Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

This article was downloaded by: [Flinders University of South Australia]

On: 30 January 2015, At: 01:48


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Communication Methods and Measures


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/hcms20

Contributing to Theory and Knowledge in


Quantitative Communication Science
a a
Michael D. Slater & Laurel S. Gleason
a
The School of Communication , The Ohio State University
Published online: 05 Dec 2012.

To cite this article: Michael D. Slater & Laurel S. Gleason (2012) Contributing to Theory and
Knowledge in Quantitative Communication Science, Communication Methods and Measures, 6:4,
215-236, DOI: 10.1080/19312458.2012.732626

To link to this article: http://dx.doi.org/10.1080/19312458.2012.732626

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Communication Methods and Measures, 6:215–236, 2012
Copyright © Taylor & Francis Group, LLC
ISSN: 1931-2458 print/1931-2466 online
DOI: 10.1080/19312458.2012.732626

ARTICLES
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Contributing to Theory and Knowledge in Quantitative


Communication Science
Michael D. Slater and Laurel S. Gleason
The School of Communication, The Ohio State University

We propose a framework for understanding how quantitative researchers typically seek to contribute
to theory and generalizable knowledge in communication, noting some of the strengths and weak-
nesses of the various approaches typically used. The major approaches are described, and a content
analysis of articles from four issues each of Journal of Communication, Human Communication
Research, and Communication Research spanning parts of 2008 and 2009 provides information
about their relative frequency of use in the field. We discuss how these categories can be employed
by researchers when conceptualizing and articulating the theoretical contributions planned in their
research and by reviewers, editors, and researchers in critically assessing such contributions in the
work of others.

Contributing to theory and generalizable knowledge is typically necessary for publication in


major journals in the field of communication and elsewhere. However, advice about what it actu-
ally means to contribute to theory and generalizable knowledge in this field in ways that merit
publication tends to be rather vague. The purpose of this article is to characterize typical ways
researchers in quantitative communication research seek to contribute to theory or otherwise con-
tribute to knowledge. We include a content analysis of one year of articles in three major general
interest communication journals that illustrates not only that the approaches we characterize are
used but also provides a rough estimate of their relative frequency of use in the field’s leading
journals.
Our focus is on understanding how researchers attempt to contribute to theory and knowl-
edge creation in quantitative communication science. We do not address theory development or

The authors appreciate the contributions of the students in The Ohio State University School of Communication’s
graduate course in Theory Development, out of which this paper has developed.
Correspondence should be addressed to Michael D. Slater, Social and Behavioral Sciences Distinguished
Professor, 3022 Derby Hall, School of Communication, The Ohio State University, Columbus, OH 43210. E-mail:
slater.59@osu.edu
216 M. D. SLATER AND L. S. GLEASON

knowledge creation as practiced in cultural and critical studies. We are not providing critiques
of communication theory, as have been offered by several communication scholars (Bryant &
Miron, 2004; Pavitt, 2000, 2009). We are not critiquing the breadth or depth of existing commu-
nication theory or making suggestions regarding techniques to stimulate creative thinking about
new theory (see Berger, Roloff, & Roskos-Ewoldsen, 2009). We share with these authors a pas-
sion for explanation and for understanding underlying mechanisms regarding communication
phenomena. Our focus, though, is on better understanding the day-to-day intellectual work of
most empirical communication scholars—the normal-science process of building knowledge and
contributing to existing theory. Our effort differs from prior work in that we take a bottom-up
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

approach to developing a descriptive framework of empirical research approaches in communi-


cation science rather than a prescriptive top-down model based on a given conceptualization of
what theory is or should be.
We believe that no theory, irrespective of merit, is ever a finished product. Concepts are refined,
operationalizations are challenged, boundary conditions are explored, hypothesized relations are
tested more rigorously or richly using new methods, explanatory mechanisms are examined, and
new applications of the theory are proposed. These tasks are the usual daily work of a com-
munication scientist. Thinking clearly about what a study is doing to contribute to theory and
generalizable knowledge, and articulating that contribution clearly, is much of what makes a
study a useful contribution. It is our hope this article will facilitate such efforts.

A FOCUS ON THEORY AS PROPOSING CAUSAL PROCESSES

Authorities on theory in the social sciences point out that most theory attempts to provide expla-
nations of a causal process (Jaccard & Jaccoby, 2010; Reynolds, 2007; Shoemaker, Tankard, &
Lasorsa, 2004). Berger et al. (2009) also strongly endorse the centrality of explanatory theo-
ries that delineate mechanisms explaining communication phenomena. Similarly Pavitt (2009)
regards explanation as the principal contribution of theory.
Therefore, we here focus on the primary role of theory in communication science as the pro-
vision of explanation, of proposing causal processes, the explanation of “how” and “under what
circumstances” in ways that result in empirically testable and falsifiable predictions. This, as
Popper (1959) has argued, is the essence of the scientific enterprise. Theory permits empirical
studies to create knowledge applicable across a variety of situations, contexts, and populations,
by creating understanding of underlying processes, mechanisms and principles. Theory-focused
research will typically seek to test such theories, extend their utility and explanatory power, and
identify the contingencies and boundary conditions limiting their applicability.
Communication science is concerned with more than theory testing and development.
Research may be primarily methodological and contribute to building knowledge by creating
tools for scientific examination of phenomena and for developing and testing theory (Greenwald,
2012). Other research may be primarily observational and descriptive. Observation typically pre-
cedes theory—theoretical explanation of phenomena is hardly possible without prior observation
of such phenomena. Observation may be informal; it may involve rigorous qualitative research;
it may, as in content analysis, be quantitative and empirical. Theory is typically then developed
or applied to explain observed phenomena such as effects, through search for causal explanation
and boundary conditions (McLeod & Reeves, 1980).
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 217

We build upon the Shoemaker et al. (2004) and Jaccard and Jacoby (2010) discussions of
approaches to causal process theory by providing a concise account of ways researchers in
communication typically seek to contribute to such theory. In so doing, we note some of the
opportunities and limitations characterizing these strategies. To assess the reliability and utility of
this framework, we use it to parse approaches used to contribute to theory over the course of a year
of research published in three major general interest communication science journals. The results
provide us a sense of the relative frequency with which various strategies are employed, and
demonstrate that the framework can be used to characterize contributions of empirical research
articles in the discipline.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

THEORY ELABORATION AND REFINEMENT

Strategy 1: Addressing Fundamental Conceptual Issues

Writers on theory, including Reynolds (2007), Stinchcombe (1968), and Shoemaker et al. (2004),
emphasize the centrality of concept explication (Chaffee, 1991) as the starting point in theory
development. Accordingly, one way to contribute to theory is to clarify, refine, or challenge con-
ceptualization in a given theory. Because conceptual distinctions and operational definitions are
inextricably related in social science research, theory contributions of this sort can be identified
in several ways. One set of strategies for contributing to theory involves addressing fundamental
conceptual issues with regard to a theory or previously theorized relations.

Strategy 1a: Redefining and re-operationalizing a key concept

Sometimes a concept is reconceptualized and redefined and alternative operations are intro-
duced to reflect those conceptual refinements—a marriage of reconceptualization and improved
measurement or manipulations. One researcher, for example, may conceptualize violent media
content as media representation of harm of any kind. Another may conceptualize violent media
content specifically as showing harm intentionally inflicted upon another, and a third may restrict
a definition of violent media content to harm that takes place outside of acceptable social rules—
excluding law enforcement activity as a form of media violence. Similarly, some researchers
studying social support might conceptualize support as inherently prosocial; others may be con-
cerned as well with unwanted support or support received for maintaining destructive feelings
or behaviors (see Vangelisti, 2009). A study that noted and compared such variations in concep-
tualization and operationalization would have the potential to reconcile seemingly inconsistent
results in prior research, thereby moving theory forward.
Oliver, Yang, Ramasubrmanian, Kim, and Lee (2008) provide an example of this strategy when
they suggest a means of reconceptualizing the third-person effect. The third-person effect, in its
simplest formulation, is the oft-found belief that others are more influenced by media content
than oneself is, which may lead to naïve lay assumptions about strong media effects. Oliver and
colleagues argue that people also implicitly take into account beliefs about audience selectivity
and interpretation with respect to media content. As a result, the third-person effect may take the
form of believing that media content serves to reinforce others’ existing beliefs rather than to
change them, especially, as is often the case, when the direction of potential influence of media
218 M. D. SLATER AND L. S. GLEASON

content is somewhat ambiguous—a rather less naïve lay view of media effects than has been
previously argued. In other words, Oliver and colleagues reconceptualize the “effect” in the third
person effect to include beliefs about attitude reinforcement as well as attitude change.

