Information Sufficiency and Risk Communication
Robert J. Griffin
College of Communication
Marquette University
Milwaukee WI 53201-1881
Phone: 414-288-6787 (o); 414-476-7033 (h)
E-mail: [log in to unmask]
Kurt Neuwirth
Department of Communication
University of Cincinnati
Cincinnati OH 45221-0184
Sharon Dunwoody
School of Journalism and Mass Communication
University of Wisconsin-Madison
Madison WI 53706
Presented to the Communication Theory and Methodology Division, Association for
Education in Journalism and Mass Communication, 1998 annual convention,
Baltimore, MD. Address inquiries to Griffin.
Information Sufficiency and Risk Communication
Abstract
Analysis of a survey of two Great Lakes cities develops and tests part of a
model that focuses on characteristics of individuals that might predispose them
to seek and process information about risks in different ways. Support is found
for the model's propositions that information sufficiency (a person's sense of
the amount of information needed to cope with a health risk) is based partially
on affective response to the risk, which is based in part on perceptions of key
characteristics of the risk. Support is less strong for the proposition that
felt normative pressures to possess information may also affect information
sufficiency.
Acknowledgment
This research has been supported in whole or in part by the federal Agency for
Toxic Substances and Disease Registry.
Information Sufficiency and Risk Communication
Introduction
Most current risk communication research and most risk-related public
information programs are concerned with message effects on audiences, assuming
that the "right" message intervention can cause people to change risk-related
behaviors. Yet a number of researchers counsel that, to develop a truly useful
understanding of the role and effects of risk communication in the daily lives
of audience members, researchers and practitioners must pay more attention to
the communication and information-evaluative behaviors of audiences as seekers
and users of information about risks. This audience-based perspective is
supported by the growth of theories in psychology and communication concerned
with the ways that people seek and process information of various kinds, what
motivates such communication activities, and the effects of these processes on
cognition, affect, and behavior.
The purpose of this research program is to develop and test a model, derived
from Eagly and Chaiken's (1993) "Heuristic-Systematic" model, that focuses on
characteristics of individuals that might predispose them to seek and process
information about risks in ways involving more or less effort and critical
thinking. Furthermore, it is expected that persons who analyze risk
information more critically will develop cognitions, attitudes and behaviors in
regard to the risk that are stable, more resistant to change. Since the goal
of many health information campaigns is to convince people to make long-term
changes in behavior, the information processing conditions leading to such
stability are important considerations. The model, originally designed to desc
ribe information seeking and processing about health risks, is being tested
initially by applying it to three hazards involving the Great Lakes. Two of
the hazards indeed involve potential health risks to oneself: risks from
consuming contaminated fish caught in the Great Lakes and risks from consuming
drinking water drawn from the Great Lakes. The third hazard D risks to the
health of the Great Lakes' ecosystem D represents a potential application of
the model to information about risks affecting the environment.
The study has been funded by a four-year grant from the federal Agency for
Toxic Substances and Disease Registry. That grant supports a three-year,
three-wave panel design telephone sample survey in two communities (Milwaukee
and Cleveland) located on the Great Lakes, concentrating primarily on
communication about risks from eating Great Lakes fish. This paper reports a
first-wave test of part of the model. A general overview of the model will be
presented, but more attention will be paid to those variables active in this
analysis.
Background
Most scholars who study risk communication view information use as a
way-station on the road to understanding individuals' reactions to risks in
their environment. That is, research typically employs some operationalization
of exposure/attention to risk information as an independent variable, as a
potential predictor of what individuals know about a risk, how they feel about
it, or what they may do about it.
But inherent in all these studies is the assumption that information "does"
something to individuals. If one can illuminate that causal process, goes the
argument, one can then design message interventions that will cause people to
buckle their seatbelts, recycle, or adopt low-fat diets. This top-down
approach, no matter how well-intentioned, runs counter to suggestions by many
risk perception researchers that risk communication be used to facilitate a
bottom-up process (see, for example, Krimsky & Plough, 1988; National Research
Council, 1989; Juanillo & Scherer, 1995). That is, information providers are
counseled to provide individuals with the types of information they need rather
than giving them only what others with expertise feel they should have. The
bottom-up approach assumes the individual is a reasonable soul who, when it
makes sense to do so, can become engaged intellectually in the risk issue at
hand.
Consistent with this bottom-up scenario, we argue, is the reconfiguration of
information-seeking as a dependent rather than an independent variable.
Instead of asking how messages may influence people, the bottom-up approach
calls for a focus on understanding the evaluative behaviors of the information
user. Thus, we are studying factors that may influence the ways in which
people seek and process risk information.
Why care specifically about the nature of information seeking and processing
as dependent variables? We offer two responses to this question:
First, studies of everyday reasoning strategies suggest that such strategies
can be marred by incompleteness and bias. Perkins et al. (1991), for example,
find that, although individuals engage in what the researchers call "situation
modeling" to solve a real-world problem, the resulting models constructed
frequently invoke only a few common-sense, causal or intentional principles and
only rarely entertain competing arguments. Other scholars (see, for example,
Nisbett et al., 1983, and Griffin & Tversky, 1992) similarly encounter both
flawed reasoning strategies and an enduring tendency to make decisions on the
basis of truncated amounts of information. Given the complicated nature of
risk D and particularly of those social constructions that we call risk
estimates D it makes sense to seek the conditions under which individuals are
more or less systematic in their search for risk information and in their
processing of it.
Second, we feel it is useful to focus on the nature of information seeking and
processing as a primary dependent variable because a body of theory has grown
in both psychology and communication studies to facilitate the use of these
concepts, most specifically Petty and Cacioppo's (1981) elaboration likelihood
model (ELM) and, more recently, Eagly and Chaiken's (1993) heuristic-systematic
model (HSM). Although these models have their roots in persuasion contexts of
communication, Chaiken, Liberman and Eagly (1989) invite researchers to apply a
generic version of HSM to a variety of other contexts in which people "are
exposed to information about themselves, other persons, and events, and have to
make decisions or formulate judgments about these entities" (p. 239).
