meta-Analysis of the Impact of Conversation and Gender Mike Allen, Tara M.Emmers-Sommer, and Tara L.Crowell
16 Argumentativeness and Its Effect on Verbal Aggressiveness: A Meta-Analytic Review Mark A.Hamilton and Paul J.Mineo
17 Sexual Coercion and Resistance Tara M.Emmers-Sommer
18 A Meta-Analytic Interpretation of Intimate and Nonintimate Interpersonal Conflict Barbara Mae Gayle, Raymond W.Preiss, and Mike Allen
V: META-ANALYSIS AND INTERPERSONAL COMMUNICATION THEORY GENERATION 19 An Analysis of Textbooks in Interpersonal Communication: How Accurate are the Representations? Mike Allen and Raymond W.Preiss
20 How Does Meta-Analysis Represent Our Knowledge of Interpersonal Communication Daniel J.Canary and Michelle J.Mattrey
21 Better Living Through Science: Reflections on the Future of Interpersonal Communication Mary Anne Fitzpatrick
22 The State of the Art of Interpersonal Communication Research: Are We Addressing Socially Significant Issues? Michael E.Roloff
Preface On Numbers, Narratives, and Insights Regarding Interpersonal Communication Raymond W.Preiss and Mike Allen
The study of interpersonal communication is one of the more vibrant domains for social scientific theorizing and investigation. This interest is warranted, as the dyad has long been viewed as the nexus for message exchange and relationship evolution. As might be expected in an area dedicated to the study of the nuances and perplexities of social discourse, the complexity of the relational issues embedded in the interpersonal context is both intriguing and bewildering. Those interested in systematically understanding the richness of social life must address germinal issues: how and why individuals are attracted to certain others, how talk synchronizes perceptions and behaviors, or how and why individuals employ strategic messages to achieve relational outcomes. Of course, the list of “fundamental” issues is long and our journals provide a record of the conversations between scholars seeking adherence to various positions on that long list. This book is about those conversations. Contributors approach their tasks from various perspectives and with numerous agendas. All of us, however, share the commitment to establish reliable generalizations about inter-personal communication in ways that can be properly described as “scientific.” We search for stable, unbiased, predictable generalizations that operate within clearly defined interpersonal parameters (see Allen & Preiss, 1993, for a discussion of these qualities). Locating stable, unbiased, predictable generalizations has been complicated by the splintering of the macrodomain of interpersonal communication into subgroupings or rubrics that share common (or disparate) features. Issues
of gender, conflict, communication competence, or group cohesion blur the oversimplified titles of our courses and force us to broadly consider the “long list” as we explore the ways human utterance functions interpersonally. We seek the generalizations that unify and explicate the interpersonal context. This quest is a daunting task. Although every theory is assessed by its ability to produce meaningful generalizations, there are many theories to consider and many experimental findings to evaluate. Virtually every issue on the long list is contested and advocates offer important observations and key findings that are consistent with their perspectives. Faced with disparate theoretical approaches, assumptions, and methods, it is not surprising that progress has been uneven and that some theoretical questions resist interpretation. There is, it would seem, no lack of interpersonal communication research. There is, it is certain, little consensus on the interpersonal communication generalizations rooted in the experimental evidence. We believe that social scientific progress in understanding interpersonal communication would be served best if the literature could be simultaneously expanded and summarized. Before people accept claims regarding the planetary origins of men and women, the literature on relational communication should be gathered and systematically summarized. These summaries would identify voids in the literature and draw attention to accepted issues that are supported by only a limited number of findings. In short, a comprehensive review of interpersonal communication would provide fertile soil for the next generation of studies. As research accumulated, the new primary research would be folded into ever-widening reviews of the issues explored in the investigations. Two methods for summarizing literature and reviewing findings are currently being used: narrative summaries and meta-analysis.