Strategy 1b: Critiquing and testing key theoretical assumptions

A variant on this approach is more explicitly a critique of existing theoretical formulations:


a key theoretical assumption that has gone untested or has been inadequately tested is identi-
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

fied, operationalized, and subjected to empirical examination. For instance, Valkenburg and Peter
(2008) point out that previous research had assumed effects of adolescent on-line experimenta-
tion on self-concept result from the breadth of interaction partners possible on-line, but that this
theoretical assumption had not been previously explicated and tested,. This therefore became one
of the objectives of their study.

Strategy 1c: Critiquing prior operationalizations that have conceptual as well as


methodological implications

Still another variant is a fundamental critique of operationalization of concepts in existing


theory, with significant implications conceptually as well as methodologically. For instance, a
researcher might argue that operationalizing aggression in terms of symbolic acts such as attack-
ing inflatable dolls or avatars in computer games is conceptually inappropriate because such acts
do not transgress boundaries of socially acceptable norms and only such transgressions should be
considered enacted aggression. We note that the distinction between this strategy and Strategy 1a
regarding conceptual refinement relates to the source of the concern—does it arise from a revis-
iting of how a concept is explicated and how a variable is defined, or does it arise from the obser-
vation of nonobvious and perhaps unintended conceptual implications of operational choices?
At times, this strategy arises as a result of methodological development (consistent with
Greenwald’s 2012 argument concerning the theoretical importance of methodological innova-
tion), permitting empirical researchers to both theorize and test questions that otherwise might be
inaccessible to quantitative examination, or to test them in ways that are more rigorous and appro-
priate than in prior research. For example, theories of the interplay of social groups and individual
behavior can be tested using multi-level models, resulting in the potential to test theories of social
influence using methods better matched to the underlying theoretical ideas as well as encouraging
such cross-level theorizing. For example, Cho (2008) uses such multilevel methods to examine
theoretical claims regarding how the amount of exposure to political ads (a media-market level
variable) influences individual political information-seeking and interpersonal communication
behavior.
It should be acknowledged that the above approaches to contributing to theory can be double-
edged swords. On one hand, overly broad conceptualizations can be refined, sloppiness in the
relationships between conceptualization and operationalization or analysis strategy corrected, or
the inherent complexity embedded in a concept elaborated. On the other hand, simply increasing
the range of conceptual definitions used can lead to theoretical confusion, with the same term
evidently meaning different things to different researchers and resulting in alternative research
designs and analytic strategies—a situation all too familiar to anyone who has reviewed the
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 219

literature in a well-developed research area. Efforts that seek to integrate or compare competing
definitions of a construct, identify conceptual slippage that may have led to confusion in ear-
lier research, or compare alternate operationalizations that may have led to conflicting inferences
about theory in past research will typically have substantially greater value than those that simply
further multiply the many conceptualizations already being used.

Strategy 2: Extending the Range of Existing Theory and Empirical Findings

Strategy 2a: Adding a theoretically relevant independent variable


Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Sometimes the central claims of previous work are elaborated or the range of the prior work’s
applicability is extended, typically by adding variables or testing the theory in new contexts or
populations. One such approach is taken when a theoretically relevant independent, exogenous
variable is added. Obviously, one can always add a new independent variable to an existing the-
ory or to supplement existing findings. Doing so does not necessarily represent a significant
advance worthy of publication. A sound theoretical and substantive argument for the importance
of incorporating the new variable is essential.
For example, Dixon (2008) included a measure of television network news exposure in addi-
tion to a measure of television local news exposure in his investigation of the connection between
news viewing and stereotypical perceptions of African Americans. In so doing, he expands evi-
dence of news exposure effects on increasing stereotypic racial perceptions, and raises socially
important concerns about the content and effects of national news. Such research increases
the robustness of theoretical claims and their substantive value in describing socially important
phenomena.
It is worth noting how Dixon (2008) also made an argument for the theoretical importance of
his findings by noting that national news viewers might otherwise be presumed to be relatively
well-informed and less likely to hold stereotypic beliefs, and then provided a sophisticated the-
oretical causal process explanation for why such viewing would occasion greater stereotyping
rather than being associated with reduced stereotyping. The contribution made by a manuscript,
of course, is a subjective assessment made by reviewers and editors. Often, as in this example, it
is not one major theoretical insight but rather the combination of substantive advances with inter-
esting theoretical implications, intelligently interpreted and methodologically well-executed, that
adds up to a contribution deemed worthy of publication in leading journals.

Strategy 2b: Testing existing theories/prior findings in a new communication


context

Perhaps the simplest, though sometimes the least interesting approach, to contributing to the-
ory is testing a theory in a new context. There is always a new population, a new social situation, a
different problem, to which a theory can be applied. The important question, as elaborated below,
is—does it matter, and why? We have found it useful in clarifying our own thinking on this matter
to distinguish two types of context for extending theory: different communication phenomena or
other social contexts and outcomes.
Strategy 2b involves exploring or testing established theoretical perspectives in a new com-
munication channel, genre, or other class or category of messages. This would be instantiated
220 M. D. SLATER AND L. S. GLEASON

by studies that explore theoretical issues relevant to relatively recent communication technolo-
gies such as mobile telephony or social media, topics of obvious interest to communication
researchers. For example, Wojcieszak and Mutz (2009) examined online discussion groups to
determine whether such groups were a venue for deliberative discussion concerning political
issues. They found that this was the case for groups that were not political in intent and in which
political discussion was incidental. This study not only provides information about political dis-
cussion in an emerging context, but also raises intriguing theoretical questions about the nature of
interpersonal discussion: namely, whether deliberative discussion may be more often found when
discussions are not primarily or overtly political.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Strategy 2c: New populations, locations, outcomes

Strategy 2c involves testing a theory in a new study population (such as testing a theory
developed in experiments with undergraduates with a noncollege student youth or adult popu-
lation, or a theory developed with a mostly majority population in a minority population), or a
new geographic location (such as testing the cross-cultural generality of a theory). This strategy
also includes employing a new and conceptually distinct dependent measure (either a measure
developed for the purpose or importing an existing measure not previously used in this research
context). For example, Clark, MacGeorge, and Robinson (2008) examined the evaluation of sup-
port strategies among children and young adolescents, a population in which this question had
not previously been addressed. Dixon (2008) also noted that the expansion of research of news
effects on racial stereotyping from a student to a general population sample was a substantively
important contribution to generalizability and the robustness of theoretical claims. As in many
studies, it is often an accumulation of various contributions such as population generalizabil-
ity, expanded context, and some nonintuitive findings, especially in a socially important context,
that lead to a contribution deemed worthy of publication in a major journal. An example of the
use of a novel dependent measure with important theoretical implications is the study by Appel
(2008), who extended cultivation research in an unexpected direction. Cultivation research often
provides evidence that exposure to television increases perceptions that the world is a danger-
ous place. Appel provides evidence for contexts in which exposure to fictional narratives might
encourage “just world” beliefs—beliefs that wrong-doing is reliably punished. In other words,
a similar theoretical mechanism might lead to two different outcomes with respect to audience
members’ confidence about the social world.
Generally, though, the theoretical contribution involved in simply testing a theory in a
new context or population may be modest at best, though the substantive importance may be
important—for example, finding that a theory such as the Extended Parallel Processing Model
(Witte, 1992) can be applied to constructing health behavior messages in the context of a disease
or other risk with which it has not been previously tested. Such research is more likely to find a
home in a specialty or applied journal than in a journal that serves the discipline as a whole.
While it is often the case that the theoretical (as opposed to substantive) contribution involved
in testing a theory in a new context or population is modest, it is possible that such research
provides a test of underlying conceptual issues. For example, Hayes, Matthes, and Eveland
(in press) show, using data from eight countries, that the effects of fear of social isolation on
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 221