Both ELM and HSM describe dual forms of human processing of information, one
more superficial (and which people tend to use unless motivated otherwise) and
the other deeper and more effortful. Generally, attitudes developed through
the more intense forms of processing are more stable and longer lasting than
those developed through superficial processing. In HSM terms, people tend to
adopt the form of processing that they use for a given message based on (1)
their capacity to process the information in each manner, and (2) their
motivation to go beyond the more superficial ("heuristic") processing to engage
in "systematic" processing, which can occur along with heuristic processing
(Eagly & Chaiken, 1993).
According to the HSM formulation, a person's desire for sufficiency motivates
systematic processing. The sufficiency principle, state Eagly and Chaiken
(1993), "asserts that people will exert whatever effort is required to attain a
'sufficient' degree of confidence that they have accomplished their processing
goals" (p. 330). For example, the personal relevance of the message topic to
the individual can motivate people to do systematic processing of the message
(Eagly & Chaiken, 1993). Relevance elevates the amount of judgmental confidence
people need to have (the "sufficiency threshold") in their own attitudes (e.g.,
do they square with relevant facts? are they defensible? are they socially
acceptable?) and/or the confidence they need to have in the validity of the
message (Eagly & Chaiken, 1993). In a risk communication experiment, Griffin,
Neuwirth & Dunwoody (1995) found that subjects who were given information that
described a hazard as high risk and severe in its effects were less likely than
other subjects to have their judgments of personal risk from the hazard
affected by stylistic factors in the message. One interpretation is that
relevance might have motivated subjects in the "risky and severe" condition to
process the risk information more systematically and to rely less on heuristics
that might have been influenced by superficial cues in the structure and style
of the message.
By proposing that people need varying levels of confidence in the information
(relevant facts) that they hold about a topic D especially as the basis for
developing their beliefs, attitudes, and behavioral intentions toward their own
preventive health behaviors (see Griffin, Neuwirth & Dunwoody, 1995; Ajzen &
Timko, 1986) D we incorporate, extend and adapt some of the basic principles of
HSM in part of our model of risk information seeking and processing. As part of
our adaptation, we will concentrate some of our efforts on investigating a
variable we will call "information sufficiency" (rather than the term
judgmental confidence as used by Eagly and Chaiken), reflecting the nature of
the variable concerned.
Our study expectations
Figure 1 illustrates the variables of primary concern in our Model of Risk
Information Seeking and Processing, and indicates the strongest paths that we
expect to find among these variables. Specifically, we hypothesize that three
factors D (1) information sufficiency, (2) perceived information gathering
capacity, and (3) relevant channel beliefs D will influence the extent to which
a person will seek out risk information in both routine and non-routine
channels and the extent to which he or she will spend time and effort analyzing
the risk information critically (i.e., systematically rather than merely
heuristically). Consistent with past research (e.g., Gantz, Fitzmaurice &
Fink, 1991; Johnson & Meischke, 1993; Perse & Courtright, 1993; Perse, 1990a,
1990b, 1990c; Griffin, Dunwoody, Zabala & Kamerick, 1994), our model adds to
the heuristic-systematic distinction the dimension of multi- channel informati
on seeking and proposes that the sufficiency principle can also motivate
non-routine seeking of information D that is, active attempts to gather
relevant risk information (e.g., calling the doctor) that go beyond
[--- ??? Graphic Goes Here ---]
Information
Sufficiency
High
Sufficiency
Threshold
Current
Knowledge
Perceived
Knowledge
Low
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
Affective Response
Individual
Characteristics
Relevant
Hazard
Experience
Political Philosophy
Demographic/
Sociocultural
Gender
Ethnicity
Age
Social
Status
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
Perceived
Hazard
Character-
istics
Relevant Channel Beliefs
Informa-
tional
Subjective
Norms
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
Perceived Information Gathering Capacity
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
[--- ??? Graphic Goes Here ---]
Information
Seeking/Processing Behavior
Processing
Seeking
H
S
R
RH
RS
N
NH
NS
Avoidance
Key:
R Routine
N Non Routine
H Heuristic
S Systematic
Processi ng
I ignoring/ avoiding
H heuristi c processing
P active processing
R reflective integration
Seeking
R routine
N non- R
Special Note:
RI=ignoring
NI=avoiding
[--- ??? Graphic Goes Here ---]
habitual or routine sources a given individual might usually use for such
information (e.g., evening newscast). Similarly, people might devote more or
less effort to avoiding risk content if, for example, it produces worries with
which they cannot cope. Some may simply not pay attention, while others might
go out of their way to avoid such information.
Based on Eagly and Chaiken's (1993) motivational factor, we argue that the
size of the gap between information held and that needed will ultimately affect
the information-seeking and processing styles that might be employed by the
individual to learn more about the risk. To incorporate underlying notions of
confidence and relevance, we will measure "information sufficiency" mainly in
terms of the amount of information people say they need in order to deal
adequately with a given risk in their own lives. Similarly, based on Eagly and
Chaiken's (1993) capacity factor, we expect that the relationship between
information sufficiency and information processing effort will be mediated by
an individual's assessment of his or her ability to learn more about the risk
("perceived information gathering capacity"). To this mix we add "relevant
channel beliefs," the individual's everyday beliefs about channels of risk
information (e.g., beliefs that the news media are biased), adapting work on
media images by Kosicki and McLeod (1990). We would expect these beliefs to
affect information seeking and processing in combination with self-perceived
capacity and motivation. Based on Eagly and Chaiken (1993), we expect that
persons who analyze risk information more critically will ultimately develop
attitudes and even behaviors in regard to the risk that are more resistant to
change. [Relevant behavioral measures, operationalizing Ajzen's (1988) Theory
of Planned Behavior, are included in the second and third waves of the survey
so assessment can be made of stability or change in those variables over time.]
The model proposes that information sufficiency, perceived information
gathering capacity, and relevant channel beliefs, in turn, are affected to one
extent or another by (1) affective response to the risk (e.g., worry, anger),
(2) subjective norms about knowledge and information-gathering about the risk,
(3) perceived hazard characteristics, and (4) selected characteristics of the
individual. For example, an affective response to the risk, such as worry,
could influence one's judgment of the amount of information one feels one needs
to have in order to cope with the risk (e.g., to take effective action).
Similarly, the one's perception that valued others expect one to keep on top of
information about the risk (the subjective normative component) could also
affect one's judgment about how much information one needs to have about the
risk. Either or both variables could affect information sufficiency.
