Download multilevel analysis in mass communication research

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
EBSCOhost Full Display
Page 1 of 24
Result 3 of 3
Title:
Subject(s):
Source:
Author(s):
Abstract:
[Go To Full Text] [Tips]
Multilevel analysis in mass communication research.
ANALYSIS (Philosophy); CONTENT analysis (Communication) -- Evaluation
Communication Research, Apr91, Vol. 18 Issue 2, p140, 34p, 1 diagram
Pan, Zhongdang; McLeod, Jack M.
Presents an epistemological view of levels of analysis in mass communication
research. Reason for developing a multiple perspective of the field; Three factors to
consider on the demarcation of levels of analysis; Four types of relationships and a
metatheoretical framework; Five dimensions that locate many theoretical propositions
about macro and micro linkages; Applications of the multilevel framework.
9707265671
0093-6502
12668
Academic Search Elite
AN:
ISSN:
Full Text Word Count:
Database:
Print: f
g Click here to mark for print.
c
d
e
View Item: XML Full Text
View Links: Check MadCat for electronic or print copies
[Go To Citation]
Select Language
MULTILEVEL ANALYSIS IN MASS COMMUNICATION
RESEARCH
This article presents an epistemological view of levels of analysis. According to this view, four types of
relationships need to be differentiated: macro-macro' macro-micro, micro-micro, and micro-macro. The two
within-level relationships are linked by the two cross-level relationships that, in turn, are explicated by various
theories of organizational, institutional, and social processes. Mass communication is thus conceived of as a
process from production to consumption that occurs at both micro-individual and macro-social levels. The
contributions of this multilevel view of mass communications to theoretical development in the field is illustrated
by analyzing three prominent theories in our field: the knowledge gap, cultivation, and the spiral of silence, finally,
the article discusses the available research techniques and strategies for dealing with multilevel research
questions.
This special issue of Communication Research testifies that communication researchers are determined to
confront difficult levels-of-analysis issues. They are not alone. Rapid growth in the literature specifically
addressing these issues can be found in very diverse areas of social science.(n1) In mass communication research,
we saw similar efforts in Cushman and Craig's (1976) system analysis, in Dimmick and Coit's (1982) analysis of
mass media decision making, and more recently in the organizational structure of the Handbook of
Communication Science (Berger & Chaffee, 1987).
The basis for developing a multilevel perspective of this field is that mass communication consists of persons in
bureaucratic organizations producing diverse messages for multiple consumers with impact on individuals,
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 2 of 24
families, and social systems. The very nature of mass communication processes compels multilevel consideration.
Despite its being a "variable" rather than a "level" field (Paisley, 1984), however, mass communication research
lacks any semblance of cross-level integration. Researchers from different intellectual backgrounds tend to work
within a particular level, using the concepts and methods most common within that level, and rarely venture
beyond their level of "residence." Practical considerations have restrained their theoretical and empirical endeavors
from going either "vertically" across different levels or "horizontally" to consider both production and
consumption. The result has been theoretical "islands" mapped by distinctive terminology, each surrounded by
mutually noncomparable "seas" of research evidence. Development of the field needs not only a broader
horizontal spectrum to consider mass communication as a process from production to consumption but also
vertical multilevel perspectives to link various theoretical "maps."
Other symptoms also indicate a lack of analytical multilevel mass communication perspectives. The literature of
the field has tended to produce "metaconcepts," such as free flow of information, New International Information
Order, agenda setting, knowledge gap, and so on. These are concept conglomerates, each of which refers to a
domain of phenomena rather than to a specific set of concepts and their connections. They are often ambiguous as
to levels in that they do not specify or even imply translation across levels. When these metaconcepts gain status as
legitimate labels, they become reified, and special effort is needed to recover the concepts that they subsume. Even
more difficult is their "decomposition" at various levels of abstraction.(n2)
We begin here by proposing an alternative way of treating levels-of-analysis issues. We then present a
metatheoretical framework for multilevel analyses. This framework is illustrated by analyzing three prominent
theories (or pretheories) in our field: the knowledge gap (Tichenor, Donohue, & Olien, 1970; Donohue, Tichenor,
& Olien, 1975); cultivation (Gerbner & Cross, 1976); and the spiral of silence (Noelle-Neumann, 1984). Each of
these theories has stimulated research for more than a decade. Our analysis neither endorses them as exemplars
nor condemns them as hopelessly flawed. They serve as useful illustrations of both the problems and promises of
research in the field. Finally, we suggest various research options. Our central theme is that construction of mass
communication theories would benefit from multilevel perspectives and that intellectual and analytical apparatus
is available for multilevel analyses.
Levels of Analysis Reanalyzed
Despite previous efforts, ambiguity has frustrated meaningful discussion of levels of analysis in mass
communication. One source of confusion is the mingling of levels of discourse and levels of analysis (cf. Chaffee
& Berger, 1987, p. 144). Another is the lack of consensus as to whether levels are ordered as 8 continuous
hierarchy or as a set of discrete categories (cf. Eulau, 1986).
A discourse of scientific inquiry refers to the language employed in a scientific discipline for the purpose of
describing and explaining a finite set of phenomena. This language consists of concepts, statements, conceived
mechanisms' as well as assumptions about nature. In the discourse of empirical science, an ontological view of the
vertical ordering of scientific disciplines was developed from a hierarchical conception of the natural world
(Comte, 1830-1842/1983; see also Bunge, 1977; Paisley, 1984): physical, chemical, biological, psychological,
social, and 80 on. Each of these levels is seen as having emerged from other lower levels and as being governed by
laws of its own (Bunge, 1977). Scientific disciplines are thus assumed to form naturally in accordance with the
divisions of the ontological levels. Each of the levels differs in terms of degrees of generality, so that laws of
physics are more general than those of biology, which in turn, are more general than those in social sciences, if one
could find any laws in the social sciences.
Although the ontological view may call for multilevel analysis in "variable fields," it has problems in dealing with
multilevel issues in "level fields," such as sociology and anthropology. Additional difficulties for this ontological
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 3 of 24
view result from the historical evolution of science. For example, whereas Comte (1830-1842/1983) denied
psychology the status of a scientific discipline and classified research on human affective and intellectual faculties
as part of biology (pp. 182-194), psychology today is indisputably regarded as a distinct field and perhaps as being
a level field as well (Paisley, 1984).
The ontological view of hierarchical levels has created obstacles to empirical research. Not only is the assumption
about the hierarchy of nature limited by what is known at a particular historical period, but the ontological view of
natural hierarchy fails to provide solutions to empirical questions of cross-level linkages. Different discourses,
based on the prevailing ontological view of nature and under the influences of tradition, employ very different
terminologies that are not easily translated. Making cross-level understanding even more difficult is the fact that
scientific knowledge continues to evolve in the direction of increased division of disciplines and specialization of
researchers. Neither of these helps facilitate between-discourse communication.
Further, in the realm of ontology, disputes over the primacy of units at one heel in relation to units at another level
are empirically unresolvable. For example, Alexander and Giesen (1987, p. 14) list five different approaches to
micro and macro relations in social theories. One major dimension along which these theories vary is whether the
individual or society has ontological primacy.(n3) Intellectual discussion among the five schools of thought may
be stimulating and interesting, but the chicken-egg primacy issue is not among the problems capable of empirical
resolution.
An epistemological view of levels of analysis may be more fruitful. According to this view, to explain variations
among units in a realm of discourse, it is sometimes necessary to incorporate concepts whose variance is located
either among the aggregates of these units or among the components within each of the units. A level of analysis is
thus defined as a category of concepts whose variations or covariations are to be explained. The units are the
carriers of the attributes defined by the corresponding concepts (see Eulau, 1977, p. 46, for a similar definition).
In the epistemological approach, "the choice of the empirical level on which analysis is to be conducted is a
substantive rather than a methodological problem" (Eulau, 1977, p. 44), and such choice does not depend on
holding a particular position in an ontological debate. In both level and variable fields, substantive questions call
for clarification of concepts, which means identifying the locus of variation within a concept as well as among the
concepts, ant specification of theoretical as well as operational linkages of the concepts (Hage, 1972). The locus of
the variation and the nature of the mechanisms determine the appropriate level of analysis.
Some concrete examples may be helpful here. In the "gatekeeping" literature, the early studies examined the
question of how an individual editor selected news stories (White, 1950). Although the initial empirical study
observed only one wire service editor, the research question was such that the focus was on variations among
individual editors, in terms of their perceptions of news values, readers' interests, their own values, and so on.
Dimmick (1979) took over the concept of gatekeeping to examine organizational activities and uncertainty
reduction in the process of information control. In this context, variations in organizational goals, structure, and
conflict management were employed to account for differences in information control among organizations. It is
possible to move up one more level. Donohue, Tichenor, and Olien (1972) argue that mass media can be conceived
as "social subsystems" that serve both distribution and feedback control functions. In this context, media
information control is examined in terms of the relationships between media systems, professional and
occupational systems, media ownerships and community power structure, and so on. This macro-level conception
certainly requires analyses conducted at the macrosocial level. All these examples demonstrate that the locus of
variations to be explained within a theoretical perspective determines the specific level of analysis.