attention to press reports of public opinion (predicted by spiral of silence theory) to some extent
vary based on cultural differences. In the case of the Clark et al. (2008) article, the framing
of the authors’ theoretical contribution was two-fold. First, the research was intended to be in
part descriptive—characterizing how different support strategies were perceived by children and
younger adolescents, a topic about which very little was known, and which might provide a basis
for developing theory-based work in the future. A second element—the one most relevant to
this strategy—was to provide a basis for distinguishing how support communication may operate
differently as a function of developmental changes from childhood through adulthood. In other
words, the authors’ argument was that because of developmental changes, the causal process of
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

support communication may well have to be conceptualized differently at different ages (e.g.,
for younger, more concrete thinkers, the offer of companionship may be more meaningful than
the provision of empathic verbal support). In this way, the research also became an examination
of theoretically-derived boundary conditions, as discussed later with respect to moderation, and
leads to the conclusion that different causal process models are needed for children versus young
adults.

THEORIZING ABOUT STRUCTURE OF A CAUSAL PROCESS

Strategy 3: Elucidating Mechanism and Theorizing Causal Process: Mediation

Why mediation and indirect effects matter

At the heart of theories of causal process is the notion of mechanism. To say that A causes C is
an assertion. Documenting such assertions certainly is important and plays a key part in commu-
nication research by providing evidence regarding the existence of a phenomenon of interest. One
cannot theorize about phenomena of which one is unaware. The fact that a given effect exists, say,
exposure to sexualized media content increases willingness to attribute blame to victims of sexual
assault, may be consistent with a theory of media effects. However, as Reynolds (2007) repeatedly
emphasizes, social science theory is concerned largely with explanation. This including explain-
ing how causal relationships take place. Therefore, hypothesizing intervening processes that
explain effects and testing mediation is a primary way in which researchers contribute to theory.

What is mediation?

Mechanism is typically represented as mediation; an independent variable A influences an


intervening (or endogenous, or mediating) variable B, which in turn influences an outcome C.
This is called an indirect effect. In the prior example concerning sexual media content, to what
extent would the effect be due to physiological impact of sexual arousal on judgment? Would the
effect be due to more accessible exemplars of seductive sexual behavior in memory? To making
self-concept associated with sexual aggression more accessible? By influencing beliefs about the
sexual willingness of victims? In other words, as Baron and Kenny (1986) and Hayes (2009)
point out, mediation addresses the fundamental question of “how?”
222 M. D. SLATER AND L. S. GLEASON

For example, Valkenburg and Peter (2008) examine whether effects of on-line interactions on
adolescent self-concept are mediated by on-line interactions with a wide variety of communica-
tion partners. And Kline, Simunich, and Weber (2008) examine how effects of equivocal versus
nonequivocal corporate reputation messages are mediated by perceptions of the severity of the
crisis.
A full discussion of mediation is beyond the scope of this article and has been summarized
elsewhere (e.g., Preacher & Hayes, 2008). However, because of the growing importance of study-
ing mediation in order to develop theoretical understanding of explanation and mechanism, we
look at various approaches to studying mediation in some detail. If methods with significant
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

limitations are inappropriately used, contribution to theory is substantially diminished; often,


causal processes may be inadequately theorized and specified as a result of using some of the
cruder approaches to examining mediation (a situation the first author has run into all too often
as a journal reviewer). Moreover, various approaches can provide distinctive opportunities to
contribute to theory, as described below.

Not all tests of mediation are of equal interest

First, a caveat: not all mediation arguments are created equal with respect to importance of
theoretical contribution. As Berger et al. (2009) point out, creating and testing a causal model that
includes mediation and indirect effects is not synonymous with creating and testing a theory, nor
does a slight change in the causal process proposed by an existing theory merit being called a new
theory. It is likely, given any causal relationship one can empirically identify, that some mediating
variable or indirect effects can also be identified and tested. Such ad hoc model development is
of limited value to the discipline. The important question, as Berger et al. (2009) note, is how
much the mediating or indirect relationship matters conceptually. That is, do the relationships
proposed arise from some underlying explanatory mechanism that helps us understand in a more
general way communication processes and effects? Obviously, this is a subjective judgment by
reviewers and editors. Researchers are well-advised to carefully develop the theoretical logic from
which their proposed mediating or indirect relationships are derived and to carefully articulate
the theoretical and substantive significance of the causal process models proposed. If the logic
of the mediating or indirect relationship is unclear or the theoretical and substantive significance
of the relationship is questionable, there is little reason to think that the findings will deserve
scarce journal space. Such logic is best reflected in clear hypotheses, well-developed rationale
for these hypotheses, and appropriate statistical analyses. In the following paragraphs we discuss
implications of different ways in which mediation can be assessed.

Strategy 3.1: Asserting causal process because a series of hypothesized links are
found in the data

Occasionally, one finds papers that hypothesize a causal chain of associations. In other words,
the research may hypothesize that A influences B and that B influences C. If these associations
are found, the researcher may claim support for a causal process in which A influences C through
B. Sometimes this claim is explicit, sometimes it is implicit. In either event, claim for a causal
process in which A influences C via B requires an explicit hypothesis concerning indirect effects,
appropriately theoretical justification for that hypothesis, and (as discussed below) an appropriate
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 223

statistical test. Simply hypothesizing and testing a series of bivariate relations does not in fact
represent theorizing and testing a causal process.

Strategy 3.2: Asserting a causal process because a structural equation model of the
proposed process provides good fit

Sometimes the above simple path model approach is bolstered by the use of structural equation
models and provision of a fit statistic suggesting that the proposed associations are a good fit to
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

the data. However, it may be that a variety of other causal arrangements inconsistent with the pro-
posed causal model would provide equally good fit (MacCallum, Wegener, Uchino, & Fabrigar,
1993). Moreover, there are a variety of ways to maximize fit that have little to do with assess-
ing a given proposed causal process (Holbert & Stephenson, 2008). Therefore, neither strategies
3.1 or 3.2 are to be recommended as stand-alone tests, as the mediating or indirect relationships
theorized are not in fact tested and likely have not been fully specified in the theoretical argument
or hypotheses.

Strategies 3.3 and 3.4: Assessing attenuation in a direct effect due to a hypothesized
mediator

A commonly used approach to assessing mediation is based on recommendations by Baron


and Kenny (1986), in which each link in the causal process among three variables is assessed, and
the attenuation in the relationship between A and C—between initial cause and effect—due to the
presence of the mediator is assessed (Strategy 3.3); a variant adapted for use in SEM (Holmbeck,
1997) is occasionally used instead (Strategy 3.4). Problems with the Baron and Kenny approach to
testing indirect effects are well-known and are thoroughly discussed in Hayes (2009). The bottom
line is that the Baron and Kenny approach does not provide a straightforward inferential statisti-
cal test, and is not recommended to assess theoretical claims about the presence of mediating or
indirect effects. In our view, one aspect of the Baron and Kenny approach can be useful descrip-
tively, if used to supplement and not supplant inferential tests of indirect effects. Sometimes, in
addition to the theoretical question about the presence of an indirect effect, there is also a the-
oretical or substantive concern regarding how much of a statistically significant relationship is
accounted for by a given mediator. Such a concern is likely to arise, for example, when theo-
rizing about mechanisms that might explain previously demonstrated relationships. Obviously, a
mediator that accounts for 50% of a relation explains more of that relation than one that accounts
for 5% or 10%.