Various attributes of an individual (e.g., social status, past experience
with a hazard) are expected to affect his or her views of the characteristics
of a given risk (e.g., level of personal risk, the extent to which risk
management institutions are acting in a trustworthy way, the extent to which
one can protect oneself from the hazard, the extent to which a risk threatens a
personally-held value) which in turn should influence his or her affective
response (e.g., level of personal worry or anger about that risk).
Concurrently, many of those same individual attributes are expected to
influence the individual's assessment of the amount of information he or she
already possesses about the risk, perceived information gathering capacity, and
perhaps perceived social norms for information gathering and holding.
The next section of this proposal offers a brief conceptual scaffolding for
each stage of the model predicting to information sufficiency (the difference
between knowledge currently held and that needed to cope with a risk), which is
the primary focus of this analysis. A separate analysis, not reported here,
will examine information seeking/processing behavior and its more proximate
predictors (i.e., the balance of the model). Similarly, analysis of the
relationship of risk information seeking and processing to risk-related
behavior will await the gathering of data on an array of behaviorally-related
variables in subsequent waves of this survey.
Filling In the Model
To construct this model, we have borrowed concepts from several disciplinary
groups, among them risk perception researchers, psychologists who study
attitude formation and change, and communication scholars. The front of the
model is indebted largely to the first group, while the back of the model owes
much of its inspiration to the second. Communication research D particularly
work that focuses on risk communication concepts D infuses the whole model and
serves both as a kind of glue and as a tool for customizing concepts to better
fit communication needs.
Individual characteristics. Our model begins with a set of variables
representing the demographic/ sociocultural background of respondents as well
as their political philosophy and experience with the hazard.
Demographic/Sociocultural variables are found in most studies of risk
perception to help absorb variance in the dependent variable of interest.
Typically, some subset of those constructs does account for statistically
significant amounts of variance, although they rarely engage the lion's share.
Typical of their contribution is that found in an extensive survey of public
attitudes in the United States about the management of such risks as nuclear
power, handguns, auto travel and industrial chemical production (Gould et al.,
1988). The researchers found that the addition of variables such as age,
education, gender and political philosophy did account for variance in
respondents' judgments about the acceptability of technology safety regulations
(the dependent variable) but accounted for increases, on average, of only 3
percentage points.
Use of these kinds of demographic variables is further problematized by their
atheoretical employment. For example, studies frequently demonstrate a
difference in risk perception between men and women; typically, women perceive
a higher level of risk than do men and signal greater levels of personal worry
about that risk than men. How should one give meaning to such a gender
difference? Psychologist Paul Slovic found not a gender difference in risk
perception but a difference between white males and everyone else (i.e., white
women and minority men and women). The gender difference, he suggests, may be
masking a difference in one's sense of control over one's environment. White
males may simply feel more control than others (Slovic, 1994). Such findings
suggest that we need to select and interpret individual characteristics
carefully. We use the following for this study:
y Gender has been such a systematic predictor of variance in the past that we
will employ it here. But we also utilize a measure of perception of
personal
control to control for that suggested dimension of gender meaning.
y Ethnicity similarly has been found to relate to risk perception, perhaps
because of differences in (1) perceptions of personal control and/or (2)
other
factors such as differences in levels of exposure to (or experience with)
risks
(Vaughn & Nordenstam, 1991). As controls, we include measures of relevant
hazard experience.
y Age has typically D but not always D had a positive relationship with risk
perception. That is, the older one is the more fearful one is. However,
this
reaction may be risk-specific, e.g., concern about longer-term risks might
decrease with age (e.g., Fischer et al, 1991; Griffin, Dunwoody, Dybro &
Zabala, 1994).
y Socioeconomic status also has an uneven history of prediction. However,
status, like ethnicity, could affect exposure to various risks. And
education
D one of the major dimensions of SES D tends to be such an important
predictor
of an individual's ability to seek, process and retain information that
"knowledge gaps" can develop between higher and lower status segments of
society (Griffin, 1990; Olien, Donohue & Tichenor, 1983). Moreso than for
lower status persons, higher status persons may even be expected by their
friends and relatives to be keep up to date on the topic. One upshot is
that
blue-collar workers, for example, probably have a greater likelihood of
personal harm from workplace hazards such as hazardous chemicals, but might
not
have the education that would enhance their ability to seek and process
risk-relevant information.
Political philosophy (liberalism-conservatism) could affect acceptance of
hazard reduction regulations (Gould et al., 1988) and, more generally, trust
in risk management institutions.
Relevant hazard experience represents personal experience with a risk. Grunig
(1983) observes that people tend to apply a referent criterion from past
situations (e.g., a cognition that can serve as a decision rule) as an initial
guide when deciding how to think, behave or communicate in a new situation, and
that they develop new criteria only if the referent criterion fails to work in
outlining the new situation. In our model, we propose that a major referent
criterion that people might apply to a hazard is their past experience with the
same hazard or with what they might nominate as a related hazard, including
their experiences with associated preventive behaviors. Personal experience
with a risk has served as a strong predictor, in many studies, of a number of
dependent variables, among them individuals' perceptions of the risk of
recurrence of that hazard in future (Griffin, Dunwoody, Zabala & Kamerick,
1994), their perception of the extent of risk posed by a related hazard
(Griffin, Dunwoody, Dybro & Zabala, 1994), and information seeking (Lenz, 1984;
Johnson and Meischke, 1993).
Informational subjective norms. Informational subjective norms represent
perceived social normative influences motivating the desire for information
sufficiency. A person's perceptions that relevant others believe he or she
should (or should not) perform a particular behavior can be at least as
important a predictor of behavior as the person's own cognitions and attitudes
about performing that behavior (Ajzen and Fishbein, 1980; Fishbein & Ajzen,
1975) and a person's own sense of control or capacity in performing that
behavior (Ajzen, 1988). In this case, we are examining the extent to which
such subjective norms might affect the person's behavior of risk information
seeking and processing, and by extrapolation, to the knowledge a person
believes he or she would be expected to hold about the risk. We expect that
these subjective norms could be affected by various individual characteristics,
such as social status as noted above, and that they might in turn influence
information gathering and processing via effects on information sufficiency.