Historically, distinctive vocabularies have tended to become useful in analyzing particular types of units. For
example,the concepts used to analyze media organizations (e.g., size, structure) are quite distinct in form and
meaning from those used to analyze the media use behavior of individuals (e.g., exposure, attention, reliance). To
the extent that the concepts used in a particular realm of discourse are clearly defined, however, communication
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 4 of 24
about theory is possible. Mindless mixing of variables from different levels is of little value. Attempting to
"explain" individual level variations of television news effects, for example, by tossing "psychological,"
"sociological," and "media" variables into a step-wise regression is likely to result in an alchemic stew contributing
essentially nothing to coherent theory construction. The epistemological view of levels of analysis thus calls for a
clear definition of levels (Reeves, 1989) as well as for greater attention to cross-level mechanisms and processes
(McLeod & Pan, 1989).
This epistemological view of levels of analysis also holds that analytical levels fall on a continuum of micro and
macro (Eulau, 1986; 1977), and the terms of micro level and macro level are meaningful only in specified
theoretical domains. As Eulau (1986) says, "what in this continuum is micro and what is macro depends on the
point on the micro-macro scale where the observer `dips in,' where he fixes his object unit of analysis." Further,
"what is micro and what is macro . . . changes with changing theoretical standpoints" (Eulau, 1986, p. 90).(n4)
The continuum conception of levels of analysis can be seen as a partial solution to the particular difficulties that
mass communication researchers face in specifying units of analysis. These difficulties are more daunting, for
example, than comparable specification of units by sociologists analyzing educative processes or political
scientists studying voting behavior. It is intuitively meaningful to discuss school effects or classroom effects on
students' learning (e.g., Hauser, Sewell, & Alwin, 1976; Jencks & Brown, 1975; Gamoran, 1986). For political
scientists, there is a clear progression of voting units from individual voters to precincts, counties, states, and so on
For communication researchers, there are no such "natural units," except for production organizations (Dimmick
& Coit, 1982; Whitney, 1982). The continuum view of levels sensitizes us to conceptualize the appropriate levels
of analysis.
Viewing levels as a continuum also implies that there are multiple ways of conceptualizing levels of analysis. It
does not necessarily contradict the notion of hierarchical structure of the levels. There are various levels-ofanalysis classifications in mass communication research; for example, the "atom-molecule-cell" model of
hierarchy of nature (Paisley, 1984), of four distinct forms of communication (Chaffee & Berger, 1987), or of three
mayor types of social and behavioral theories, for example, theories of social systems, microsocial theories, and
cognitive/physiological theories (McLeod & Blumler, 1987). A clear conceptual and operational demarcation of
levels of analysis depends on theoretical explication of concepts and linkages of the concepts. Specifically, we
suggest that three factors need be considered.
First, does a set of concepts describing the units of a proposed level correspond to a set of global (or integral)
properties (Eulau, 1986; Lazarsfeld & Menzel, 1961) of the units? Global properties refers to those properties of 8
unit that pertain only to the unit as a whole and are not based on information about its components (Eulau, 1986, p.
98). For example, media organizations may be described in terms of ownership (private ownership vs. state
ownership), organizational goals (commercial vs. public service organizations), and so on. These properties are not
based on measures of the individuals or specific branches within an organization; rather, they are measured by
looking at each organization as a whole. Further, no single member or branch alone displays these properties. It
would be difficult to make a strong case for a level if all the important concepts at that level were aggregates of
micro-level units below it.
Second, do changes at the proposed level exhibit a time frame that differs from other levels? For example, when
we study individuals, the appropriate time frame may be chronological age. When we study a media organization,
we may need to use time units such as a cycle of investment return, a season of programming, and so on. We have
no reason to assume that these time units coincide with those of members' aging or moving through their life
stages. A media organization does not necessarily change when its employees age or get fired.
Third, and perhaps most important, are there theoretically distinctive mechanisms that can be specified to connect
various concepts at different levels? For example, explanations of individuals' opinion changes may resort to
mechanisms such as learning and information processing, but explanations of a media organization's productivity
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 5 of 24
may need notions such as management style, intraorganizational communication, interorganizational relationships
with its patrons, flow of capital, and so on. Specification of theoretical mechanisms at various levels clearly is
crucial for achieving logical coherence. When the major theoretical mechanisms are identified the individual level
(e.g., priming, categorization, prototyping), the corresponding theory must call for micro-level analysis.(n5)
None of these factors relies on identifying physical attributes of "natural units." Eulau (1986, p. 114) makes this
point dear: "Levels are not distinguished by the relative size, complexity, or structural elaboration of the units."
Various levels of analysis may form a hierarchy only in that attributes of the units from a lower level, or attributes
of relationships of these lower level units, could be indispensable explanatory variables for explanations of
processes at a specific level. For example, an account of variations in media organizations' performance may need
variables such as employees' qualifications and interpersonal relationships among employees within each
organization.
Four Types of Relationships and a Metatheoretical Framework
Mass communication research extends across a wide spectrum of analytical levels. Studies range from research
conducted at the psychophysiological level, using measures such as EEG, skin temperature, heart rate, and 60 on
(e.g., Reeves, Thorson, & Schleuder, 1986) to studies making comparisons across nations (e.g., Blumler, 1983;
Gudykunst, 1987). A mayor challenge facing researchers in this field is to develop strategies for "integration of
levels" (Chaffee & Berger, 1987, p. 108).
The central question of achieving cross-level integration is not simply a matter of statistical data manipulation,
such as data reduction or aggregation, even though these are among the necessary operational procedures; rather, it
is a matter of first specifying "cross-level auxiliary theories" that "connect micro- and macro-processes" (Hannan,
1971, pp. 4-5) and then testing the hypotheses derived from these "auxiliary theories." Auxiliary theories here refer
to functional relationships between micro and macro variables (Hannan, 1971). The importance of specifying such
auxiliary theories is implied by the fact that a simple aggregation of micro relationships requires linear additive
functions relating micro variables to their aggregate-level counterparts (Hannan, 1971; van Daal & Merkies, 1984).
(n6) For example, whereas family income may have a linear relationship with household monthly expenditure on
entertainment' income inequality in a community, measured by variance of income distribution, may or may not be
linearly related to variation in community expenditure on entertainment, because nonlinear functions are involved
in linking micro-level attributes (family income and household expenditure) to macro-level attributes (income
inequality and variations in expenditure).
There are times, furthermore, when cross-level linkages are central components of explanations of certain
phenomena under investigation. In this case, the otherwise auxiliary theories become the focal theoretical linkages
to be examined empirically. For example, in the "structural model" of the knowledge gap (Tichenor, Donohue, &
Olien, 1973; Donohue, Tichenor, & Olien, 1973, 1975), two crucial hypotheses are proposed: that the degree of
community conflict on an issue is positively related to greater interest in the issue among the residents in the
community and that the more media coverage there is on this issue, the more interpersonal conversations there will
be about the issue. Interest and interpersonal communication in turn lead to increases in knowledge across all
sectors of the community. Two kinds of mechanisms are at work here: information control (macro-micro
influence) and information dissemination through media use and interpersonal conversations (micro-macro
influence). Both mechanisms affect knowledge-gap reduction in relation to community conflict; therefore, both
occupy a central theoretical position.
This multilevel perspective borrows from Coleman (1986, 1987, 1990) the distinction of four types of
relationships: macro to macro, macro to micro, micro to micro, and micro to macro. Consistent with Coleman's
formulation, we consider that the basic units at the micro level are individuals, because individuals are the basic
agents participating in the processes examined in social and behavioral sciences (Eulau, 1986; Coleman, 1990).
We also agree with Coleman that the units at the macro level can be properties of social-system-level units such as
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 6 of 24
community pluralism, inequality, autonomy of media systems, and so on. But we also make an important
modification of Coleman's model by including organizational, institutional, and social processes explicitly in the
model as theoretical explications of the two cross-level linkages (see Figure 1).
The modification of Coleman's model reflects our departure from Coleman's formulation in three important ways.
First, we do not accept his "methodological individualism," which attributes "more stable and general"
explanations of a social system to analysis of its internal components (Coleman, 1990, pp. 2-5). By advocating a
multilevel perspective, we do not mean to suggest a reductionist approach to communication processes (cf.
Lemert, 1981). Second, we see no need to accept Coleman's restriction of micro analysis to individuals' rational
actions. If anything, recent cognitive research has demonstrated that individuals are far from rational in processing
information, drawing inferences, and forming their behavior decisions (e.g., Kahneman, Slovic, & Tversky, 1982).
Third, we also believe that all four types of linkages are equally important for constructing mass communication
theories, although different emphases might be placed on them in answering a particular research question.
Therefore, although we agree with Coleman that explication of the micro-to-macro relationship has been a major
challenge to social science research, we do not think that a disproportionate focus on this relationship would lead
to a more rapid development of understanding communication processes than would more balanced attention to all
four relationships.
This view of the multilevel communication process assumes that individuals are implicated in any communication
process. Not only do individual professionals select and construct media content, but individuals also consume
mass media content. It is inevitable that individuals in either their cognitive or physiological domains are likely
candidates for the micro analytical level. However, all individual mass communication behavior is performed in s
social, cultural, institutional, and organizational context. Mass communication research has examined the
influences of social change on individuals' media-use behavior and individual-level effects (e.g., Ball-Rokeach &
DeFleur, 1976) as well as the impact of mass communication behavior on social change and organizational and
institutional outcomes (e.g., Dimmick & Coit, 1982; Gerbner, 1973; Katz & Szecsko, 1981; Tichenor et al., 1970).