Strategy 3.5: Testing alternative models

One of the most interesting and sophisticated approaches to exploring mediation from a theory
development standpoint is to compare model fit for alternative model specifications in structural
equation modeling. For recent examples, see Eveland, Hayes, Shah, and Kwak (2005) and Cho
and Boster (2008). The alternative models approach is more than simply another methodology for
testing mediation claims. This method (preferably paired with inferential tests of indirect effects)
permits the researcher to compare the fit of alternative causal paths in a given data set. In other
224 M. D. SLATER AND L. S. GLEASON

words, it provides a way to compare alternative theoretical explanations, that is, as long as the
alternative explanations are thoughtfully theorized. We believe that examination of alternative
theoretical explanations deserve wider use in communication.
For example, the researcher’s theory may suggest that political interpersonal discussion medi-
ates effects of advertising on willingness to make contributions to a candidate. Testing of alter-
native models allows the researcher to compare the fit of this hypothesized causal order with the
possibility that interpersonal discussion predicts exposure to the ads (people who are interested
in politics and engage in political discussion may pay more attention and recall ads better).
As a theory development strategy, alternative model testing acknowledges and explores alter-
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

native causal process explanations. It is important when interpreting the implications of such
comparisons for theory to recall that often more than one competing model provides adequate
fit to a given data set, though the fit of one may be superior to the fit of another. In such cases,
evidence may be said to favor the model with better fit, but does not exclude the alternative model.

Strategies 3.6 and 3.7: Statistical tests of indirect effects

Neither the Baron and Kenny approach nor the testing of alternative models provide a direct
inferential test of the indirect effect of A on C through B, though they can readily be combined
with such approaches. The indirect effect is sometimes tested by using techniques such as the
Sobel test (Sobel, 1988). This test is noted here as Strategy 3.6. Statistical problems with the Sobel
test are addressed through the use of bootstrapping techniques (Strategy 3.7), and software is
available which implements bootstrap approaches (see Preacher & Hayes, 2008). In other words,
such approaches support theorizing consistent with the emphasis on causal process discussed in
Jaccard and Jacoby (2010), Reynolds (2007), and Shoemaker et al. (2004).

Strategy 3.8: The desirability of longitudinal data for studying causal process

The above approaches to testing mechanisms can be adapted to longitudinal data (see Cheong,
MacKinnon, & Chew, 2003; MacKinnon, 2008; Preacher & Hayes, 2008). Longitudinal data
has advantages with respect to exploration of dynamic social processes employing mediation
tests supported by the temporal ordering of data. We refer to testing mediation in panel data as
Strategy 3.8. Tests of mediation in cross-sectional data are often suspect given the lack of temporal
order, and use of longitudinal data can strengthen claims when assessing causal process. In the
absence of longitudinal data, cross-sectional mediation analyses especially benefit from tests of
theoretically plausible alternative models to help assess questions of causal order. It should also
be noted that the utility of many longitudinal examinations of mediation to inform theory can also
be strengthened through hypothesizing about and testing alternative models, as competing causal
influences and orders can still be proposed in longitudinal data.

Strategy 3.9: Experimental tests of mediating relationships

Some researchers have argued that the most compelling way to demonstrate mediation is
by directly manipulating the presence or absence of the mediator (e.g., Spencer, Zanna, &
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 225

Fong, 2005). For example, by having confederates provide white participants with nonver-
bal cues similar to those often experienced by African-Americans, researchers (Word, Zanna,
& Cooper, 1974) were able to elicit behaviors consistent with African-American stereotypes
among these white participants. In other words, experimental manipulation was used to show
that nonverbal cues helped explain the influence of majority individuals on stereotypic behav-
iors. Spencer et al. (2005) also acknowledge that such manipulation of intervening variables
is often not possible or practicable, and that the more commonly used statistical methods are
necessary.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Strategy 4: Elucidating Contingency: Moderation, Interactions, Boundary Conditions

Strategy 4.1: Studies of moderation (interactions) and boundary conditions

Studies of simple causal or mediating processes often assume a causal process is uniform
in its effect across people, situations, and other contexts. As McLeod and Reeves (1980) have
noted, such uniformity of effect is rare in communication, and theory development often properly
involves identification of the contingencies of effects of theoretical interest, also referred to as
moderation or interaction. Often, a full grasp of a phenomenon can be characterized as a grasp of
the contingent conditions involved—when will something happen and when won’t it? What does
it depend upon? As Baron and Kenny (1986) note, while mediation answers the question “how?”,
moderation answers the question “when?”
Typically, moderation is concerned with circumstances—psychological differences, situa-
tional contexts, message variations—in which causal processes work differently. For example,
studies suggest effects of violent media content depend on audience member individual differ-
ences and social context, how the violence is portrayed, and the outcome of the violence for the
persons portrayed (see Huesmann & Taylor, 2006). Generally, the theoretical value and publisha-
bility of a moderation relationship depends on how much insight that moderation provides about
the underlying phenomenon—or, more directly, how smart and interesting the underlying theo-
rizing is. It is often possible to find a significant difference based on some measured individual
difference variable. What is harder is to find a difference that matters in a larger sense, one that
tells us something theoretically important in terms of understanding the underlying phenomenon.
For example, if a theory about effects of media violence focuses on the effects of exemplars
in memory on people with limited capacity for behavioral inhibition, demonstrating modera-
tion by some measure of inhibition, self-regulation, or other self-control is central. Moderation
by gender, when the outcome measure is physical aggression, may be of limited interest with
respect to that theory given hormonal and cultural differences that render such contingencies
unsurprising.
Moderation studies may focus on boundary conditions—circumstances under which a given
theory cannot be expected to hold (Reynolds, 2007, as noted above, regards this as one of the fun-
damental questions in theory development). A theory regarding how framing may influence views
of a candidate may apply to those who do not have strong and committed identification with that
candidate. More trivially, it is unlikely to influence persons who by reasons of language or disabil-
ity cannot process the message or by virtue of residence or citizenship have no concern about the
election outcome. Health messages intended to evoke fear about a health threat may be effective
only with populations who have the health care access, skills, self-efficacy, or financial resources
226 M. D. SLATER AND L. S. GLEASON

available to address that threat (Witte, 1992). Boundary conditions may also examine population
questions: Does an effect hold when a study population is culturally distinctive, at a different
stage developmentally, or facing a different disease condition? Testing boundary conditions can
be an important component of theory development when the boundaries tested are predicted by
the theory, or should be. However, sometimes the reason for assessing boundary conditions is
more applied. If there are important social implications but no important theoretical implications
can be identified and articulated, such research probably will be of interest to a specialty journal
rather than a general interest journal.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Strategy 4.2: Moderated mediation

There is increasing awareness that the question of contingent conditions can be much richer
than simply identifying bivariate causal relations that are stronger or weaker under differing
circumstances. Causal process theories may work differently for different populations or under
different circumstances. Different mechanisms may hold. Different mediating relationships may
be found. For example, using our previous example, effects of violent media content might be
mediated by aggressive fantasy primarily for persons high in some personality construct such as
a tendency to become absorbed in narrative, or for those who report higher than average levels
of other stressors in their life. For others, the mediation relationship may not hold. This would
be theorizing about moderated mediation, which is amenable to statistical test (Preacher, Rucker,
& Hayes, 2007). Such models may be of particular importance in the study of causal process, as
they assert the process itself varies in theoretically predictable ways.