THE LOGIC OF NARRATIVE SUMMARIES The narrative review or summary is the traditional verbal description of a body of literature (Pillemer, 1984) and the qualitative method for evaluating research on a given topic (Rosenthal, 1984). In most instances, the narrative reviewer will explicate a basic assumption or question and classify existing research using a vote-counting system (Did the studies on the roster of germane research detect a significant effect? Was the significant effect in the predicted direction? Were the significant effects attributable to a competing theory?). The reader is asked to tally the votes (confirming or nonconfirming tests) and render a judgment regarding the question of interest.
The outcome of the narrative summary may range from strong support (a uniform confirming vote count) for some proposition to no support (the failure to detect confirming votes). Of course, interesting questions and the subtle texture of interpersonal communication issues will rarely produce a clear-cut vote count. If 60% or 70% of the votes confirm a proposition, the narrative reviewer must question the robustness of the relationship and provide a rationale for nonconfirming outcomes. If 30% or 40% of the votes are confirming, the reviewer must impugn the relationship and express the concern that little progress has been made in understanding such an important issue. In either outcome, more research will certainly be required. The difficulty with the narrative review process involves the tendency for narrative reviewers to treat each qualifying vote as being 100% accurate. Because empirical findings are assessed probabilistically, it is a tautology that the findings of any one study may be the result of sampling error. Usually, narrative reviewers do not consider the possibility of Type I (false positive) or Type II (false negative) error as factors influencing trends in the primary research. Instead, experts tend to introduce intervening variables that explain apparent inconsistencies in the experimental record. In instances where discrepancies resist the philosopher’s stone of the confounding variable, narrative reviewers may assess sample characteristics, research designs, or statistical methods as the source of contradictory findings. This produces a web of issues, theoretical and methodological, that deflect attention away from hypothesized relationships. Of course, expert narrative summaries are essential components of the scientific enterprise, as they provide an avenue for subjective interpretation, reformulation, and reappraisal. If the goal is to assess evidence, however, this approach usually possesses the liability of nonrepresentativeness. In the course of making the case for an innovative interpretation or conclusion, the reviewer elevates certain studies as exemplars of the feature of interest. The difficulty here is that narrative reviews usually do not employ explicit rules or the methods used to locate primary evidence, how the reviewer determined which studies were germane to the analysis, and what criteria were used to determine whether or not an effect was present. The reader is asked to consider the exemplar studies in the context of a theoretical narrative or story that explains what the findings mean. Although reviewers make the case that the confirming evidence (the vote) is consistent with a novel interpretation, the reader is often not told why nonconfirming exemplars were excluded from the review. The narrative summary offers an important venue for experts to advance their informed conclusions about a domain of literature. The
risk is that in making the case for one interpretation, the ballot box for the vote count can be “stuffed” with nonrepresentative example studies. When nonconforming studies are mentioned, new variables may be introduced to explain discrepancies. The intervening variables may have been studied in only a limited number of investigations, and applying them to an entire domain runs the risk of overgeneralization. For these reasons, it is difficult for the vote-counting method used by narrative summaries to present a balanced portrayal of a large domain of literature.
THE LOGIC OF META-ANALYTIC SUMMARIES The term meta-analysis (Hedges & Olkin, 1985; Hunter, Schmidt, & Jackson, 1982) refers to a cluster of research procedures for aggregating primary research findings and estimating the direction and magnitude of effects. By coding and transforming outcomes into a common metric, it is possible to combine results across studies, estimate average effects, and detect moderator variables. In this book, we use variance-centered forms of meta-analysis (Hunter & Schmidt, 1990). This procedure requires the calculation of a weighted average correlation that is tested for homogeneity of the sample. The homogeneity test, using chi-square, compares the expected variability in the sample of correlations to the actual variability in the observed correlations. A significant chi-square indicates the existence of greater variability than would be expected due to random sampling error. This would be a sign of the probable existence of a moderator variable and a reason to interpret the average effect cautiously. These techniques quantify the magnitude of an effect across a body of research (Johnson, Mullen, & Salas, 1995) and may be used in a variety of strategic ways. Preiss and Allen (1995) suggested that metaanalyses can serve four functions in the review process. They grouped the meta-analytic procedures in a 2×2 matrix based on the researcher’s goal (providing a historical record vs. hypothesis testing) and the scope of the review (resolving a specific question vs. summarizing a research domain). This scheme shows the utility of meta-analytic procedures in addressing questions encountered by researchers interested in interpersonal communication (see Fig. P1). Of course, the categories are extreme cases, and middle positions exist. They serve the purpose of illustrating the flexibility of meta-analysis as a research tool.