Hazard characteristics and affective response. Most studies of information use
and processing employ some measure of issue salience, involvement or relevance
as a predictor. As noted earlier, an issue judged to be personally relevant or
important is more likely to generate systematic processing efforts than is an
issue relegated to lower levels of importance or relevance (see, for example,
Donohew, 1990; Petty & Cacioppo, 1981; Eagly & Chaiken, 1993). Constructs like
salience or relevance, although clearly valuable to information studies, may be
overly broad for use in studies of information and risk perception. Risks by
definition have a very particular negative valence, and it is reasonable to
assume that individuals' cognitive evaluations of a risk produce a response
closely related to such constructs as relevance and salience but more specific
and negatively valenced than either. We propose two stages of our model to
capture this process: (1) perceived hazard characteristics, which include
measures related to a sense of personal risk, trust in risk management,
personal control over the risk, and a sense of threat to personal values; and
(2) affective responses to the risk (e.g., personal worry). We expect that the
set of cognitive variables that comprise perceived hazard characteristics might
combine in various ways to influence affective response and, like the relevance
variable, will eventually affect the intensity of information seeking and
processing.
Perceived hazard characteristics. Risk perception researchers have
convincingly argued that risk judgments are multidimensional, that is, that
they take into account more than just estimates of the likelihood of coming to
harm from exposure to a risk.
Perhaps the most visible proponent of multidimensional risk judgments is
psychologist Paul Slovic, whose surveys of individuals' perceptions of risks
and benefits led to a set of empirically defined groupings of risk
characteristics and, ultimately, to groupings of risks themselves. He refers
to his approach and the theoretical framework within which it is embedded as
"the psychometric paradigm" (Slovic, 1992). Slovic originally demonstrated the
utility of some 15 different risk characteristics but found that they could be
grouped into a smaller set of higher-order characteristics that reflected the
degree to which a risk is understood and the degree to which it evokes a
"feeling of dread" (p.121). Gregory and Mendelsohn (1993), in a reanalysis of
some of Slovic and colleagues' original survey data, found that six variables
accounted for significant variance in respondents' assessments of perceived
risk and of dread: (1) An estimate of the number of deaths that would take
place "if next year is average"; (2) a judgment of the potential for
catastrophic outcome; (3) an assessment of the immediacy of the effect; (4) an
assessment of the economic benefits of the risk; (5) an assessment of the
pleasure benefits of the risk; and (6) the estimated impact of the risk on
future generations. In applying these variables to health risks in our model,
we adjust variables 1 through 5 to represent perceptions of risk, seriousness,
immediacy of onset, and benefits for the self rather than for society at
large. We also supplement them with three other precursors:
y (7) Personal control D A self-evaluation of the level of personal control that
the individual has over susceptibility to harm from the hazard (see, for
example, Weinstein, 1993; Schwarzer, 1992; Ajzen & Timko, 1986; Rogers, 1985);
y (8) Trust in risk management D A judgment of the amount of trust the
respondent has in the ability of others to prevent the respondent from coming to
harm. This variable reflects a growing sentiment among risk perception
researchers (see, for example, MacGregor et al., 1994; Flynn et al., 1992;
Slovic, 1992; Wynne, 1992; Kasperson et. al., 1987) that individuals' judgments
of how much they can trust "responsible" agencies and institutions may play a
major role in the kinds of risk perceptions that we attempt to measure.
y (9) Perceived threats to personal values D Although few would argue with the
proposition that values undergird much if not all of human activity, the
relationship between values and risk judgment remains largely unexplored.
Theoretically, Earle and Cvetkovich (1995) argue that judgments of risk increase
when values are seen as being challenged or threatened.
Affect, trust, and risk judgments. Although a link between an emotional
response and risk judgments has a well-established history in the "fear appeals"
literature (Dillard, 1994; Witte, 1994), affect's importance to researchers
working outside of this tradition has been recognized only recently (Johnson
and Tversky, 1983; Dunwoody and Neuwirth, 1991; Peters and Slovic, 1996). Quite
typically, terms such as fear, dread, worry, and outrage are bandied about with
scant attention paid to their theoretical status. More recently, Dillard and
colleagues 1996) provide evidence that messages about a hazard designed to
elicit fear can also rouse a variety of additional emotional responses.
Beliefs based on social trust also influence risk judgments. Indeed,
researchers have identified trust as a key mediating factor in circumstances
requiring collective action (Slovic, 1993; Earle and Cvetkovich, 1994). For the
individual, social trust serves as a cognitive heuristic that decreases the
complexities of social life to workable levels. Upon encountering potentially
threatening events or conditions, people often make risk judgments based on
social trust, a expectation that assigns to others the responsibility for
working on some necessary task (Slovic, 1993; Earle and Cvetkovich, 1995). In
essence, social trust entails a trade-off between internal and external
reactions to risk. For example, increased social trust in an institutional
arrangement such as municipal water utility means that a person will experience
less personal worry and perceived vulnerability to the hazard of water borne
parasites. Thus, social trust is seen as mutable, subject to change and
modification as new information is received. When circumstances require
cooperative social action, social trust derives from perceptions of
institutional arrangements and social processes designed manage hazards (Rayner
and Cantor, 1987; Rayner, 1992; Earle and Cvetkovich, 1995). The concept of
social trust, in our view, extends beyond institutions formally charged with
managing environmental risks to scientific institutions which provide the
knowledge and technology to reduce risks and organizations such as the mass
media that provide audiences with relevant information about hazards (Kosicki
and McLeod, 1990). In our model, trust in media is considered part of "relevant
channel beliefs" while trust in risk management and scientific institutions is
part of "perceived hazard characteristics."
Affect and information processing. A growing body of research indicates that
emotional reactions and moods influence both heuristic and systematic
processing. The main research finding is that positive moods and emotions are
associated with heuristic information processing, while negative affective
states are linked to systematic processing (Batra and Stayman, 1990; Kuykendall
and Keating, 1990; Bohner and Apostolidou, 1994; Bohner, Chaiken and Hundyadi,
1994). However, extremely negative affective (fear) states appear to result in
greater heuristic processing or avoidance (Jepson and Chaiken, 1990).