Linking the two poles of the micro-macro continuum in this fashion requires construction of theories linking the
mechanisms of information processing and behavior formation with the processes of societal change ant
development.
Figure 1 is intended as a heuristic device to sensitize theorists to different types of relationships. By using this
diagram, we reiterate our earlier arguments that cross-level theorizing is an integral part of theory construction in
mass communication research (McLeod & Pan, 1989). The focus here is on the two cross-level relationships.(n7)
Macro-to-micro and micro-to-macro are dearly differentiated, in that they involve different sources of enactment
mechanisms. For example, although social control is a "top-down" process from macro to micro, collective
decision making involves a "bottom-up" process from individuals to groups through negotiations of the group
members. A major problem in the mass communication field is that the cross-level linkages are omitted, not
tested, or are assumed to be captured by simple aggregation or reduction.
Cross-level linkages are clearly the area where empirical sciences encounter the most difficulties. One reason is
that it is not empirically straightforward to test cross-level linkages. Another is that inclusion of variables from
venous levels in a statistical model complicates statistical analysis greatly (Goldstein, 1987; Mason, Wong, &
Entwisle, 1983). Due to these difficulties, we cannot test every specific cross-level hypothesis. We must rely on
theoretical seductions or empirical knowledge (Hannan, 1971, p. 59) in analyzing cross-level processes, that is,
social interactions and organizational processes.(n8) These processes might provide (a) theoretical descriptions of
the micro-macro connections, (b) logical explanations of macro and micro variables, and (c) theoretical
interpretations of the meanings of the macro and micro variables and their relationships. The point here is that the
two within-level linkages, that is, macro-macro and micro-micro, must be logically consistent with the cross-level
theorizing (see Price, 1988, for an example). Without these descriptions, cross-level inferences may have no
theoretical basis and must resort to "social telepathy" (Erbring & Young, 1979).
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 7 of 24
From this multilevel perspective, we can see that various theoretical traditions have projected a number of routes
through which individual members of a society are connected to societal properties. They can be classified by the
following dimensions:
Unidirectional versus bidirectional. Theories differ as to whether their cross-level linkages are conceived of as
unidirectional or bidirectional. Concepts such as power, institutionalization, control, division of labor, and
allocation of resources and opportunities are important explanatory mechanisms in the unidirectional conception of
macro-to-micro linkages. Notions like aggregation and mainstreaming represent unidirectional conceptions of
micro to macro linkages, because these notions assume that societal-level properties do not involve interactions
and structural constraints among members of the society (Bunge, 19 77; Eulau, 1986). Socialization, conformity
and coorientation involve bidirectional processes between individuals and society, in that societal norms, values,
and structural constraints are recognized, learned, or adopted by individuals.
Direct effects versus internalization. Theories also differ as to whether their cross-level mechanisms are viewed as
direct effects or as indirect effects mediated by individuals' internalization of societal influences. Although many
unidirectional cross-level concepts do not require internalization, some may involve subjective processes. For
example, power relationships associated with coercion do not require internalization processes. But power
relationships manifested through institutionalization may involve recognizing and accepting institutional rules,
norms, and values. Not all the bidirectional cross-level concepts require internalization. For example, conformity
may or may not involve internalization (Asch, 1956). Although modeling in the operant-conditioning paradigm
does not require internalization, the accumulation of a behavioral repertoire in social learning theory may indeed
require internal processing by individuals (Bandura, 1977).
Individual-centered versus society-centered. Differences also exist among theories as to whether individuals or
social systems are conceived of as being the locus of enactment. For example, socialization researchers differ
greatly in whether they place individuals in a passive learning position or characterize socialization as symmetric
interaction processes (Parsons & Bales, 1955; Easton & Dennis, 1969; Gecas, 1981; Wentworth, 1980). Notions
such as coercion and social control clearly attribute the primary role to society. In contrast, concepts such as social
interaction and coorientation allow individuals more freedom of action.
Conspiracy versus constraint. In constructing cross-level linkages, social processes can be conceptualized either as
being dominated by a group of individuals who seem intentionally to gain benefits or as individuals being affected
by socioeconomic status, norms, organizational hierarchies, and other sources of structural constraints. Conspiracy
theories (e.g., Marxist theory) tend to focus on conflict and are oriented toward large-scale or revolutionary
change; structural constraint theories (e.g., functionalist theory) tend to be homeostatically oriented toward social
order and structure. Concepts such as power and control are consistent with conspiracy theories, whereas
structuration (Giddens, 1984) and socialization tend to be constraint theories. This dichotomy should not be
equated with the dichotomy of conflict versus consensus theories. Conflict theories need not rest on conspiracy in
power arrangements. Constraint theories can also incorporate the concept of conflict; for example, community
conflict is a crucial concept in the knowledge-gap literature (Tichenor, Donohue, & Olien, 1980).
Static versus dynamic. Another important dimension among various cross-level linkages is whether their focus and
implications are more closely related either to social change or to maintenance of the system equilibrium. All
theories of social change involve cross-level mechanisms. For example, in the spiral of silence theory, the spiral of
public opinion toward the direction of majority opinion implies maintenance of the status quo. However, the
eventual gaining of ground by the avant-garde through the same spiral process may represent social change. In
diffusion-of-innovation research, the processes of inducing individuals to adopt innovations may result in social
change (Rogers, 1983). Concepts such as social control, institutionalization, and socialization may be more directly
related to social stability.
The five dimensions locate many theoretical propositions about macro and micro linkages. Such cross-level
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 8 of 24
linkages may help us develop coherent theories of mass communication process. It should not come as a surprise
that examples of almost all of the concepts can be found in the mass communication literature.
A multilevel perspective conceives of mass communication as a process consisting of both horizontal relationships
of media production and consumption linked by media content and vertical linkages of macro and micro concepts
through social and organizational mechanisms. There are two macro processes (media production in the context of
organizational and institutional relationships and media effects on social changes and stability) and two micro
processes (media professionals' decision making and production behavior and individual audience members'
processing media stimuli and forming their cognitive, affective, and behavioral responses). On both production and
consumption sides, there are two types of cross-level linkages: (a) social, institutional, and structural constraints on
individual media professionals or audience members and (b) integration or aggregation of individuals' opinions and
behaviors into macro-level social changes and stability.
Mass media content in this process functions as a carrier of thoughts and information that are jointly acted upon by
producers and consumers at various levels. Content is thus a product of social construction, and its conception in
terms such as "message systems" (Gerbner, 1973) is appropriate. However, upon its production, it also has
analytical independence, in that it imposes certain limitations over its interpretations (see Popper, 1972/1979, for
the notion of "World 3"). These limitations are closely related to the norms and codes of production process (Gans,
1979), and they are extended or strengthened by the public's assumptions about media content and prevailing
cultural norms and values (Kosicki & McLeod, 1990).
Content limitations materialize through social interactions and organizational processes on both sides. For
example, on the production side, professionalization and organizational constraints are major channels for
implementing professional norms and codes as well as behavioral rules (Breed, 1955; McLeod & Hawley, 1964;
Tuchman, 1978). On the consumption side, at the microsocial level, family communication patterns may be
important mechanisms for translating media content into culturally compatible terms (Chaffee, McLeod, &
Wackman, 1973; McLeod & Chaffee, 1972).
This multilevel framework of the mass communication process thus has two implications for constructing mass
communication theories. First, mass communication theories must explicitly convey propositions or assumptions
about both production and consumption of media content. It is theoretically beneficial to the field to state these
propositions clearly at multiple levels. Second, all theories of mass communication, regardless of their units of
analysis, contain stated or unstated propositions about how both societies and individuals work. They involve
certain assumptions about social structure, social dynamics, and social processes and also about how persons
process information and learn. The multilevel view of mass communication process forces researchers to examine
these assumptions and ideally to test the hypotheses that can be derived from them.
Application of the Multilevel Framework
In this section, we apply the multilevel framework to analyzing three prominent theories or perspectives about
media effects in our field. This analysis is not a complete review of the many empirical studies each of the theories
has generated; rather, it illustrates how the major theoretical components of each theory might be connected by the
four within-level and cross-level linkages. We hope to illustrate that such cross-level specifications may stimulates
theoretical development in these areas.
Knowledge Gap
The knowledge-gap hypothesis states how mass media may reinforce the positive relationship between
socioeconomic status and knowledge growth (Tichenor et al., 1970). Two theoretical perspectives have been
developed in the knowledge-gap research. One is social-psychological at She micro individuals level, focusing on
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 9 of 24
the individual's knowledge acquisition as a function of socioeconomic status and mass media use. The explanatory
variables invoked have included individual differences in cognitive ability, communication skills, levels of
interest, and so on (Ettema &i Kline, 1977). The other theoretical perspective is social-structural, focusing on
information control and distribution. The explanatory variables specified have been macro social in nature, such as
levels of community conflict, community pluralism, information control in media organizations, and so on
(Donohue et al., 1975; Tichenor et al., 1973, 1980).