FIG. P1. Approaches to meta-analysis.
Meta-analyses in this category address a single question or assess a limited number of conceptual issues. Usually the answer or implications of the findings will radiate to a larger set of theoretical issues. An example in the area of interpersonal communication is the Allen and Burrell (chap. 8, this volume) meta-analysis on the effects of parents’ sexual orientation on the adjustment of their children. The evidence on this narrow, substantive question has theoretical, legal, and policymaking ramifications. This sort of meta-analysis is often used when earlier reviews (narrative or meta-analytic) have located unresolved inconsistencies and demonstrated how broad issues may by clarified by untangling a limited set of outcomes. Narrow-descriptive meta-analyses should be employed in ways that reflect the literature in the domain, not by restricting the definition of the domain. For example, if the interpersonal researcher was interested in the narrow question of the relation between shyness and loneliness, she or he would aggregate all instruments used to assess these variables, code for possible moderator variables resulting from the various measurement instruments, and report the results of the moderator search. Unless there are compelling theoretical reasons to do so, it would be less appropriate to restrict the domain by selecting a single, commonly used instrument measuring each variable and summarizing only the studies using that instrument. Of course, less desirable still would be
aggregating studies using obscure, specialized instru- ments when accepted, standardized tools were available. In any event, theoretical justifications for all such decisions are essential and restricted definitions or capricious selection criteria merit cautious interpretation. Narrow-Inferential Meta-Analysis
Although meta-analyses in this category address a single question or a limited number of conceptual issues, the goal of the meta-analysis is to test hypotheses. If the researcher establishes generalizations within a narrow set of conditions, additional theory development is possible and the scope of the findings may be increased. An example in the area of interpersonal communication is the Ah Yun (chap. 7, this volume) metaanalysis on the effects of attitude similarity on attraction. Virtually all interpersonal researchers have followed the discussion between scholars holding competing views on this issue. Ah Yun tests three hypotheses, searches for moderators, and offers methodological suggestions that would allow for tests of alternative interpretations. Narrow-inferential meta-analyses are now commonly used to summarize the evidence in a given area. This popularity is probably due to the recent arrival of meta-analysis as a research tool. Because decades of narrative reviews have failed to resolve rather basic issues, meta-analytic investigations are being devised that set the stage for more sophisticated research. By establishing whether a relation exists and if so, under what conditions, the researcher establishes baseline summaries. We believe that, over time, progressively more complex issues will be examined meta-analytically. Comprehensive-Descriptive Meta-Analysis
Early research using meta-analysis treated the empirical summary as a historical document (Glass, 1976, 1983). The logic here is that metaanalytic assessment constitutes a pragmatic tool, a historical report card of the research in a given area. Some of the early literature asserts that empirical aggregation is atheoretic, but that view has faded (see Preiss & Allen, 1995). When a comprehensive review of an entire body of research is conducted, the results have theoretical and pragmatic implications. The Hamilton and Mineo (chap., 16, this volume) review of the literature on argumentativeness and verbal aggressiveness is an example of this approach to meta-analysis. The contributors review a substantial body of literature focusing on an important theoretical relationship. They probe the theorized associations from several directions, assess instruments, and offer a critique of assumptions and applications.