The influence of different emotions and their intensity in the formation of
risk judgments and subsequent information seeking and processing is a relatively
unexamined area, especially in field settings. In his project, we are
emphasizing worry, anger, and uncertainty about a health risk. Consistent with
previous findings (Griffin, Dunwoody, Zabala and Kamerick, 1994), we would
expect worry to motivate information seeking and processing about a risk, even
if indirectly, more than would the cognitive components of risk perception.
Worry is considered here to be a manifestation of anxiety, distinguished by a
recurrent negative affective state provoked by a future hazard. Anger is
considered to be related to an attempt by the person to reassert control over
the risk. Uncertainty, while not considered to be a "classical" emotional state,
is often associated with negative emotional states such as fear, anxiety, and
anger and is associated with dimensions such as the unknowability of outcome or
consequences and a perceived loss of control. It also reflects aspects of
judgmental confidence (or lack thereof) underlying the Eagly and Chaiken (1993)
Heuristic-Systematic Model. In general, investigating the role of positive and
negative affective responses to risk (Dillard et al., 1996; Ekman and Davidson,
1994; Plutchik and Kellerman, 1989; Frijda, 1986; Scherer, 1984) should greatly
enhance our understanding of these connections.
Information sufficiency. Our model proposes that subjective information
gathering norms and affective response to a risk (e.g., personal worry) will
affect the confidence one wants to have in one's knowledge about the risk
(information sufficiency threshold), in particular about how to behave (i.e.,
protect oneself) in the face of the risk, which should manifest itself in a
judgment of the amount of information the respondent feels he or she needs
(e.g., to take effective action). Further reflecting the approach of the HSM
model (Eagly & Chaiken, 1993: 330-332), we propose that more effortful
information seeking and processing will be motivated when the sufficiency
threshold is higher than the amount of such information the respondent feels he
or she currently has ("current knowledge").
Various individual characteristics, especially relevant hazard experience and
social status, should affect what one actually knows about some hazards, as
might some of the perceived hazard characteristics. We specify none of these
relationships in the model, except for the expectation that social status in
particular should account for a significant amount of variance in the perceived
amount of information one currently has about a hazard.
Research Questions and Hypotheses
Our analysis will concentrate on examining the precursors to information
sufficiency as illustrated in Figure 1, in particular examining relationships
among informational subjective norms, institutional trust, risk judgment,
affective response, and information sufficiency.
The first research question (RQ1) is: What are the relationships among
informational subjective norms, institutional trust, risk judgment, affective
response, and information sufficiency? We expect that:
H1a. Institutional trust will be negatively related to "risk
judgment" (perceived risk x seriousness).
H1b. Institutional trust will be negatively related to affective
response.
H1c. Personal control will be negatively related to risk
judgment.
H1d. Personal control will be negatively related to affective
response.
H1e. Risk judgment will be positively related to affective
response.
H1f. Affective response will be positively related to
information sufficiency.
H1g. Informational subjective norms will be positively related
to information sufficiency.
The second research question (RQ2) is: What are the relationships
between individual characteristics and risk judgment, affective response,
and information sufficiency? We expect that:
H2. Education (as an indicator of social status) will be
positively related to self- reported current knowledge.
Method
The purpose of the study is to test the Model of Risk Information
Seeking and Processing by applying it across different risks and across
different communities. The current analysis examines only part of the model
and applies it to two potential health risks related to the Great Lakes:
eating Great Lakes fish and drinking tap water drawn from the Great Lakes.
Another part of the study, not reported here, will seek to find whether the
model, originally developed to describe individuals' responses to personal
health risks, can also be applied to reactions individuals might have to
threats to the health of the Great Lakes ecosystem.
Great Lakes fish consumption is the health risk of primary concern in
this study. Fish in the Great Lakes, like fish from other waters, can
contain various chemicals, most notably polychlorinated biphenyls (PCBs).
Human consumption of PCB-laden fish is a suspected cancer risk and has been
associated with developmental problems in infants whose mothers had
regularly eaten PCB-contaminated fish. Every year for the past quarter
century, states surrounding the Great Lakes, including Wisconsin and Ohio,
have issued advisories that warn people to avoid or limit consumption of
certain sizes and varieties of fish and that suggest ways to prepare the
fish to reduce exposure to chemical contamination. This information is
available in pamphlets, sometimes in news media, and potentially via other
sources as well.
The second health risk of concern, potential hazards lurking in
municipal drinking water, is of course not limited to the Great Lakes.
Municipal drinking water can contain substances such as chemicals and lead,
as well as organisms that occasionally slip past municipal water treatment
systems. In recent years, the United States has seen an increase in major
outbreaks of various waterborne illnesses. Probably the most salient
outbreak took place in 1993 in the Great Lakes community of Milwaukee,
Wisconsin. A tiny parasite, cryptosporidium, entered the city drinking water
from Lake Michigan and produced the largest recorded outbreak of waterborne
disease in the nation's history D as well as national headlines. Milwaukee
has since installed special monitoring equipment and is installing special
treatment equipment. Nonetheless, cryptosporidium is difficult and
expensive to detect and purge from municipal water systems and could
potentially strike somewhere again. Thus, we are concentrating on an
examination of people's responses to potential hazards from waterborne
parasites.
The two health risks also offer different risk scenarios. Although
both contain an element of uncertainty, as does nearly any risk, the
uncertainties of the health impacts of eating PCB-contaminated fish are far
greater than those of the hazards of drinking contaminated water. Health
hazards from eating PCB-laden fish tend to be longer term in development and
effects and relatively serious, while the hazards from drinking water
infested with parasites such as cryptosporidium or giardia tend to be shorter
term in development and effects and, for most people, less serious (e.g.,
usually a bout with cramps and diarrhea). Thus, our model can be tested for
now under these two conditions.
Sampling and Interviewing
Two communities on the Great Lakes D Milwaukee, Wisconsin, on Lake
Michigan and Cleveland, Ohio, on Lake Erie D were chosen as the research
sites. These two medium-sized American cities, each on a different lake,
have diverse populations that draw their drinking water from the lakes and
have relatively ready access to commercially caught and sport-caught fish
from the lakes. The two communities might respond similarly to risks from
eating Great Lakes fish but somewhat differently to risks from drinking the
water because of Milwaukee's recent bout with cryptosporidiosis.