The two theoretical perspectives have generated empirical research at both individual ant societal levels. At the
micro individual level, research has examined rates of knowledge growth and levels of knowledge as the
dependent variables. The knowledge-gap phenomenon is thus conceived as primarily the extent to which low
socioeconomic status impairs one's learning from mass media and thus the information poor fail to learn as much
from media as do the information rich. At the macrosocial level, research has focussed on the processes of
information control and its relationship to community structure and power hierarchy. Little attention has been
devoted to integrating the two levels.
Careful reading of the literature, however, reveals a number of cross-level linkages that might be useful in
connecting the two within-level linkages: micro individual-level learning processes and macro-level processes of
information control and distribution. On both the production and consumption sides, cross-level linkages need to
be specified in order to advance a coherent theory of knowledge gaps.
On the production side, a key micro-macro proposition can be derived from structural analyses of communities in
terms of the relationships between community power structure and individual editors (Donohue et al., 1972). We
may hypothesize that structural locations of the editors, the size of news hole, and community conflict, restrict the
editors' decisions on news coverage of certain controversial issues. More specific theoretical analysis might reveal
that these constraints may be mediated by individual editors' cognitive processes, including internalized
professional values and behavioral codes a: well as their perceptions of issues.
On the consumption side, two crucial cross-level linkages might be specified. One is the relationship between
community structural characteristics and levels as well as patterns of interpersonal communication. For example,
we may observe more within-status interpersonal communication in a highly, pluralistic community. In a highly
homogeneous community, interpersonal communication may be subject to fewer restrictions of social barriers.
Controversy may lead to better awareness of common interests in a relatively homogeneous community but to
polarization in very pluralistic communities Therefore, controversy over an issue may result in different patterns of
interpersonal communication in homogeneous versus pluralistic communities.
Another cross-level linkage is the relationship between individual knowledge acquisition and aggregate-level
knowledge distribution. In a population that already has inequality in knowledge distribution, equal rates of
acquisition maintain the existing pattern of knowledge distribution. To close the knowledge gap in this case would
require faster rates of knowledge gain among those who are initially located at the lower end of the knowledge
distribution.
A greater integration of the micro and macro studies in this area may also help advance our theoretical thinking on
the relationship between knowledge distribution and conditions of social change and stability. (Tichenor et al.,
1970; Robinson, 1972). To be more specific, a multilevel perspective directs attention to the following cross-level
linkages: (a) relationships between the characteristics of a media market, media accessibility, and diversity of
media use among individuals; (b) relationships between variations in individual's knowledge level and
distributions of effective social and political participation; and (c) the impact of widening or narrowing knowledge
gaps on further information seeking and learning among individuals.
Cultivation
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 10 of 24
The cultivation hypothesis states that because of the ubiquitousness of the symbolic environment created by
television and the homogeneity in television content as a system, television tends to cultivate "standardized roles
and behavior" (Gerbner & Gross, 1976). A micro version of this hypothesis is that heavy television viewers are
more likely than light viewers to develop a perception of social reality similar to the reality depicted by television.
The cultivation hypothesis is based on a series of assumptions about society, and media institutions and
organizations, as well as about individual audience members. Some of these assumptions could be specified as
testable theoretical propositions. This would necessarily lead to advancement of both within-level and cross-level
linkages. Again, our analysis is organized along the division of production and consumption.
On the production side, the key concepts are institutional roles and their relations to power (Gerbner, 1973).
Political and military positions and economic resources furnish sources of power. They strongly influence the
decision making of executives in media management and professionals in media organizations. Ideological
orientations of power holders are embedded unobtrusively in the system of television content as a whole. These
arguments clearly indicate a macro-centered unidirectional process in message production.
However, for the propositions stated above to be validated empirically, the research focus must be directed toward
decision-making processes at the individual level. Three cross-level linkages require examination: (a) the
restrictions of institutionalized norms and rules over media professionals; (b) the intraorganizational as well as
extraorganizational (e.g., relationships between media organizations and their patrons as well as competition
among media organizations) constraints over media professionals; and (3) the relationships between media
professionals and other actors in a society, such as sources and audiences. Examination of these cross-level
linkages might help us to understand how institutional roles and power relationships function in creating
ideological orientations in media content.
On the consumption side, theoretical efforts have been made to specify individual-level mechanisms: for example,
incidental learning (Hawkins, Pingree, & Adler, 1987) and selective forgetting (Shapiro, 1991). Empirical
research needs to examine the propositions about individual audience members as isolated, passive, and ritualistic
viewers of television. More important, theoretical efforts should examine the roles played by interpersonal
relationships in the process of mass communication (Chaffee, 1972; McLeod & Chaffee, 1972). Such cross-level
linkages may aid the empirical investigation of whether cultivation means individuals' accepting "TV reality" or,
alternatively, television serving as one among various sources of individuals' experiences with social reality. An
important consequence of such cross-level theorizing might be specification of the conditions under which there is
a shrinking of variance in the perceptions of reality called "mainstreaming" and under what conditions there is an
increasing variance in the perception of reality called "resonance."
A multilevel perspective would force us to ask the following questions essential to formulating a coherent
cultivation theory: First, to what e extent do individual media professionals internalize the behavioral codes that
reflect power relationships? To what extent and how are they constrained by organizational structure and
institutional restrictions in their decision making process? Second, is television so ubiquitous that there is little
diversity in our symbolic environment? What, if any, contributions do social networks make to our symbolic
environment? Third, are there any active elements in audience members' television viewing? If yes, what are their
contributions to the process of construction of social reality, rather than acceptance of TV reality? Fourth, what are
the microsocial processes that may account for homogenization, represented by mainstreaming, and diversification,
represented by resonance? Under what conditions will one be more dominant than the other?
Spiral of Silence
Among the three theories analyzed here, Noelle-Neumann's (1984) spiral of silence theory is closer to a theory of
social dynamics, in that it specifies not only predictions but also theoretical mechanisms. Briefly, the theory says
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 11 of 24
that individuals have a quasi-statistical ability to perceive the societal opinion climate. Due to fear of isolation,
those individuals who perceive themselves to be on the minority side tend to remain silent, whereas those
individuals who perceive themselves on the majority side tend to be more willing to spears out. To the extent that
whichever side is being expressed influences the perceptions of the opinion climate, the process spirals over time,
leading to increasing dominance of one direction of opinion. Under such an opinion climate, only two small groups
of people, representing different social forces, are willing to insist on their own minority views: the avant-garde
and the hard core. The former represent changes that will come eventually, whereas the letter represent the
resistance to change. During this process, mass media, due to their ubiquitous presence and consonant content,
tend to amplify one kind of opinion, thus facilitating the spiral process.
In this theory, the micro individual mechanism is characterized by the relationships between perception of the
aggregate opinion climate and opinion expression behavior. Interpersonal relationships constitute the context for
both. But on the micro-macro side (using the representation of Figure 1), interpersonal relationships are relatively
unimportant, in that aggregation of individuals' opinions constitutes the macro-level opinion climate. The theory
can be formulated more precisely and can generate more fruitful empirical research if all four types of
relationships are clearly stated.
On the production aide, the theory assumes that journalists are unproblematic at the individual level, in that they all
follow their professional codes and chase "news values" (Noelle-Neumann, 1981). The aggregate of journalists
shows "a peculiar kind of consensus": Journalists express, in their reports, the opinions of their fellow journalists.
The governing principles of news values in everyday journalistic work "evoke a remarkable consonance" of media
content (Noelle-Neumann, 1981, p. 141). These theoretical arguments lead Noelle-Neumann to reach a conclusion
about media content similar to that of the cultivation hypothesis: Media content is not only consonant but also
cumulative and ubiquitous.
The homogeneity hypothesis of media content has to be examined empirically, however. Of even greater
importance is theoretical work examining not only the institutional restrictions and organizational constraints over
journalists but also journalists' relationships with their peers, with audience members, and with sources. One may
very well borrow theoretical insights from Blumler and Gurevitch (1981) about role relationships between
journalists and sources and from Donohue et al. (1972) about structural and power relationships between
journalists and other actors in the social structural and power hierarchies.
Detailed theoretical formulation of how journalists cover news events is needed. Currently, the theory assumes an
automatic process of journalists following "professional norms" and chasing news values. But more fruitful
empirical research might examine journalists' decision-making processes. These processes include such microlevel mechanisms as perception, learning, judgment formation, and expression, as well as the macro-micro
processes outlined in the previous paragraph.
On the consumption side, researchers need to develop a theoretical framework of how individuals receive "cues" of
opinion climate from mass media and how they make judgments about the direction of majority opinion based on
the external cues. These processes involve micro-level mechanisms such as learning, cognitive heuristic, judgment
formation, and so on. It remains an empirical question whether individuals form their judgment about the opinion
climate based on simple aggregation or, alternatively, whether they calculate the macro opinion climate based on
social categorization and levels of salience of various opinions. It also remains to be investigated how the relative
weights assigned to media news accounts or to reports of public opinion polls are balanced against the perceived
frequency of opinion expression in various types of social situations.