THEORY COMPARISON, THEORY DEVELOPMENT, AND THEORY SYNTHESIS

Strategy 5: Theory Comparison

Sometimes the focus of a study is specifically to compare the utility of theories in a given domain.
In general, this kind of comparison is most convincing or powerful when different theories result
in opposed predictions and alternative falsifiable hypotheses can be compared, as pointed out by
Popper (1959). For example, Lee (2008) endeavored to determine whether resource depletion
or reduced attention offers a better explanation of how computer modality moderates the extent
to which people evince gender stereotypes toward computers by deriving alternative hypotheses
based on each explanation.
Some research may also attempt to compare theories in terms of explanatory power, not com-
peting predictions. However, such comparisons are often difficult to interpret, as explanatory
power can be confounded with quality of measurement and other artifacts (Cooper & Richardson,
1986). Another potentially useful approach is to employ structural equation models incorporat-
ing latent variables to correct for measurement reliability. Such models can be used to assess the
relations predicted by alternate theoretical arguments and identify which is the better fit to the
data. For instance, Cho and Boster (2008) found their proposed model to explain the relationship
between perceived message quality, perceived effect on self, and various perceived message char-
acteristics was not well supported by a measure of model fit, and they developed an alternative
model that better explained relationships found. While this was not an a priori test of competing
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 227

models derived from theory (as in the case of Eveland et al., 2005), it does illustrate the utility of
such an approach in refining causal process theories.

Strategy 6: Synthesis among Theories: Creating Overarching Models or Theoretical


Frameworks

Sometimes the numbers of theories that can be applied to any given kind of problem can be
overwhelming, and their relation to one another can be hard to sort out. Some larger questions
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

may not lend themselves to precise theoretical statement in terms of formal hypotheses because
the exact nature of the hypotheses might vary depending on specific domains, each with its own
boundary conditions and contingencies. In such cases, development of overarching models or
theoretical frameworks can have considerable heuristic and theory-building value (see Shoemaker
et al., 2004, for a discussion of such models).
A prototypical example in the domain of health behavior is the Transtheoretical or Stages of
Change model (Prochaska, DiClemente, & Norcross, 1992), which organizes a variety of theories
of behavior change and intervention in terms of the readiness of an individual to address addictive
behaviors. This model has been adapted to integrate multiple theories of media communication
interventions in health behavior (Slater, 1999) as well as applied to a great many specific health
behaviors (perhaps inappropriately in some cases, as if it were a theory itself rather than an
organizing structure for selection of relevant theories). Such models typically cannot be tested via
a single study; their value is better assessed by the impetus they provide for theory development
and application of theories.

Strategy 7: Proposing a New Theory or Model in Communication

Sometimes existing theories are inadequate to address a social phenomenon and an entirely new
theoretical approach is justified. Such successful ventures are rare, though of substantial potential
impact when brought to successful fruition. Another approach well-suited to the communication
discipline’s interdisciplinary roots is to import theory from other disciplines and adapting such
theories into a communication context. Communication has been described as a “variable field”
(Paisley, 1984). That is, unlike “level-field” disciplines concerned with a single level of analysis
(such as intra-personal process for psychology and social processes for sociology), communi-
cation examines communication variables across a variety of levels of analysis. As a result, it
is likely to benefit from methodological and theoretical innovation from researchers working in
those different levels (intrapersonal or psychological, group/organizational, and sociological).
The communication contribution may be highly creative; often ideas developed at one level (e.g.,
psychology) do not really tackle the complex problems of real-world messages and communi-
cation contexts, so adapting these theories also causes them to become richer, better developed,
and more valuable. An example is the use of social categorization theory to explain mechanisms
behind some well-established communication phenomena (Hogg & Reid, 2006), or the devel-
opment of the extended parallel process model (EPPM) to synthesize two psychological models
of fear appeals (the parallel process model, Leventhal, 1971, and protection motivation theory,
Rogers, 1975), and to develop methods to apply them to a variety of risk communication contexts
(Witte, 1992).
228 M. D. SLATER AND L. S. GLEASON

Strategy 8: Description, Observation, Hypothesis Generation

Most of the approaches to contributing to theory described above consist of elaborating exist-
ing theory, explaining mechanisms behind previously theorized processes, comparing theories,
changing assumptions of theories, and importing theories from other disciplines. New theoret-
ical innovations, however, often come from observation of social phenomena. Sometimes such
observation is informal. A researcher, based on experience, recognizes that existing theory is
inadequate for describing what one has seen first-hand, experientially or as an applied researcher,
and embarks on research intended to expand theoretical explanation to address the apparent gap.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Sometimes such observation is formal, rigorous, and publishable.

Strategy 8a: Content analysis

A typical research strategy that occasionally tests or develops theory, but more often describes
communication phenomena about which one may theorize social or psychological processes and
effects, is content analysis. Content analysis is typically interesting insofar as content discover-
ies suggest socially important effects and underlying mechanisms; cultivation theory (Gerbner,
Gross, Morgan, Signorielli, & Shanahan, 2002) is a paradigmatic example of theory which has its
roots in content analysis. Much of the scientific interest of content analysis, then, derives from the
extent to which findings suggest important theoretical as well as substantive possibilities for sub-
sequent research (McLeod & Reeves, 1980). For instance, Coyne and Whitehead (2008) exam-
ined indirect or social aggression (e.g., spreading rumors, social exclusion) in animated Disney
films. Beyond the substantive interest inherent in such a descriptive analysis, such work provides
a potential basis for developing theory about the nature of indirect aggression and the influence
of such portrayals on the beliefs and behaviors of children, including possible gender differences.

Strategy 8b: Qualitative observation

In our view, the value of qualitative studies often derives from the theory generation possibil-
ities inherent in rigorous qualitative observation (see Silverman, 2004). For example, Howard
(2008) analyzed more than 15 hours of pilot–air traffic controller dialogue to better under-
stand “problematic communication.” It is regrettable that the progression from observation to
theory/hypothesis generation to quantitative data collection and test, which characterizes much
of physical science, is often not evident in communication or other social sciences, and qualita-
tive and quantitative research approaches are sometimes and unnecessarily treated as competing
rather than complementary strategies with which to develop theory. Theory—the development of
broad, generalizable explanatory principles and mechanisms—often originates from close obser-
vation of phenomena. Qualitative research provides systematic and disciplined means with which
to conduct such observation.

Strategy 8c: Documenting a communication effect of substantive and potential


theoretical interest

In mass communication research, a classic research type is the pure effects study (McLeod
& Reeves, 1980) intended to establish that media use has an impact on attitudes and behavior
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 229

and perhaps to identify a parameter estimate for the size of the effect. One example is a study
of the long-term effects of a single sun safety campaign on employees of 26 ski areas in North
America (Anderson et al., 2008). From a theory contribution point of view, such research can be
important for two reasons. First, sometimes an effect is sought because it is predicted based on an
existing theory or theories and provides a test and extension of that theory. Sometimes, the effect
is primarily of substantive interest, but once established, explanation of underlying mechanisms
and explanations becomes fertile ground for theory development, extension, and test.

Strategy 9: Overview of Theory: Review or Meta-analysis Articles


Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Another way to advance theory is to intellectually organize theoretical work on a given topic.
This is typically done either through reviews of literature and theory in a given domain or a
more empirical examination using meta-analysis as a tool. For example, Sun, Pan, and Shen
(2008) meta-analyzed 106 studies on the third-person effect. The theoretical utility of such an
effort is not only to confirm the robustness of the effect across multiple studies and methods,
but to identify moderators and boundary conditions—such as perceived social desirability of the
influence, vulnerability of others, and others’ similarity to self—found in the analysis of the
pooled data. Ideally, such work provides a clear picture of the empirical state of knowledge with
respect to the conditions under which the predictions of the theory are most likely to hold.
The technical issues and concerns regarding meta-analyses are beyond the scope of this
discussion (e.g., see Johnson, Scott-Sheldon, Snyder, Noar, & Huedo-Medina, 2008). From a
perspective of concern with theory, a potential problem with meta-analyses is that they may focus
on evidence for a given association or relationship and may not look at evidence for moderation
or indirect effects (either because these were not of interest to the meta-analyst, or there were
insufficient studies to permit their analysis). In such cases, a meta-analysis may provide only a
pixelated image of the state of knowledge on a given topic.