The comprehensive-descriptive meta-analysis provides a historical record that is either consistent or inconsistent with theoretical expectations. This justifiably creates tensions that prompt reflection and revaluation. A consistent history draws attention toward application and policy. For example, a hypothetical finding that communication skills training increases relational satisfaction would encourage counselors to employ this technique in their practices. An inconsistent record on this issue would force researchers and practitioners to reconsider and refine their views on the role of communication in the process of relationship maintenance. Comprehensive-descriptive meta-analyses are compelling because they usually embrace large bodies of literature on complex topics. The results may validate decades of practice or cast a shadow over commonly accepted maxims. For this reason comprehensivedescriptive reviews are inherently political and may significantly affect future research. Comprehensive-Inferential Meta-Analysis
Perhaps the most ambitious enterprise, the comprehensive-inferential form of meta-analysis aggregates large bodies of literature and tests the assumptions and outcomes predicted by one or more theories. Usually these endeavors involve hundreds of studies, scores of moderators, and, often, the construction of mathematical models. Dindia’s (chap., 10, this volume) reviews of self-disclosure, Allen’s (chap., 13, this volume) work on comforting strategies, or Hample and Dallinger’s (chap., 11, this volume) examination of compliance-gaining appeals might be classified in this category. Usually, comprehensive-inferential meta-analyses serve to consolidate bodies of literature and become the benchmark for new theoretical contributions. The act of isolating generalizations does not imply “correct answers.” Rather, the stable findings become a platform for innovative interpretations and original insights.
THE ORGANIZATION OF THE BOOK We believe that both narrative and meta-analytic reviews are vital to the scientific study of interpersonal communication. The risks associated with narrative reviews involve distorting average effects by considering example findings as being 100% accurate, building theoretical stories that ignore counterexamples, and introducing intervening variables that may obscure fundamental generalizations. In later chapters, Canary and Mattrey (chap., 20, this volume) and Fitzpatrick (chap., 21, this volume) point to risks asso- ciated with meta-analytic reviews. They address
search procedures and coding decisions as a potential source of bias. The dynamic interplay of the narrative review and the meta-analytic review deepens understanding and enriches both approaches. In selecting and arranging chapters in this volume, we were mindful of the issues addressed and the potential of chapters to establish significant generalizations about interpersonal communication. It became clear that meta-analytic methods were not novelties in investigations focusing on interpersonal communication. We also sought out issues that were fundamental to discussions underway in communication journals and that reflected current (or enduring) controversies. There was no shortage of potential candidates. We present the issues using five general themes. In Part I, we set the stage for the independent meta-analyses. Two chapters make the case for using meta-analysis to address meaningful issues in the interpersonal communication literature. In Part II, we present an overview of individual characteristics in interpersonal communication and three meta-analyses reflecting this theme. Contributors discuss sex differences in self-esteem, sex differences in powerful language, and social skills. In Part III, literature emphasizing the dyadic approach to interpersonal communication is considered. In this part, we provide an overview and contributors present summaries on sexual orientation and children, similarity and attraction, self-disclosure, and compliance gaining. Part IV explores the interactional approach to interpersonal communication. Following an overview, meta-analyses are presented on comforting, social support, safe-sex interactions, argumentativeness, conflict styles, and sexual coercion. In Part V, contributors consider the impact of the meta-analyses on our understanding of interpersonal communication. Authors discuss pedagogy and textbooks, assess representations of the interpersonal communication literature, evaluate the prospects for future theory development, and chart the course for new investigations. The issues summarized in these chapters reinforce Hamilton and Hunter’s (1998) contention that meta-analysis is a necessary tool for theory construction. The conclusions reached by the contributors to this volume provide three invitations for future research. First, the findings may be baselines for new, primary investigations that may be integrated into more sophisticated meta-analyses. A second invitation is for scholars to empirically summarize the variables that are conceptually aligned with the findings of these meta-analyses. The goal here is to establish sets of interlocking generalizations that provide breadth of understanding about interpersonal communication. Finally, readers are invited to replicate and refine the summaries. Allen and Preiss (1993) offered preliminary guidelines for these replications and possible benefits, including testing mathematical models, enlarging samples, and generalizing results.
Meta-analysis is an inherently public enterprise. All meta-analysts explicitly state search procedures, study inclusion standards, coding rules, and aggregation formulas. Any informed reader of any theoretical or ideological persuasion is welcome to sort through the primary literature and add, recede, reformulate, or reinterpret it. When readers accept the invitations to replicate, interlock, and extend empirical summaries, we believe that interpersonal communication researchers will, indeed, see advances through meta-analysis.