In spiral of silence theory, the perception-to-expression-to-reinforcement process constitutes a mayor cross-level
mechanism with bidirectional flow of causal forces. To this extent, the theory makes an explicit effort to formulate
cross-level linkages. But such formulation may be greatly broadened by including additional within-level as well
as cross-level linkages. In addition to the aforementioned micro-level mechanisms, we may also examine the
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 12 of 24
processes of opinion articulation and mobilization. It is highly questionable whether aggregation is the only
channel linking micro individual opinions to macro-level public opinions. Rather, we may need to examine the
conditions under which (a) certain opinions are articulated more systematically and therefore, carry more weight
and salience; (b) certain groups representing particular opinions are better mobilized to express their opinions
despite their being in the minority; and (c) certain opinions are more saliently presented by mass media due to the
opinion holders' more advantaged positions in the power hierarchy.
In summary, development of spiral of silence theory would require increased effort in four areas. The first is to
formulate precise micro-level mechanisms of journalists' decision-making processes. The second is to develop
testable hypotheses about the institutional, organizational, and social constraints over journalists' professional
practices. The third is to examine the cognitive processes of individuals acquiring external cues of opinion
climates, making judgments about opinion climates and deciding whether to express their opinions overtly. The
fourth is to examine the multiple micro-macro processes from micro individual opinions to macro public opinions.
Summary
The analyses presented above demonstrate that all three theories need further specification of the four types of
relationships. The basic point is that theoretical efforts in these areas have not been precise enough to specify both
within-level and cross-level linkages explicitly. Because of such neglect, the overall quality of each theory cannot
be adequately judged; therefore, theoretical advancement is impeded. Put in more positive terms, we hope that our
analyses of the three theories show that theoretical development in our field would benefit from clarifying the four
types of relationships in a multilevel perspective.
Research Options
Multilevel perspectives in mass communication research clearly require methodological pluralism involving
integration of "hard" and "soft" data obtained from diverse sources describing units of analysis at different levels.
A major challenge facing us is how to apply available research options and develop new research strategies. It
should be helpful to review multilevel research strategies and methodological tools that may be applicable to mass
communication research.
Measurement
In multilevel research, measures of theoretical concepts must be built upon conceptual work linking macro- and
micro-level concepts. Based on different kinds of the cross-level linkages, researchers have specified various
properties of macro-level units (Lazarsfeld & Menzel, 1961; Eulau, 1986), including global, structural, relational,
and analytical properties.
Global properties, as they are defined earlier, are diversely measured through procedures such as field observations
or the use of existing archival records and media market data. Relational properties refer to the characteristics of a
unit that describe the relationships among its components. Direct measures of relational properties can be obtained
through techniques such as paired-sample interviews, observations of interpersonal interactions, and sociometric
survey questions. Structural properties refer to the organizational characteristics that result from formally or
informally defined positions and their relationships Measures of these properties involve such techniques as direct
observation of group or organizational functioning and analysis of the intragroup and intraorganizational
behavioral rules and regulations. Institutional analysis of organizations and groups can often provide data on these
properties. Analytical properties may involve aggregates of individual data, that is, transformation of observations
obtained from units at some lower level (Hannan, 1971). For example, use of television in various communities
might be compared by examining differences between samples of individuals in each community in terms of the
mean or dispersion of time they spend watching television.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 13 of 24
This differentiation of macro-level properties calls for specification of auxiliary theories that "express the rules
according to which observations are grouped" (Hannan, 1971, p. 17). First, we should carefully guard against the
atheoretical practices that use aggregate data to represent all macro properties. Not all macro-level variables have
direct micro-level counterparts. For example, variables describing global and structural properties, and even some
relational properties, have at best very indirect and incomplete micro manifestations. Aggregations should "be
treated as special cases of the more general theory that allows for the operation of organizational
variables" (Blalock & Wilken, 1979, p. 15).(n9)
Sample Design
Multilevel research requires observation of units of different theoretical levels. The essence of multilevel research
is its making comparisons across units at each of the micro and macro levels.
In developing a multilevel research design, we may conceive of response variance as being partitioned into macrolevel and micro-level components Research objectives sometimes require consistent and unbiased estimates of
population variance at both the macro and micro levels, along with their covariance. When this is is the case,
sampling designs should allow for nested sampling, including random selection of both macro and micro units. For
more restrictive purposes, alternative designs may be useful (Przeworski & Teune, 1970).
Multistage probability sampling design. The crucial issue here is to define hierarchically nested populations with
units residing at different levels. For example, studies of student academic achievement may consider effects of
schools and of classrooms, as well as of individual characteristics. Three populations--school classroom, and
individual--are nested in a hierarchical order such that an entire lower-level population is nested in the population
at an adjacent higher level. Because of the complete nested structure researchers can proceed by first sampling
schools from the school population, followed by sampling classrooms from each school selected and individuals
from selected classrooms. In an analogous multilevel study of journalists, it would be necessary to draw sequential
probability samples of media conglomerates, media organizations, and journalists.(n10)
Maximizing variance or minimizing variance designs. Przeworski and Teune (1970) propose two designs that can
serve different research purposes The maximizing variance ("most different systems") design can be useful to test
the generalizability of within-system micro relationships across different macro units. If no difference is found in
micro relationships between two "most different" macro units, macro-level variables examined can be regarded as
being largely irrelevant to the explanation of the variations in the lower-level response variables.
The minimizing variance ("most similar systems") design attempts to identify macro-level variables that do differ
between systems and that are crucial for explaining variations in the dependent variables. With a priori knowledge
of the macro-level variables that are relevant to micro-level relationships under examination, one can identify the
macro-level variables that contribute to differences in micro-level relationships.
We should recognize that because the macro units are not selected with probability sampling, neither of these two
designs allow estimates of macro population parameters. They may be useful nonetheless for identifying
differences and similarities across macro units.
Observational case study. Survey research is often regarded as being responsible for the individual focus of the
social and behavioral sciences (Coleman, 1981; Stipak & Hensler, 1982). The natural units of observation in
survey research undoubtedly are individuals. Nevertheless, survey research can be supplemented with
observational case studies that may be very useful for describing intragroup or intraorganizational processes and
for identifying global, structural, and relational properties of macro units. Examples of such studies can be found in
mass communication (Sigal, 1973; Sigelman, 1973; Gans, 1979; Ettema & Whitney, 1982).
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 14 of 24
Statistical Techniques
Statistical analysis of multilevel data has posed major problems to the behavioral and social sciences. Space does
not permit a detailed technical presentation here. A brief review, however, provides references to different
statistical techniques that are potentially applicable to various kinds of multilevel data.
Contextual analysis. Contextual analysis is basically an analytical approach to predicting micro-level variations
with macro-level variables, or, "contextual variables," in addition to micro-level predictors. A systematic treatment
of contextual analysis can be found in Boyd and Iversen (1979). Additional sources on issues related to contextual
analysis include Farkas (1974), Firebaugh (1979), and Stipak and Hensler (1982). Note that these treatments of
contextual analysis rely on the assumptions of independent micro-level observations between, as well as within,
macro units and equivalence of micro-level measures across macro units. They focus on single-equation models,
that is, direct effects of micro- as well as macro-level variables.
A causal modeling approach to contextual analysis based on different conceptions of causal mechanisms of macro
influences is proposed by Erbring and Young (1979). They focus on the problem of specifying theoretically
meaningful mechanisms that explain the impact of macro variables on micro-level relationships and argue that a
genuine contextual process is "interaction among individuals within a particular social structure" (p. 407) and
contextual effects can be modeled as "endogenous feedback" (p. 408). Their approach is one step closer to theoryguided contextual analysis.(n11)
Multilevel linear models. Multilevel linear models consist of various linear models for hierarchical data:
observations made at both micro and macro levels. The general models are flexible enough to allow for estimates
not only of the effects of micro- and macro-level predictors but also of various correlations among the micro- and
macro-level predictors. Contextual analysis models are therefore a special case of the multilevel linear models, in
that the micro- and macro-level predictors are assumed to be uncorrelated. Systematic treatment of multilevel
linear modeling can be found in Goldstein (1987), Mason et al. (1983), and Raudenbush and Bryk (1986). Special
purpose computer packages have also been developed for estimating various types of multilevel linear models.
(n12)
Simultaneous equations models. A simultaneous equation model allows for simultaneous estimation of a system of
equations that is constructed on the basis of causal theories. This approach can be used to examine two other
multilevel research questions. When independent samples from two or more macro units are available and the
research question involves testing the hypothesis of equality of within-system relationships, multigroup models
can be estimated via LISREL routines (e.g., Schoenberg, 1972; Specht & Warren. 1976).
Another multilevel research question is how to decompose variations of micro units into similarities due to
common experiences associated with being in the same macro unit, such as community, school, family, and so on
and differences due to individual characteristics. Statistical techniques for such simultaneous estimation of both
macro- and macro-level parameters are available in the LISREL framework (e.g., Hauser & Mossel, 1985).
Other econometric and demographic models. Econometric and demographic literature contain various statistical
models applicable to multilevel mass communication questions. In econometrics, for example, two kinds of
micro- to macro-level effects have been conceptualized: those due to changes in marginal distributions and those
due to changes in micro-level functions (Stoker, 1985). Consideration of these two types of effects ant their
statistical treatment are provided in much less mathematically complex ways by Boyd and Iversen (1979). The
demographic literature contains models for conceptualizing and estimating effects due to changes in demographic
composition and social changes associated with them, that is, cohort and period effects (e.g., Danowski &
Ruchinskas, 1983; Fienberg & Mason, 1985).