A CONTENT ANALYSIS OF A YEAR IN MAJOR GENERAL INTEREST


COMMUNICATION JOURNALS

We conducted a content analysis of the empirical research articles published in three major
general-interest publications over the course of one year (the most recent year at the time that
this project was originally undertaken). Our primary objective was to use our framework to
characterize the use of these strategies in recent articles in major general interest journals in
communication, to gain at least an initial sense of their relative frequency, albeit in a snapshot
of a single year of work in the field. For these purposes, aiming at a census of the most recent
research in key journals at the time of data collection seemed an appropriate strategy. An ancil-
lary goal was to determine whether two readers could reliably distinguish between articles that
differed with respect to the approaches used and outlined above.
We wished to use journals that serve the discipline as a whole and that focus on the publi-
cation primarily of quantitative, empirical research. A bibliometric analysis of communication
journals (Feeley, 2008) found that, averaged across the years 2002–2005, the highest impact fac-
tors were associated with Human Communication Research (HCR), Communication Research
(CR), Journal of Communication (JOC), and Personal Relationships. Personal Relationships was
excluded as a specialty journal focused on a subdisciplinary area, and the other three journals were
230 M. D. SLATER AND L. S. GLEASON

clearly central to citation networks in the discipline. More recent citation reports continue to list
HCR, CR, and JOC as the highest-impact general interest communication journals. Moreover, the
editorial statements of each of the selected journals emphasize the importance of theory contri-
bution to publication. As our purpose is to characterize strategies for contributing to knowledge
and to theory in contemporary empirical communication research practice, our purpose is served
by analyzing articles from these journals. Therefore, our content analysis focused on HCR, CR,
and JOC.
As we are interested in a snapshot of contemporary research approaches rather than character-
izing the history of the field, we content-analyzed four issues each of one year of these journals
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

for a total of 12 issues and 89 articles (between spring 2008 and spring 2009 using the most
recent available journal issues at the time the content analysis was conducted). A special review
issue of JOC was excluded from the analysis because it did not focus on empirical studies. The
most recent four of six issues that were available in hard copy to the authors at the time of coding
were selected from journals that published six times per year so that each of the three journals
had equal numbers of issues in the sample.
We used the categories described earlier (see Table 1 for a summary). We emphasize that the
various strategies are not mutually exclusive. Therefore each strategy was assessed simply as
being present or absent for each article coded. The strategies were first developed a priori based
on our experience reading, reviewing, and writing research mss. They were then modified and
refined during the process of developing intercoder reliability.
We assessed 50% of the sample for reliability using Krippendorff’s alpha (Hayes &
Krippendorff, 2007) with two coders. Subcategories within the major categories were typically
too difficult to code reliably, though they provided a useful basis for discussion of author intent.
Therefore, all articles were double-coded and discrepancies resolved after discussion for both the
major categories and the sub-categories for the purpose of generating percentages for the table
presented here.
Table 1 provides the reliability assessments for the major categories, as well as a summary
of strategies and sub-strategies that we also used as a reference to our coding definitions on our
coding sheet. We also provide Table 1 as a tool for researchers or graduate students who wish a
quick reference to the list of strategies and their component elements, for use when assessing their
own work or that of others. Krippendorff’s alphas ranged from 0.63 to 1.00, indicating that the
major strategy distinctions can be independently coded with adequate reliability but still reflective
of the degree of subjective judgment involved in making such category assessments.
The use of articles from a single year, the several years that have passed since time of data col-
lection, and use of just three journals limit the robustness of results with respect to characterizing
such strategies as implemented in the communication discipline. With that caveat, we provide a
brief overview of the results of the content analysis (see Table 2 for a summary of frequencies for
each strategy by journal).
Consistent with McLeod and Reeves (1980) dictum about the importance of contingent con-
ditions in communication, assessment of moderation was the most frequently used strategy,
employed in 54% of the articles analyzed, followed by elaboration of theory (40%, mostly by
testing in new contexts) and then by tests of mediating explanatory mechanisms (22%). A few of
the tests of mediating mechanism, even in these major journals, did not employ the more rigorous
techniques for assessing mediation hypotheses (e.g., three provided component paths only with-
out assessing indirect effects statistically or even showing attenuation of a main effect), though
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 231

TABLE 1
Coding Definitions by Major Theory Development Strategy (with Krippendorff’s Alpha) and Subcategories

Strategy Coding Definition Alpha

1 Theory Revision: Refining key concepts and assumptions .63


1a When a concept is redefined and alternative operationalization(s) is/are introduced to reflect those
conceptual refinements
1b When a key theoretical inference or assumption previously untested or inadequately tested is
operationalized and tested
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

1c When fundamental operational problems with existing work are pointed out that have significant
conceptual implications
2 Extending Theory Range: Adding variables, context, populations .68
2a When a new, conceptually distinct independent variable (not mediator or moderator) is added to a
program or line of research previously established, except as identified in 1d.
2b When an established theory or effect is tested using a new communication channel, genre, or
class/category of message(s).
2c When an established theory or effect is tested using a new subject population or in a new geographic
location, or when such theory or effect is tested using a new, conceptually distinct dependent
variable.
3 Elucidating Mechanism: Theorizing causal process by testing one or more mediating variables .81
3.1 By testing component paths only (pseudo-test of mediation)
3.2 Using SEM with fit statistic for single model or models
3.3 Using Baron & Kenny steps (report of main relationship w/ and w/o mediator and new beta or % or
attenuation; including going from S to N/S)
3.4 Using Holmbeck procedure (Baron & Kenny adopted to SEM)
3.5 Using SEM with alternative models (express fit comparison)
3.6 Using Sobel or equivalent statistical test of indirect effect not bootstrapped
3.7 Using bootstrapped test for indirect paths
3.8 Using panel data to include time order in assessing mediation
3.9 Experimental manipulation of proposed mechanism
4 Elucidating Contingent Conditions: When an established “effect” (relationship) is theorized as .82
being non-uniform across people, situations, or other contexts (i.e., moderation or interaction
effects).
4.1 Unmediated moderation (interaction hypothesized and tested)
4.2 When moderated mediation or mediated moderation is hypothesized; i.e., mediation “occurs” on a
non-uniform basis across people, situation, or other contexts or when “impact” of moderator on
established “effect” (relationship) is mediated by another variable (mediator).
5 Theory Comparison: When alternative theories resulting in distinctly different predictions are n/a
compared in relation their ability to explain observed phenomena, predictive power, or
variance explained.
6 Synthesis across Theories: When an overarching model and/or theoretical framework is n/a
offered to explain and/or organize a variety of related but distinct existing theories.
7 Theory Creation: When an entirely new theoretical approach is offered (not including theory n/a
elaboration or revision as described above is relabeled as a new theory).
8 Description, Observation, or Hypothesis Generation
8a Quantitative content analysis primarily descriptive or hypothesis-generating in nature .63
8b Rigorous qualitative observation (normally hypothesis-generating, occasionally .85
hypothesis-testing without statistical inference) .
8c When an effect primarily of primarily of substantive interest is identified/described but not as n/a
part of theory development
9 Overview of Theory: Review or meta-analysis articles 1.00

Note. Major categories assessed for intercoder reliability (Krippendorff’s alpha) in bold. Not applicable (N/A) when
no examples were found in reliability sample by either coder.
232 M. D. SLATER AND L. S. GLEASON