From October 1996 to March 1997, the Wisconsin Survey Research
Laboratory, a professional research organization associated with the
University of Wisconsin-Extension, conducted a sample survey by telephone of
1,123 adult residents of the two metropolitan areas (579 in Milwaukee and
544 in Cleveland). The combined response rate was 55.2% (61.3% in Milwaukee
and 50% in Cleveland). Residences were contacted by random-digit-dialing
(RDD) and respondents were chosen randomly within households.
Interviews took approximately 20 minutes apiece. Applicable human
subjects and informed consent practices were followed throughout. The
interviews constituted the first wave of a three-wave, panel-design study to
be conducted over three years. (Waves two and three will also include a
series of behavioral questions not included in the first wave.)
At the start of the interview, respondents were assigned to one of
three "paths" through the questionnaire. One path was comprised of
questions dealing with the fish consumption risks, one path concerned the
tap water risks, and the third path was composed of questions about risks to
the Great Lakes ecosystem. All questions in the tap water path and most
questions in the ecosystem path were identical in construction to questions
in the fish path. This parallel construction was designed to allow meta-tes
ting of the model by combining data across risks as much as possible. When
respondents were to be presented with a series of items to be answered on
the same kind of scale (e.g., five point, Likert-type, agreement scale), the
starting point in the series was chosen randomly.
Since applying the model to fish consumption risks was our primary
goal, the interviewers' first questions were designed to net respondents for
whom eating Great Lakes fish was a relevant personal matter. Respondents
were assigned to the fish path if they had eaten a meal of Great Lakes fish
that year or if they had made a decision to avoid these fish specifically
because of health concerns. In all, 634 respondents (326 in Milwaukee, 308
in Cleveland) were assigned to the fish path. The balance of respondents
were randomly assigned to the other two paths in the questionnaire D in
particular, 252 to the water path (137 in Milwaukee and 115 in Cleveland).
We assumed that drinking local tap water was also relevant personal matter
for residents in the two communities. Two potential analytical drawbacks of
this approach are (1) the relatively small subsample size in the water path
and (2) the fact that respondents in that path are unlike the majority of
respondents who eat Great Lakes fish or avoid them for health reasons.
Questionnaire Development
To aid in the development of the questionnaire, the Wisconsin Survey
Research Laboratory conducted four focus groups with a random sample of
Milwaukee area residents in the spring of 1996. The focus groups were
designed to gather information about various components of the model that
needed some exploratory investigation, including affective responses and
information needs in regard to Lake Michigan and fish contamination.
Intelligence from the focus group analyses was used to help prepare draft
questionnaires distributed to a convenience sample of 301 students at the
researchers' three universities in the summer of 1996. These questionnaires
operationalized all of the model components across a variety of risks,
including risks from consuming contaminated fish and drinking water, risks
from exposure to the sun, and risks to the aquatic ecosystem. Item and
scale analyses, conducted primarily by combining the data across risks and
universities, yielded the measures to be used in the actual survey. The Wisc
onsin Survey Research Laboratory then conducted three telephone pretests of
the resulting questionnaire with random samples of Milwaukee and Cleveland
residents before actual interviews began in late October 1996. Budget
constraints required the exclusion of some variables, including measures of
personal benefits related to the risks and perceived threats to personal
values.
Measurement
Individual Characteristics. Three sets of individual characteristics
are used as control variables in this analysis: (1)
demographic/sociocultural variables, (2) political philosophy, and (3)
hazard experience variables.
Demographic/Sociocultural variables include age, gender, annual
income, education (measured in terms of the highest grade or year of school
completed), and ethnicity (whether the person is a member of a minority
group). Reliability of a social status index, to be comprised of income and
education, was only marginal (Cronbach's alpha=.49), so instead both income
and education are used as individual predictors in this analysis. The
respondent's community D Milwaukee (coded as 1) or Cleveland (2) D is also
included as a control variable in analyses involving both communities.
Political Philosophy was measured on a single, five-point scale
ranging from "liberal" (coded as 1) to "conservative" (5) and is represented
by the variable "political conservatism" in the tables.
Hazard Experience variables include whether the person reports ever
having been ill from a waterborne parasite such as cryptosporidium or
giardia and whether the person reports ever having been ill from food
poisoning. ("No" or "don't know" responses were coded as 0, "yes" as 1).
Both are risks from ingestion that persons in the surveyed communities
should be able to relate to somewhat readily. Persons who had become ill
from a waterborne parasite might judge tap water parasite risks as more
probable or more serious. Contracting cryptosporidiosis can even affect a
person's judgments of risk from drinking tap water containing trace amounts
of lead, a longer-term hazard that is related to cryptosporidiosis only by
virtue of sometimes sharing the same vehicle of transmission (Griffin,
Dunwoody, Dybro and Zabala, 1994). Therefore it is possible that contracting
illness from a waterborne parasite might, by extension, also affect
judgments of risk from eating fish. Similarly, having contracted food
poisoning might affect judgments of risk from consuming Great Lakes tap
water and fish.
Informational Subjective Norms. Informational subjective norms were
measured by a single item similar to Ajzen's (1988) formulation for
assessing subjective norms for any behavior:
"People who are important to me would expect me to stay on top
of information about...[the risk from eating Lake (Michigan) (Erie)
fish] [the risk from drinking Lake (Michigan) (Erie) tap water]."
Specific wording, as illustrated, depended on the respondent's community
and path through the questionnaire. Responses were recorded on a
five-point, Likert-type, agree-disagree scale. Greater agreement yielded
higher scale values.
Perceived Hazard Characteristics. Perceived Hazard Characteristics were
limited in this analysis to three essential variables: (1) personal control
over the risk, (2) institutional trust, and (3) risk judgment.
Personal control was measured by a single item responded to on a
five-point, Likert-type, agree- disagree scale:
"In my life, it would be easy for me to avoid becoming ill from
[eating contaminated Lake (Michigan) (Erie) fish] [drinking Lake
(Michigan) (Erie) tap water that was contaminated]."
Greater agreement yielded higher scale values. The item was adapted from
one of Ajzen's (1988) measures of perceived behavioral control. It
represents a summary judgment the individual makes about both personal
efficacy (i.e., whether one can perform a health-protective action) and
response efficacy (i.e., whether the action to be taken is efficacious in
preventing illness) in a health risk context (see Bandura, 1977).