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 15 of 24
Two General Approaches to Cross-Level Theorizing
In his model of micro and macro relationships, Hannan (1971, pp. 18-23) suggests that there is only one "degree of
freedom" in specifying two within-level and two cross-level relationships. In other words, given two of the
linkages, a researcher is "free" to formulate the other linkages. This is the case because in his formulation, the two
cross-level linkages are regarded as being in the same category of "aggregation relations." More generally,
however, the two cross-level linkages are theoretically different. Despite differences in cross-level formulation,
Hannan's two general approaches in aggregation can be modified to suit our purpose of cross-level theorizing.
Hannan's first approach is that, given the variables and theoretical propositions within the respective micro and
macro levels, we can search for relationships linking micro and macro levels that are consistent with the two
within-level propositions. For example, in studies of knowledge gaps, given the macro-level proposition that
community conflict over an issue leads to more intensive coverage of the issue and thus to a narrowed knowledge
gap and the micro-level proposition that more media exposure leads to greater learning of information, a researcher
may specify cross-level linkages connecting communities and individuals so that they are consistent with both
macro and micro linkages. These linkages could include the following: Perceived conflict over an issue leads to
more interest on the issue; and given the assumption of a fixed level of total information, learning about the issue
among the individuals across all social sectors leads to reduced variance among them in levels of knowledge.
Other linkages can be specified as consistent with the two within-level linkages.
The alternative approach from Hannan is that given one within-level theoretical proposition and two cross-level
propositions, a researcher can define a set of variables at another level and search for within-level theoretical
propositions at that level that are consistent with the theoretical linkages provided. Again the knowledge-gap
research can be used for illustration. After asserting the cross-level linkage that increasing conflict over an issue
leads to a higher level of interest on the issue among individuals of all social sectors and suggesting the individuallevel learning mechanism, it is possible, given the assumption of a fixed amount of knowledge, to formulate a
macro-level proposition that increasing community conflict over an issue i& related to a narrowed knowledge gap.
It should be pointed out, however, that compared with Hannan's case, both approaches in mass communication
require additional assumptions about research practices because the cross-level linkages on the production and
consumption sides are likely to be different. Essentially, this implies that communication researchers need more
information to gain that one degree of freedom.
Concluding Remarks
In this article, we advocate an epistemological view of level-of-analysis issues, thus obviating ontological disputes
and the confusion they produce. Our view thereby defines the cross-level analysis issues more narrowly than have
some theorists (e.g., Paisley, 1984). But this epistemological view also broadens the theoretical horizon in mass
communication research to include recognition of the continuum of micro and macro levels with multiple ways of
constructing hierarchically ordered analytical levels, multilevel analysis of mass communication processes, and
specification of within-level and cross-level linkages. We have illustrated how mass communication theories
might be advanced by analyzing both within- and cross-level linkages and have pointed out analytical strategies
available for cross-level empirical studies.
We have not discussed many important level-of-analysis issues because of space limitations; for example,
conceptual equivalence of micro variables across macro units (Przeworski & Teune, 1970), conceptual equivalence
of micro variables across macro variables across time units, simultaneous cross-level and cross-time studies (e.g.,
cohort analysis), cross-nesting of micro units in a number of macro units (Goldstein, 1987), the variations of crosslevel linkages over different stages of a macro system and its changes, and so on. We hope that this volume sets
the stage for more advanced conceptual work on these and other issues.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 16 of 24
Of central importance to the study of levels-of-analysis issues is what we call "microsocial theories." These
theories include symbolic interactionism (e.g., Blumer, 1969), coorientation (e.g., McLeod & Chaffee, 1972), role
theories (e.g., Jackson, 1972), socialization and life-span development theories (e.g., Featherman & Lerner, 1985),
theories of primary groups and reference groups (e.g., Festinger, 1954; Hyman & Singer, 1968; Shils, 1951), and
so on. All these theories may be useful in specifying linkages between macro and micro units and provide
theoretical insight into the mechanisms at work.
It should be recognized that within-level variables differ in their analytical potential for making connections across
different levels. Further, their connections need two kinds of concepts. One kind describes the intermediating units
and relationships that bridge macro and micro units, such as roles, status, families, primary groups, reference
groups, and so on. In comparison with other types of concepts, they seem particularly affected by the perceptions
and interpretations of individual actors. The other kind, what we might call "boundary-friendly" level-spanning
variables, describe the nature of the connections between macro and micro units, such as socialization, interactions,
coorientational states, and so on. These concepts provide the language for us to think and talk about cross-level
linkages. Microsocial theories contain a rich repertoire of these concepts.
Applying these microsocial theories to explicate cross-level linkages may lead in various ways to more fruitful
mass communication research: within-level linkages can be built upon more solid theoretical ground; new withinlevel linkages can be derived from these microsocial theories; and connections between micro and macro levels
can be made. Microsocial theories thus play crucial roles in multilevel research. We hope that multilevel research
is geared toward construction of mass communication theories rather than simply toward acceptance of statistical
operations that are built upon "social telepathy."
Notes
(n1.) We can find work specifically devoted to level-of-analysis issues in fields as diverse as sociology (e.g.,
Alexander, Giesen, Munch, & Smelser, 1987; Coleman, 1987; Knorr-Cetina & Cicourel, 1981); social psychology
(e.g., Doise, 1986); anthropology (e.g., De Walt & Pelto, 1985); political science (e.g., Eulau, 1986); and sociology
of education (e.g., Burstein, 1980; Oosthoek & van den Eeden, 1984; Raudenbush & Bryk, 1988).
(n2.) It is no less important to point out here that many of the metaconcepts are heavily ideology laden. The labels
themselves often imply directions of policy making Such characteristics of metaconcepts also impede their use in
analytical treatment of legitimate scientific issues.
(n3.) Ontological primacy refers to the assumption about the basic elements of objective existence as sources of
enactment. From the perspective of "individual primacy," individuals constitute the basic elements of society and
the source of enactment. Individuals are assumed to be rational and purposeful and to act according to their free
will. Societies are the creations or construction of their rational and purposeful actions. From the perspective of
"society primacy," individuals are not sociologically meaningful without their being placed in a social context.
Individuals act within the constraints of environment, social structure, and historical limitations. Individuals are
social beings. See an overview in Alexander and Giesen (1987, pp. 1-37).
(n4.) We caution readers not to interpret the term continuum in any ontological sense. Levels of analysis are
theoretical constructions. In the sense that there are infinite ways of conceptualizing reality, theoretical concepts
may be hypothetically arranged in a continuum from micro to macro. Any discrete view of this continuum may run
the risk of limiting the range of theoretical perspectives. This notion is not the same as conceiving ontological
existence as having no division of micro and macro objects.
(n5.) However, this does not exclude the possibility that higher-level explanatory factors will be relevant. It is
possible that although major theoretical mechanisms are identified at the micro-individual level, some higher-level
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 17 of 24
contextual and structural variables are still theoretically important in two ways. First, the higher-level variables
may have direct impact on the dependent variables through some macro-micro linkages (e.g., Erbring & Young,
1979). Second, the higher-level variables may constitute contingent conditions that alter the relationships between
two or more micro-level variables (see Burstein, 1980; McLeod & Reeves, 1980/1981).
(n6.) In the econometric literature, aggregation refers to a statistical procedure that translates micro-level linear
relationships into a macro-level linear relationship. This translation involves specifying four relationships: micromicro, macro-macro (aggregates), and two cross-level relationships. Only when the four relationships form a linear
system, that is, each of them is a linear additive relationship, can aggregation be mathematically possible. For
mathematical proof, see Hannan (1971).
(n7.) This is not to suggest that the two cross-level linkages are more important than the two within-level linkages.
Rather, the point of this diagram is that the cross-level linkages are the places where empirical researchers
encounter the most difficulties and yet where fruitful theoretical advancement may be made.
(n8.) The phrase social interactions and organizational process is used here to represent a broad range of theoretical
processes, including social psychological interpersonal interaction, group dynamics, and social influences, as well
as sociological, economic, and political processes of organizational, institutional, and structural constraints. This
point will become clear in the discussion of the dimensions of various cross-level theoretical mechanisms.
(n9.) In addition, simple aggregation in practice rarely tests its underlying assumptions: that is, that aggregating a
micro variable cancels out its measurement error and that the data points within each level are independent
observations not only between macro units but also within each macro unit. Violation of these assumptions may
lead to biased estimates of aggregate-level variance.
Aggregate data are also often selected by their availability or convenience without probing the concepts behind the
measures. Over the years, governmental and other research agencies have collected data such as economic indices,
social indicators, and to a lesser degree, cultural indicators. Many of these indicators can be employed as measures
of global, structural, or analytical properties of some macro units (Borgatta & Jackson, 1980). However, one needs
to keep in mind that collection of these data was guided by conceptualizations and purposes different from those of
the researcher. Construct validity of these indicators thus needs to be carefully established for their revised usage.