TABLE 2
Strategies for Contributing to Knowledge by Sub-category and by Journal Sampled (Four Issues Each,
Spanning 2008–2009)

JOC HCR CR Total


Strategy Abbreviated Coding Definition (N = 38)b (N = 27) (N = 24) (N = 89)

1 Theory Revision: 2 3 6 11
1a a concept is redefined 0 1 2 3
1b a key untested assumption tested 1 0 3 4
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

1c fundamental operational problems 1 2 2 5


2 Extending theory range: Adding variables, context, 10 11 15 36
populations:∗
2a independent variable is added 2 3 4 9
2b theory tested using a new message channel, genre, or class 6 4 7 17
2c tested using a new population, location, or dependent variable 3 4 5 12
3 Elucidating Mechanism: 5 8 7 20
3.1 component paths only 2 1 0 3
3.2 SEM with fit statistic for single model 2 3 2 7
3.3 using Baron & Kenny steps 1 4 1 6
3.4 using Holmbeck procedure 0 0 1 1
3.5 SEM with alternative models 0 1 1 2
3.6 Sobel or equivalent not bootstrapped 0 3 3 6
3.7 bootstrapped test for indirect paths 1 2 3 6
3.8 panel data to address time order in mediation 1 0 0 1
3.9 manipulating proposed mediatora 0 0 0 0
4 Elucidating Contingent Conditions:∗ 14 18 16 48
4.1 moderation only∗ 13 17 14 44
4.2 moderation and mediation 2 4 3 9
5 Theory Comparison: 3 3 0 6
6 Synthesis across Theories:a 0 0 0 0
7 Theory Creation: 0 0 1 1
8 Description, Observation: 6 3 1 10
8a content analysis 4 0 1 5
8b qualitative 3 3 0 6
8c effect identified/described only 1 0 0 1
9 Review or meta-analysis articles:∗ 9 3 0 12

Note. Each article is coded with respect to presence or absence of each strategy and substrategy listed. Each cell,
then, represents how many of the articles coded in a given journal used that strategy. An article can be coded as using
multiple strategies and substrategies. Therefore, summing across substrategies does not add up to the number reported
for the larger category; e.g., an article may assess mediation several ways but use of mediation tests is counted only once.
All codes reported were double-checked by a second coder and differences of opinion resolved through discussion.
a Exemplars of these categories were not identified in the set of articles tested for reliability.
b While the same number of journal issues (4) was studied for each journal, the number of articles per journal varied.

Chi-square tests were used to assess significant differences between counts of utilization of the strategy amongst journals
for any given category or subcategory; significant differences are asterisked.
∗ p < .05.

most did. Attempts to propose altogether new theories or frameworks were rare or nonexistent
with only one example appearing in this set of articles, unsurprising given the rarity of successful
such efforts and the likelihood that only relatively mature such efforts would find their way to the
pages of leading empirically-oriented journals.
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 233

Differences between journals were small in the year under study. CR appeared to do more
theory elaboration, (χ 2 = 7.99, df = 2, p < .05) and at marginal levels, theory revision, (χ 2 =
5.35, df = 2, p = .07). HCR reported more moderation tests (χ 2 = 7.80, df = 2, p < .05), perhaps
because of what appeared to be frequent use of experimental designs which focus on interaction
hypotheses. JOC published the most review and meta-analytic articles (even with the exclusion
of the all-review issue), and CR the least (χ 2 = 7.26, df = 2, p < .05). Given how minor these
differences are, it seems that approaches to contributing to knowledge and theory are reasonably
consistent across these three major quantitatively oriented communication journals.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

CONCLUSION

The purpose of this effort is to provide a framework for identifying and describing the theory
contribution of empirical research studies. The empirical portion of this study is based on a
single year of research in three major journals—and, for the sake of consistency, four issues
per journal. The patterns of findings are likely to be different for specialty journals that may
emphasize substantive import or particular research methods. The nature of theoretical contribu-
tions proposed are also likely to vary in least in some regards over time as a function of chance, of
intellectual fashion, and methodological advances. For example, it is likely that studies addressing
mediation are becoming more common now that the software to conduct appropriate statistical
tests is readily available and knowledge of how to conduct such analyses is likely increasing over
time. Greater attention to properly conceptualizing, hypothesizing, and testing ideas about causal
process (mediation and moderated mediation) would in our view represent significant progress in
researchers’ efforts to elucidate causal processes in communication.
We would also like to make a qualitative observation. That is, some authors are very skilled
at clearly articulating the nature of their intended theoretical contribution. Others give this less
emphasis, and often a painfully careful reading became necessary to make various coding deci-
sions. For example, sometimes it seemed researchers were concerned with mediation, but the
nature of the mediating relationships was often not clearly spelled out. Sometimes moderated
mediation was suggested but not explicitly hypothesized or described. Sometimes the way the
study built upon previous work, and the extent of conceptual refinement and elaboration involved,
had to be painstakingly inferred, occasionally with reference to the authors’ own knowledge of
the literature apart from that cited in the article itself (one typical source of differences between
coders). The process certainly enhanced our appreciation of a clear and thoughtful statement
of theoretical contributions intended, and theorized relationships to be tested, in a research
manuscript. We believe that greater efforts to think through and express precisely what a given
study contributes to generalizable knowledge and previous theory would do much to increase the
quality of published communication research as a whole.
It is our hope that the present work will be of use to others in facilitating clearer articulation
of the nature of the theoretical contribution, as it has been for us. As another qualitative obser-
vation, we note that this framework has proven useful in facilitating our own manuscript writing
and reviewing as well as in helping graduate students learn how to recognize and make original
contributions to theory, by asking them to apply it to research that they read and that they pro-
pose. (We have found this works well in the second semester or third quarter for masters or new
doctoral students; we suspect most students would have insufficient background to understand
234 M. D. SLATER AND L. S. GLEASON

the issues discussed here in their first semester of graduate school.) We hope, then, that this effort
may strengthen the communication literature, by providing scholars with a tool to sharpen their
thinking about their approaches to generalizable knowledge and theoretical contribution, and its
articulation in manuscript form.

REFERENCES

Anderson, P., Buller, D., Voeks, J., Walkosz, B., Scott, M., Cutter, G., & Dignan, M. (2008). Testing the long-term effects
of the Go Sun Smart worksite health communication campaign: A group-randomized experimental study. Journal of
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Communication, 58, 447–471.