Institutional trust is a four-item, summated index (Cronbach's alpha=.73)
of trust in governmental and scientific institutions to protect one from the
specific health risk. Risk judgments are based in part on judgments of
social trust when circumstances require cooperative social action (Rayner &
Cantor, 1987; Krimsky & Golding, 1992; Rayner, 1992; Earle & Cvetovich,
1995), such as would be the case with environmental health risks from
contaminated fish or tap water. Therefore, the following items comprised
the measure of institutional trust:
"Government officials care about the health and safety of people
like me."
"Eventually science will find a way to overcome most risks to
human health."
"Government is doing a competent job of protecting people's
health from risks related to [eating contaminated Lake (Michigan)
(Erie) fish] [drinking Lake (Michigan) (Erie) tap water]."
"I trust government to protect me from risks related to [eating
contaminated Lake (Michigan) (Erie) fish] [drinking Lake (Michigan)
(Erie) tap water]."
Respondents used five-point, Likert-type, agree-disagree scales to react to
these items. Higher scale values represent higher levels of institutional
trust.
Risk judgment is the product of two measures, one representing the
subjective probability of becoming ill from exposure to the hazard and the
other representing the perceived seriousness of the illness.1 The measure of
subjective probability was:
How likely are you to become ill in the future from [eating
contaminated fish caught in Lake (Michigan) (Erie)] [drinking tap
water drawn from Lake (Michigan) (Erie)]? Please use a scale from
zero to 10, where zero means that you would have absolutely no chance
whatsoever of becoming ill, and 10 means that you are certain to.
The measure of perceived seriousness was:
If you were to become ill from [eating contaminated Lake
(Michigan) (Erie) fish] [drinking Lake (Michigan) (Erie) tap water],
how serious do you think this illness would be? Please use a scale
of zero to 10, where zero means not serious at all and 10 means it
would be as serious as it can possibly be.
Affective Response. Affective response is the sum of three variables
(Cronbach's alpha=.85) representing the amount of worry, anger, and
uncertainty respondents felt toward the risk.2 A sample item is:
Now we'd like to know your feelings about [contaminated fish]
[the risk of contaminated tap water]. Please use a number from zero
to ten, where zero means you have "none of this feeling" and ten
means you have "a lot of this feeling." When you think about the
possible health risks posed to you from [eating Lake (Michigan)
(Erie) fish] [drinking Lake (Michigan) (Erie) tap water], how much
worry do you feel?
Respondents were also asked how much anger and how much uncertainty they
felt.
Information Sufficiency. Information sufficiency was derived by juxtaposing
in the analysis two self-report variables: (1) current knowledge about the
risk and (2) the information sufficiency threshold.3 (Please see the section
on analysis for a description of the use of these variables.)
Current knowledge was measured as follows:
Now, we would like you to rate your knowledge about this risk.
Please use a scale of zero to 100, where zero means knowing nothing
and 100 means knowing everything you could possibly know about this
topic. Using this scale, how much do you think you currently know
about the risk from [eating Lake (Michigan) (Erie) fish] [drinking
Lake (Michigan) (Erie) tap water]?
Sufficiency threshold was measured as follows:
Think of that same scale again. This time, we would like you to
estimate how much knowledge you would need to deal adequately with
the possible risk from [eating Lake (Michigan) (Erie) fish] [drinking
Lake (Michigan) (Erie) tap water] in your own life. Of course, you
might feel you need the same, more, or possibly even less,
information about this topic. Using a scale of zero to 100, how much
information would be sufficient for you, that is, good enough for
your purposes?
Analysis
The Statistical Package for the Social Sciences was used to perform a
series of hierarchical multiple regression analyses in path-analytic format.
4 The analyses concentrated on examining the relationships among
informational subjective norms, risk judgment, affective response, and
information sufficiency.
Current knowledge was entered as the first block in regressing sufficiency
threshold so that variables entered later were predicting to the difference
("information sufficiency") between current knowledge and sufficiency
threshold.5 To make results comparable across analyses, current knowledge
was also entered as the first block in regressions of risk judgment and
affective response. The first analysis in the series regressed risk judgment
on blocks of (1) current knowledge, (2) individual characteristics used as
control variables, (3) information subjective norms, and (4) the other
"perceived hazard characteristics" variables D institutional trust and
personal control. The second analyses regressed affective response on the
same blocks plus risk judgment. The third analysis regressed sufficiency
threshold on the above blocks plus affective response. Listwise deletion of
cases with missing data was used throughout.6
The primary test of these relationships was conducted by combining data
across both risks and both communities (N=801). To perform internal
replications, parallel analyses were performed by combining data about both
health risks in each community individually (n=427 in Milwaukee, 374 in
Cleveland), by examining data about each health risk individually in the two
communities combined (n=585 for the fish path, 216 for tap water path), and
finally by examining data from each health risk in each community (n=305 for
the fish path in Milwaukee, 208 for the fish path in Cleveland, 122 for the
tap water path in Milwaukee, and 94 for the tap water path in Cleveland).
Results and Discussion
The first research question (RQ1) was: What are the relationships among
informational subjective norms, institutional trust, risk judgment,
affective response, and information sufficiency?7
Trust with risk judgment, affective response. The first two hypotheses
proposed that institutional trust would be negatively related to risk
judgment (H1a) and affective response (H1b). When data are combined across
communities and across risks (see the first set of three data columns in
Table 1), institutional trust does indeed bear negative relationships with
risk judgment (beta = -.19, p<.001) and affective response (beta = -.19,
p<.001). Also consistent with the model, the direct path of influence of
institutional trust does not extend past affective response to sufficiency
threshold. This same
pattern holds true for the combined health risks within each community (see
the second and third sets of columns in Table 1), for each risk across both
communities (see the second and third sets of columns in Table 2), and for
fish risks alone in each community (see the second and third sets of columns
in Table 3). The pattern holds true for tap water risks in the Milwaukee
subsample (second set of columns in Table 4), but not in Cleveland (third
set of columns in Table 4), although the small subsample size (n=94) might
have rendered insignificant the relationship between institutional trust and
affective response (beta = -.15, ns). In general, H1a and H1b are supported
in all but one comparison.