(n10.) This elementary sampling issue has been largely ignored in research designs. The Johnstone, Slawski, and
Bowman (1976) study of 1,300 practicing journalists, for example, did not include disproportionate stratified
sampling at the macro organization level. As a result, the data could be aggregated by industry (e.g., print,
broadcast) but not at the organizational level.
(n11.) A major critique of conventional contextual analysis has been its failure to specify theoretical mechanisms
linking micro and macro variables (Hauser, 1970, 1974).
(n12.) The available computer packages specifically designed for multilevel modeling include GENMOD
(Generalized Multilevel Modeling), developed by Mason, Anderson, and Hayat (1987), and HLM (Hierarchical
Linear Models), developed by Bryk, Raudenbush, Seltzer, and Congdon (1986).
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 18 of 24
Figure 1: Four types of relationships in mass communication.
References
Alexander, J. C., & Giesen, B. (1987). From reduction to linkage: The long view of the micro-macro debate. In J.
C. Alexander, B. Giesen, R. Munch, & N. J. Smelser (Eds.), The micro-macro link (pp. 1-42). Berkeley:
University of California Press.
Alexander, J. C., Giesen, B., Munch, R., & Smelser, N. J. (Eds.). (1987). The micro-macro link. Berkeley:
University of California Press.
Asch, S. E. (1956). Studies of independence and conformity: A minority of one against a unanimous majority.
Psychology Monograph, 70(9, Whole No. 416).
Ball-Rokeach, S., & DeFleur, M. L. (1976). A dependency theory of mass media effects. Communication
Research, 3, 3-21.
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall.
Berger, C. R., & Chaffee, S. H. (Eds.). (1987). Handbook of communication science. Newbury Park, CA: Sage.
Blalock, H. M., Jr., & Wilken, P. H. (1979). Intergroup processes: A micro-macro perspective. New York: Free
Press.
Blumer, H. (1969). Symbolic interactionism: Perspective and method. Englewood Cliffs, NJ: Prentice-Hall.
Blumler, J. G. (Ed.). (1983). Communicating to voters: Television in the first European parliamentary elections.
Beverly Hills, CA: Sage.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 19 of 24
Blumler, J. G., & Gurevitch, M. (1981). Politicians and the press: An essay on role relationships. In D. D. Nimmo
& K. R. Sanders (Eds.), Handbook of political communication (pp. 467-493). Beverly Hills, CA: Sage.
Borgatta, E. F., & Jackson , D. J. (Eds.). (1980). Aggregate data: Analysis and interpretation. Beverly Hills, CA:
Sage.
Boyd, L. H., Jr., & Iversen, G. R. (1979). Contextual analysis: Concepts and statistical techniques. Belmont, CA:
Wadsworth.
Breed, W. (1955). Social control in the news room: A functional analysis. Social Forces, 33, 326-327.
Bryk, A. S., Raudenbush, S. W., Seltzer, M., & Congdon, R. T. (1986). An introduction to HLM: Computer
program and users' guide. Chicago: University of Chicago Press.
Bunge, M. (1977). Treatise on basic philosophy (Vol. 3-4). Dordrecht, Holland: D. Reidel.
Burstein, L. (1980). The role of levels of analysis in the specification of education effects. In R. Dreeden & J. A.
Thomas (Eds.), The analysis of educational productivity: Vol. 1. Issues in microanalysis (pp. 119-190).
Cambridge, MA: Ballinger.
Chaffee, S. H. (1972). The interpersonal context of mass communication. In F. G. Kline & P. J. Tichenor (Eds.),
Current perspectives in mass communication Research (pp. 95-119). Beverly Hills, CA: Sage.
Chaffee, S. H., & Berger, C. R. (1987). Levels of analysis: An introduction. In C. R. Berger & S. H. Chaffee
(Eds.), Handbook of communication science (pp. 143-145). Newbury Park, CA: Sage.
Chaffee, S. H., McLeod, J. M., & Wackman, D. (1973). Family communication patterns and adolescent political
participation. In J. Dennis (Ed.), Socialization to politics (pp. 349-364). New York: Wiley.
Coleman, J. S. (1981). Longitudinal data analysis. New York: Basic Books.
Coleman, J. S. (1986). Social theory, social research, and a theory of action. American Journal of Sociology, 91,
1309-1335.
Coleman, J. S. (1987). Microfoundations and macrosocial behavior. In J. C. Alexander, B. Giesen, R. Munch, & N.
J. Smelser (Eds.), The micro-macro link (pp. 153-173). Berkeley: University of California Press.
Coleman, J. S. (1990). Foundations of social theory. Cambridge, MA: Harvard University Press.
Comte, A. (1983). Auguste Comte and positivism: The essential writings (G. Lenzer, Ed.). Chicago: University of
Chicago Press. (Original work published 1830-1842)
Cushman, D. P., & Craig, R. T. (1976). Communication systems: Interpersonal implications. In G. R. Miller (Ed.),
Explorations in interpersonal communication (pp. 37-58). Beverly Hills, CA: Sage.
Danowski, J. A., & Ruchinskas, J. E. (1983). Period, cohort, and aging effects: A study of television exposure in
presidential election campaigns, 1952-1980. Communication Research, 10, 77-96.
DeWalt, B. R., & Pelto, P. J. (Eds.). (1985). Micro and macro levels of analysis in anthropology: Issues in theory
and research. Boulder, CO: Westview.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 20 of 24
Dimmick, J. (1979). The gatekeeper: Media organizations as political coalitions. Communication Research, 6,
203-222.
Dimmick, J., & Coit, P. (1982). Levels of analysis in mass media decision making: A taxonomy, research strategy,
and illustrative data analysis. Communication Research, 9, 3-32.
Doise, W. (1986). Levels of explanation in social psychology. Cambridge: Cambridge University Press.
Donohue, G. A., Tichenor, P. J., & Olien, C. N. (1972). Gatekeeping: Mass media systems and information
control. In F. G. Kline & P. J. Tichenor (Eds.), Current perspectives in mass communication research (pp. 41-69).
Beverly Hills, CA: Sage.
Donohue, G. A., Tichenor, P. J., & Olien, C. N. (1973). Mass media functions, knowledge and social control.
Journalism Quarterly, 50, 652-659.
Donohue, G. A., Tichenor, P. J., & Olien, C. N. (1975). Mass media and the knowledge gap: A hypothesis
reconsidered. Communication Research, 2, 3-23.
Easton, D., & Dennis, J. (1969). Children in the political system. New York: McGraw-Hill.
Erbring, L., & Young, A. A. (1979). Individuals and social structure: Contextual effects as endogenous feedback.
Sociological Methods and Research, 7, 396-430.
Ettema, J. S., & Kline, F. G. (1977). Deficits, differences and ceilings: Contingent conditions for understanding the
knowledge gap. Communication Research, 4, 179-202.
Ettema, J. S., & Whitney, D. C. (Eds.). (1982). Individuals in mass media organization: Creativity and constraint.
Beverly Hills, CA: Sage.
Eulau, H. (1977). Multilevel methods in comparative politics. American Behavioral Scientist, 21, 39-62.
Eulau, H. (1986). Politics, self, and society: A theme and variation. Cambridge, MA: Harvard University Press.
Farkas, G. (1974). Specification, residuals and contextual effects. Social Methods and Research, 2, 333-363.
Featherman, D. L., & Lerner, R. M. (1985). Ontogenesis and sociogenesis: Problematics for theory and research
about development and socialization across the lifespan. American Sociological Review, 50, 659-676.
Festinger, L. (1954). A theory of social comparison process. Human Relations, 7, 117-140.
Fienberg, S. E., & Mason, W. M. (1985). Specification and implementation of age, period and cohort models. In
W. M. Mason & S. E. Fienberg (Eds.), Cohort analysis in social research: Beyond the identification problem (pp.
45-88). New York: Springer-Verlag.
Firebaugh, G. (1979). Assessing group effects: A comparison of two methods. Sociological Methods and
Research, 7, 384-395.
Gamoran, A. (1986). Instructional and institutional effects of ability grouping. Sociology of Education, 59, 185198.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 21 of 24
Gans, H. J. (1979). Deciding what's news. New York: Vintage Books.
Gecas, V. (1981). Contexts of socialization. In M. Rosenberg & R. H. Turner (Eds.), Social psychology:
Sociological perspective (pp. 165-199). New York: Basic Books.
Gerbner, G. (1973). Cultural indicators: The third voice. In G. Gerbner, L. Gross, & W. H. Melody (Eds.),
Communication technology and social policy: Understanding the new "cultural revolution" (pp. 555-173). New
York: Wiley.
Gerbner, G., & Gross, L. (1976). Living with television: The violence profile. Journal of Communication, 26, 172199.
Giddens, A. (1984). The constitution of society: Outline of the theory of structuration. Glasgow: Polity Press.
Gudykunst, W. B. (1987). Multilevel models in education and social research. London: Charles Griffin.
Gudykunst, W. B. (1987). Cross-cultural comparisons. In C. R. Berger & S. H. Chaffee (Eds.), Handbook of
communication science (pp. 847-889). Newbury Park, CA: Sage.
Hage, J. (1972). Techniques and problems of theory construction in sociology. New York: Wiley.
Hannan, M. T. (1971). Aggregation and disaggregation in sociology. Lexington, MA: Lexington Books.