REFERENCES Allen, M., & Preiss, R.W. (1993). Replication and meta-analysis: A necessary connection. Journal of Social Behavior and Personality, 8, 9–20. Glass, G.V. (1976). Primary, secondary, and meta-analysis of research. Educational Researcher, 5, 3–8. Glass, G.V. (1983). Synthesizing empirical research: Meta-analysis. In S.A.Ward & L.J. Reed (Eds.), Knowledge structure and use: Implications for synthesis and interpretation (pp. 399–421). Philadelphia: Temple University Press. Hamilton, M.A., & Hunter, J.E. (1998). A framework for understanding: Metaanalysis of the persuasion literature. In M.Allen & R.W.Preiss (Eds.), Persuasion: Advances through meta-analysis (pp. 1–28). Cresskill, NJ: Hampton. Hedges, L.V., & Olkin, I. (1985). Statistical methods of meta-analysis. Orlando, FL: Academic Press. Hunter, J.E., & Schmidt, F. (1990). Methods for meta-analysis. Newbury Park, CA: Sage. Hunter, J.E., Schmidt, F.L., & Jackson, G.B. (1982). Meta-analysis: Cumulating findings across research. Beverly Hills, CA: Sage. Johnson, B., Mullen, M., & Salas, E. (1995). Comparison of three major meta-analytic approaches. Journal of Applied Psychology, 80, 94–106. Pillemer, D. (1984). Conceptual issues in research synthesis, Journal of Special Education, 18, 27–40. Preiss, R.W., &. Allen, M. (1995). Understanding and using meta-analysis. Evaluation and the Health Professions, 18, 315–335. Rosenthal, R. (1984). Meta-analytic procedures for social research. Beverly Hills, CA: Sage.
I Interpersonal Communication Research and Meta-Analysis
1 Meta-Analysis and Interpersonal Communication: Function and Applicability Mike Allen and Raymond W.Preiss
Meta-analysis provides a technique that allows researchers to aggregate data and summarize existing reports. The technique is used to reduce or eliminate various sources of artifact and statistical error. Additionally, meta-analysis provides an ability to examine the impact of various design issues and various theoretical possibilities. In this chapter, we explain what meta-analysis is and the justification for using the procedure. Our goal in this chapter is to enhance understanding of meta-analysis so that reading subsequent chapters is easier. The discussion does not consider a number of technical issues relating to the mathematical or methodological features of meta-analysis, as those discussions do exist and can be found in various textbooks dealing with meta-analysis (e.g., Cooper & Hedges, 1994; Hunter & Schmidt, 1990).
WHY USE META-ANALYSIS? The problem in the social sciences is the use of statistical inference to make decisions about whether the experimental or survey results are not due to random chance. Consider the simple issue of determining whether a coin flip is biased by evaluating whether heads or tails is a more likely outcome. If we flipped the coin 100 times and the coin flip 3
4 Allen and Raymond
is fair, the outcome should be theoretically 50 heads and 50 tails. We could flip the coin 100 times and if all 100 flips come out either heads or tails, we would suppose the coin is biased; that is, more likely to produce one outcome over the other. The problem is that very seldom does the theoretically random result (50/50) or the completely biased (100% heads or tails) outcome ever occur. It is much more likely that the outcome will be somewhere between totally random and totally biased. Suppose the outcome of our coin flip is 51 tails and 49 heads. The results depart from a perfectly random distribution (50 of each), but a person would probably not argue that the deviation constitutes evidence of bias. Similarly, a coin flip that generates a 99 to 1 outcome would probably make the case that the coin flip was biased. The key is that ultimately we will set up some decision rule that makes the distribution of heads and tails considered biased or acceptable. For example, any coin flip is biased if, when flipped 100 times, more than 66 heads or tails appear as an outcome. In a sense, the decision rule is arbitrary because a 66–34 split would be considered a fair coin flip, but a 67–33 split would be considered evidence of a biased coin flip. However, such a procedure is in fact the standard operating practice of the social sciences. The alpha (or Type I error rate) is considered at 5%, or more commonly as texts would say, the statistical test is considered significant at p