Appel, M. (2008). Fictional narratives cultivate just world beliefs. Journal of Communication, 58, 62–83.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research:
Conceptual, strategic, and statistical considerations. Journal of Personal Social Psychology, 51(6), 1173–1182.
Berger, C. R, Roloff, M. E., & Roskos-Ewoldsen, D. R. (2009). What is communication science? In C. R. Berger, M. E.
Roloff, & D. R. Roskos-Ewoldsen (Eds.), Handbook of communication science (2nd ed., pp. 3–20). Thousand Oaks,
CA: Sage.
Bryant, J., & Miron, D. (2004). Theory and research in mass communication. Journal of Communication, 54(4), 662–704.
Chaffee, S. (1991). Explication. Thousand Oaks, CA: Sage.
Cheong, J., MacKinnon, D. P., and Khoo, S. T. (2003). Investigation of mediational processes using parallel process latent
growth modeling. Structural Equation Modeling, 10, 238–262.
Cho, J. (2008). Political ads and citizen communication. Communication Research, 35, 423–451.
Cho, J., & Boster, F. J. (2008). First and third person perceptions on anti-drug ads among adolescents. Communication
Research, 35, 169–189.
Clark, R., MacGeorge, E., & Robinson, L. (2008). Evaluation of peer comforting strategies by children and adolescents.
Human Communication Research, 34, 319–345.
Cooper, W. H., & Richardson, A. J. (1986). Unfair comparisons. Journal of Applied Psychology, 71, 179–184.
Coyne, S. M., & Whitehead, E. (2008). Indirect aggression in animated Disney films. Journal of Communication, 58,
382–395.
Dixon, T. (2008). Network news and racial beliefs: Exploring the connection between national television news exposure
and stereotypical perceptions of African Americans. Journal of Communication, 58, 321–337.
Eveland, W., Hayes, A., Shah, D., & Kwak, N. (2005). Understanding the relationship between communication and
political knowledge: A model comparison approach using panel data. Political Communication, 22, 423–446.
Feeley, T. (2008). A bibliometric analysis of communication journals from 2002–2005. Human Communication Research,
34, 505–520.
Gerbner, G., Gross, L., Morgan, M., Signorielli, N., & Shanahan, J. (2002). Growing up with television: Cultivation
processes. In J. Bryant & D. Zillman (Eds.), Media effects: Advances in theory and research (pp. 43–68). Mahwah,
NJ: Lawrence Erlbaum.
Greenwald, A. G. (2012). There is nothing so theoretical as a good method. Perspectives on Psychological Science, 7,
99–108.
Hayes, A. F. (2009). Beyond Baron and Kenny: Mediation analysis in the new millennium. Communication Monographs,
76, 408–420.
Hayes, A. F., & Krippendorff, K. (2007). Answering the call for a standard reliability measure for coding data.
Communication Methods and Measures, 1, 77–89.
Hayes, A. F., Matthes, J., & Eveland, W. P. (In press). Stimulating the quasi-statistical organ: Fear of social isolation
motivates the quest for knowledge of the opinion climate. Communication Research.
Hogg, M., & Reid, S. (2006). Social identity, self-categorization, and the communication of group norms. Communication
Theory, 16, 7–30.
Holbert, R., & Stephenson, M. (2008). Commentary on the uses and misuses of structural equation modeling in com-
munication research. In A. Hayes, M. Slater, & L. Snyder (Eds.), The Sage sourcebook on advanced data analysis
methods for communication research (pp. 185–218). Thousand Oaks, CA: Sage.
Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical clarity in the study of mediators and mod-
erators: Examples from the child-clinical and pediatric psychology literatures. Journal of Consulting and Clinical
Psychology, 65, 599–610.
CONTRIBUTING TO THEORY IN COMMUNICATION SCIENCE 235

Howard, J. (2008). “Tower, am I cleared to land?”: Problematic communication in aviation discourse. Human
Communication Research, 34, 370–391.
Huesmann, L., & Taylor, L. (2006). The role of media violence in violent behavior. Annual Review of Public Health, 27,
393–415.
Jaccard, J., & Jacoby, J. (2010). Theory construction and model-building skills: A practical guide for social scientists.
New York, NY: Guilford.
Johnson, B. T., Scott-Sheldon, L. A., Snyder, L. B., Noar, S. M., & Medina-Huedo, T. B. (2008). Contemporary
approaches to meta-analysis in communication research. In A. F. Hayes, M. D. Slater, & L. B. Snyder (Eds.), The
Sage sourcebook of advanced data analysis methods for communication research (pp. 311–348). Thousand Oaks,
CA: Sage.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

Kline, S., Simunich, B., & Weber, H. (2008). Understanding the effects of nonstraightforward communication in
organizational discourse: The case of equivocal messages and corporate identity. Communication Research, 35,
770–791.
Lee, E. (2008). Gender stereotyping of computers: Resource depletion or reduced attention? Journal of Communication,
58, 301–320.
Leventhal, H. (1971). Fear appeals and persuasion: The differentiation of a motivational construct. American Journal of
Public Health, 61, 1205–1224.
MacCallum, R. C., Wegener, D. T., Uchino, B. N., & Fabrigar, L. R. (1993). The problem of equivalent models in
applications of covariance structure analysis. Psychological Bulletin, 114, 185–199.
MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. Mahwah, NJ: Lawrence Erlbaum.
McLeod, J., & Reeves, B. (1980). On the nature of mass media effects. In S. Withey & R. Abeles (Eds.), Television and
social behavior: Beyond violence and children (pp. 17–54). Mahwah, NJ: Erlbaum.
Oliver, M. B., Yang, H., Ramasubramanian, S., Kim, J., & Lee, S. (2008). Exploring implications of perceived media
reinforcement on third-person perceptions. Communication Research, 35, 745–769.
Paisley, W. J. (1984). Communication in the communication sciences. In B. Dervin & M. J. Voigt (Eds.), Progress in
communication sciences (Vol. 5, pp. 1–44). Norwood, NJ: Ablex.
Pavitt, C. (2009). Alternative approaches to theorizing in communication. In C. R. Berger, M. E. Roloff, & D. R. Roskos-
Ewoldsen (Eds.), Handbook of communication science (2nd ed., pp. 37–54). Thousand Oaks, CA: Sage.
Pavitt, C. (2000). Answering questions requesting scientific explanations for communication. Communication Theory,
10, 379–404.
Popper, K. (1959). The Logic of Scientific Discovery. London: Hutchinson.
Preacher, K., Rucker, D., & Hayes, A. (2007). Assessing moderated mediation hypotheses: Theory, method, prescriptions.
Multivariate Behavior Research, 42, 185–227.
Preacher, K. J., & Hayes, A. F. (2008). Contemporary approaches to assessing mediation in communication research.
In A. F. Hayes, M. D. Slater, & L. B. Snyder (Eds.), The Sage sourcebook of advanced data analysis methods for
communication research (pp. 13–54). Thousand Oaks, CA: Sage.
Prochaska, J. O., DiClemente, C. C., & Norcross, J. C. (1992). In search of how people change: Applications to addictive
behaviors. American Psychologist, 47, 1102–1114.
Reynolds, P. D. (2007). A primer in theory construction. Boston, MA: Allyn and Bacon.
Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. The Journal of Psychology, 91,
93–114.
Shoemaker, P., Tankard, J., & Lasorsa, D. (2004). How to build social theories. Thousand Oaks, CA: Sage.
Silverman, D. (2004). Qualitative research: Theory, method, and practice. Thousand Oaks, CA: Sage.
Slater, M. D. (1999). Integrating application of media effects, persuasion and behavior change theories to communication
campaigns: A stages of change framework. Health Communication, 11, 335–354.
Sobel, M. (1988). Direct and indirect effects in linear structural equation models. In J. Long (Ed.), Common
problems/proper solutions: Avoiding error in quantitative research (pp. 46–64). Thousand Oaks, CA: Sage.
Spencer, S. J., Zanna, M. P., & Fong, G. T. (2005). Establishing a causal chain: Why experiments are often more effective
than mediational analyses in examining psychological processes. Journal of Personality and Social Psychology, 89,
845–851.
Stinchcombe, A. (1968). Constructing social theories. New York, NY: Harcourt Brace.
Sun, Y., Pan, Z., & Shen, L. (2008). Understanding the third-person perception: Evidence from a meta-analysis. Journal
of Communication, 58, 280–300.
236 M. D. SLATER AND L. S. GLEASON

Valkenburg, P., & Peter, J. (2008). Adolescents’ identity experiments on the Internet: Consequences for social competence
and self-concept unity. Communication Research, 35, 208–231.
Vangelisti, A. L. (2009). Challenges in conceptualizing social support. Journal of Social and Personal Relationships, 26,
39–51.
Witte, K. (1992, December). Putting fear back into fear appeals: The extended parallel process model. Communication
Monographs, 59, 329–349.
Wojcieszak, M. E., & Mutz, D. C. (2009). Online discussion groups and political discourse: Do online discussion spaces
facilitate exposure to political disagreement? Communication Research, 59, 40–56.
Word, C. O., Zanna, M. P., & Cooper, J. (1974). The nonverbal mediation of self-fulfilling prophecies in interracial
interaction. Journal of Experimental Social Psychology, 10, 109–120.
Downloaded by [Flinders University of South Australia] at 01:48 30 January 2015

You might also like