Hauser, R. M. (1970). Context and consex: A cautionary tale. American Journal of Sociology, 75, 645-664.
Hauser, R. M. (1974). Contextual analysis revisited. Sociological Methods and Research, 2, 365-375.
Hauser, R. M., & Mossel, P. A. (1985). Fraternal resemblance in educational attainment and occupational status.
American Journal of Sociology, 91, 650-673.
Hauser, R. M., & Sewell, W. H., & Alwin, D. F. (1976). High school effects on achievement. In W. J. Sewell, R.
M. Hauser, & D. L. Featherman (Eds.), Schooling and achievement in American society (pp. 309-341). New York:
Academic Press.
Hawkins, R. P., Pingree, S., & Adler, I. (1987). Searching for cognitive processes in cultivation effects: Adult and
adolescent samples in the United States and Australia. Human Communication Research, 13, 553-577.
Hyman, H. H., & Singer, E. (1968). Readings in reference group theory and research. New York: Free Press.
Jackson, J. A. (Ed.). (1972). Role. Cambridge: Cambridge University Press.
Jencks, C. S., & Brown, M. D. (1975)/ Effects of high schools on their students. Harvard Educational Review, 45,
273-324.
Johnstone, J. W. C., Slawski, E. J., & Bowman, W. W. (1976). The newspeople: A sociological portrait of
American journalists and their work. Urbana: University of Illinois Press.
Kahneman, D., Slovic, P., & Tversky, A. (Eds.). (1982). Judgment under uncertainty: Heuristics and biases.
Cambridge: Cambridge University Press.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 22 of 24
Katz, E., & Szecsko, T. (Eds.) (1981). Mass media and social change: The Uppsala Apers. Beverly Hills, CA:
Sage.
Knorr-Cetina, K., & Cicourel, A. V. (Eds.). (1981). Advances in social theory and methodology: Toward an
integration of micro- and macro-sociologies. Boston: Routledge & Kegan Paul.
Kosicki, G. M., & McLeod, J. M. (1990). Learning from political news: Effects of media images and information
processing strategies. In S. Kraus (Ed.), Mass communication and political information processing (pp. 69-83)
Hillsdale, NJ: Lawrence-Erlbaum.
Lazarsfeld, P. F., & Menzel, H. (1961). On the relation between individual and collective properties. In A. Etzioni
(Ed.), A sociological reader on complex organizations (2nd ed. pp. 499-516). New York: Holt, Rinehart, &
Winston.
Lemert, J. B. (1981). Does mass communication change public opinion after all? Chicago: Nelson-Hall.
Mason, W. M., Anderson, A. F., & Hayat, N. (1987). Manual for GENMOD. Ann Arbor: Population Studies
Center, University of Michigan.
Mason, W. M., Wong, G. Y., & Entwisle, B. (1983). Contextual analysis through the multilevel linear model. In S.
Leinhardt (Ed.), Sociological methodology (pp. 72-103). San Francisco: Jossey-Bass.
McLeod, J. M., & Blumler, J. G. (1987). The macrosocial level of communication science. In C. R. Berger & S.
H. Chaffee (Eds.), Handbook of communication science (pp. 271-322). Newbury Park, CA: Sage.
McLeod, J. M., & Chaffee, S. H. (1972). The construction of social reality. In J. Tedeschi (Ed.), The social
influence processes (pp. 50-99). Chicago: Aldine.
McLeod, J. M., & Pan, Z. (1989). Getting levels across and crossing levels. American Behavioral Scientist, 33,
199-202.
McLeod, J. M., & Hawley, S. (1964). Professionalization among newsmen. Journalism Quarterly, 41, 529-539.
McLeod, J. M., & Reeves, B. (1981). On the nature of mass media effects. In G. C. Wilhoit & H. de Bocks (Eds.),
Mass communication review yearbook (Vol. 2 pp. 245-282). Beverly Hills, CA: Sage. (Original work published
1980)
Noelle-Neumann, E. (1981). Mass media and social change in developed societies. In E. Katz & T. Szeckso (Eds.),
Mass media and social change (pp. 137-165). Beverly Hills, CA: Sage.
Noelle-Neumann, E. (1984). The spiral of silence: Public opinion--Our social skin. Chicago: University of Chicago
Press.
Oosthoek, H., & van de Eeden, P. (Eds.). (1984). Education from the multilevel perspective: Models, methodology
and empirical findings. New York: Gordon.
Paisley, W. (1984). Communication in the communication sciences. In B. Dervin & M. Voigt (Eds.), Progress in
the communication sciences (Vol. 5, pp. 1-43). Norwood, NJ: Ablex.
Parson, T., & Bales, R. F. (1955). Family, socialization and interaction process. Glencoe IL: Free Press.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 23 of 24
Popper, K. R. (1979). Objective knowledge: An evolutionary approach (rev. ed.). Oxford: Clarendon Press.
(Original work published 1972)
Price, V. (1988). On the public aspects of opinion: Linking levels of analysis in public opinion research.
Communication Research, 15, 659-679.
Przeworski, A., & Teune, H. (1970). The logic of comparative social inquiry. New York: Wiley.
Raudenbush, S., & Bryk, A. S. (1986). A hierarchical model for studying school effects. Sociology of Education,
59, 1-17.
Raudenbush, S. W., & Bryk, A. S. (1988). Methodological advances in analyzing the effects of schools and
classrooms on student learning. In E. Z. Rothkopf (Ed.), Review of research in education (Vol. 15, pp. 423-475).
Washington, DC: American Educational Research Association.
Reeves, B. (1989). Theories about news and theories about cognition: Arguments for a more radical separation.
American Behavioral Scientist, 33, 191-198.
Reeves, B., Thorson, E., & Schleuder, J. (1986). Attention to television: Psychological theories and chronometric
measures. In J. Bryant & D. Zillmann (Eds.), Perspectives on media effects (pp. 51-279). Hillsdale, NJ: Erlbaum.
Robinson, J. P. (1972). Mass communication and information diffusion. In F. G. Kline & P. J. Tichenor (Eds.),
Current perspectives in mass communication research (pp. 71-93). Beverly Hills, CA: Sage.
Rogers, E. M. (1983). Diffusion of innovations (3rd ed.). New York: Free Press. Schoenberg, R. (1972). Strategies
for meaningful comparison. In H. L. Costner (Ed.), Sociological methodology (pp. 1-35). San Francisco: JosseyBass.
Shapiro, M. A. (1991). Memory and decision processes in the construction of social reality. Communication
Research, 18, 3-24.
Shils, E. (1951). The study of primary groups. In D. Lerner & H. Lasswell (Eds.), The policy sciences (pp. 44-69).
Stanford, CA: Stanford University Press.
Sigal, L. (1973). Reporters and officials: The organization and politics of newsmaking. Lexington, MA: D. C.
Health.
Sigelman, L. (1973). Reporting the news: An organizational analysis. American Journal of Sociology, 79, 132-151.
Specht, D. A., & Warren, R. D. (1976). Comparing causal models. In D. R. Heise (Ed.), Sociological methodology
(pp. 46-82). San Francisco: Jossey-Bass.
Stipak, B., & Hensler, C. (1982). Statistical inference in contextual analysis. American Journal of Political
Science, 26, 151-175.
Stoker, T. M. (1985). Aggregation, structural change, and cross-section estimation. Journal of the American
Statistical Association, 80, 720-729.
Tichenor, P. J., Donohue, G. A., & Olien, C. N. (1970). Mass media flow and differential growth in knowledge.
Public Opinion Quarterly, 34, 159-170.
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11
EBSCOhost Full Display
Page 24 of 24
Tichenor, P. J., Donohue, G. A., & Olien, C. N. (1973). Mass communication research: Evolution of a structural
model. Journalism Quarterly, 50, 419-425.
Tichenor, P. J., Donohue, G. A., & Olien, C. N. (1980). Community conflict and the press. Beverly Hills, CA:
Sage.
Tuchman, G. (1978). Making news: A study in the construction of reality. New York: Free Press.
van Daal, J., & Merkies, A.H.Q.M. (1984). Aggregation in economic research: From individual to macro
relations. Dordrecht, Holland: D. Reidel.
Wentworth, W. M. (1980). Context and understanding: An inquiry into socialization theory. New York: Elsevier.
White, D. M. (1950). The "gate keeper": A case study in the selection of news. Journalism Quarterly, 27, 383-390.
Whitney, D. C. (1982). Mass communicator studies: Similarity, difference, and level of analysis. In J. S. Ettema &
D. C. Whitney (Eds.), Individuals in mass media organizations: Creativity and constraints (pp. 241-254). Beverly
Hills, CA: Sage.
~~~~~~~~
By ZHONGDANG PAN AND JACK M. McLEOD
Copyright of Communication Research is the property of Sage Publications Inc. and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.
Source: Communication Research, Apr91, Vol. 18 Issue 2, p140, 34p, 1 diagram.
Item Number: 9707265671
Result 3 of 3
[Go To Full Text] [Tips]
© 2002 EBSCO Publishing. Privacy Policy - Terms of Use
file://C:\Documents%20and%20Settings\zhongdang%20pan\My%20Documents\research\Pan%20pu... 2005-10-11