Personal control with risk judgment, affective response. The next two
hypotheses proposed that personal control will be negatively related to risk
judgment (H1c) and affective response (H1d). There are no statistically
significant relationships between personal control and affective response in
any of the comparisons across all tables. Thus H1d is not supported. There
are no statistically significant relationships between personal control and
risk judgments in any of the comparisons except for tap water risks in the
Milwaukee subsample (Table 4) where lower levels of personal control are
indeed associated with higher risk judgments (beta = -.16, p<.05). Thus, H1c
is generally not supported.
It is possible that the single-item "ease" measure being used is not
sensitive to variance in feelings of personal control except when a
respondent can readily visualize the hazard, as might be expected in
Table 1:
Regression of Risk Judgment, Affective Response and Information Sufficiency by
Community, Combining Both Health Risks
Both Health Risks:
Both Communities
Both Health Risks:
Milwaukee
Both Health Risks:
Cleveland
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Current Knowledge
.04
-.03
.27***
.02
.02
.30***
.06
-.09*
.27***
R2 change
.00
.00
.07***
.00
.00
.08***
.00
.00
.07***
Individual Characteristics
Political Conservatism
-.04
-.07*
-.03
-.03
-.07
-.04
-.05
-.08
-.01
Demographic/
Sociocultural
Community
-.03
-.05
-.01
Education
-.11**
-.08*
.00
-.12*
-.05
-.03
-.10
-.11*
.02
Income
-.05
-.05
-.02
-.06
-.04
-.04
-.04
-.06
.00
Age
.06
-.04
-.10**
.07
-.04
-.10*
.03
-.05
-.09
Female Gender
.07
.09**
.08*
.07
.10*
.09*
.06
.08
.06
Racial Minority
.06
.15***
.05
.01
.13**
.13**
.14**
.16***
-.02
Hazard Experience
Food Poisoning
.03
.04
.05
-.02
.00
.05
.09
.08
.05
Waterborne Parasite
.13***
.03
-.04
.15**
.04
-.01
.06
-.03
-.12*
R2 change
.06***
.13***
.05***
.06*
.10***
.08***
.07**
.15***
.05*
Informational
Subjective Norms
.09**
.25***
.15***
.11*
.26***
.10*
.07
.25***
.20***
R2 change
.01*
.08***
.04***
.01*
.09***
.03***
.00
.07***
.07***
Perceived Hazard Characteristics
Institutional Trust
-.19***
-.19***
.02
-.20***
-.24***
-.03
-.19***
-.13***
.06
Personal Control
-.04
.00
-.06
-.07
.00
-.07
.02
.00
-.04
R2 change
.04***
.06**
.01*
.04***
.09***
.01*
.03***
.04***
.00
Risk Judgment
.34***
-.05
.33***
-.02
.34***
-.07
R2 change
.10***
.00
.10***
.00
.10***
.00
Affective Response
.26***
.21***
.31***
R2 change
.04***
.03***
.06***
Multiple R
.33***
.61***
.47***
.34***
.62***
.48***
.32***
.61***
.50***
Adjusted R2
.10
.36
.20
.09
.37
.20
.08
.35
.22
N(n)
801
801
801
427
427
427
374
374
374
Key: * p .05 **p .01 ***p .001
Table 2:
Regression of Risk Judgment, Affective Response and Information Sufficiency by
Health Risk, Combining Both Communities
Both Health Risks:
Both Communities
Fish Risks:
Both Communities
Tap Water Risks:
Both Communities
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Current Knowledge
.04
-.03
.27***
.04
-.08*
.26***
-.02
.13*
.35***
R2 change
.00
.00
.07***
.00
.00
.06***
.00
.00
.11***
Individual Characteristics
Political Conservatism
-.04
-.07*
-.03
-.07
-.07*
-.04
.04
-.16**
-.02
Demographic/
Sociocultural
Community
-.03
-.05
-.01
-.01
-.05
-.03
-.13
-.08
.05
Education
-.11**
-.08*
.00
-.12*
-.07
.02
-.08
-.15*
-.06
Income
-.05
-.05
-.02
-.06
-.05
.01
-.06
-.06
-.05
Age
.06
-.04
-.10**
.07
-.02
-.08*
.03
-.14*
-.16*
Female Gender
.07
.09**
.08*
.06
.11**
.07
.09
.05
.06
Racial Minority
.06
.15***
.05
.08
.17***
.06
-.01
.07
.04
Hazard Experience
Food Poisoning
.03
.04
.05
.02
.05
.07
.04
-.02
.01
Waterborne Parasite
.13***
.03
-.04
.10*
.05
-.07
.25***
-.10
-.02
R2 change
.06***
.13***
.05***
.06***
.15***
.05***
.14***
.15***
.08*
Informational
Subjective Norms
.09**
.25***
.15***
.08*
.26***
.15***
.10
.17**
.17**
R2 change
.01*
.08***
.04***
.01
.07***
.04***
.01
.03**
.04**
Perceived Hazard Characteristics
Institutional Trust
-.19***
-.19***
.02
-.16***
-.14***
.02
-.19**
-.23***
.00
Personal Control
-.04
.00
-.06
-.05
-.03
-.05
-.12
-.01
-.03
R2 change
.04***
.06**
.01*
.03***
.04***
.00
.05**
.09***
.01
Risk Judgment
.34***
-.05
.32***
-.09*
.36***
.13
R2 change
.10***
.00
.10***
.00
.10***
.03**
Affective Response
.26***
.29***
.17*
R2 change
.04***
.05***
.02**
Multiple R
.33***
.61***
.47***
.31***
.61***
.46***
.45***
.61***
.53***
Adjusted R2
.10
.36
.20
.07
.35
.19
.15
.33
.23
N(n)
801
801
801
585
585
585
216
216
216
Key: * p .05 **p .01 ***p .001
Table 3:
Regression of Risk Judgment, Affective Response and Information Sufficiency for
Fish Risks, by Community
Fish Risks:
Both Communities
Fish Risks:
Milwaukee
Fish Risks:
Cleveland
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Current Knowledge
.04
-.08*
.26***
.05
.00
.28***
.03
-.17***
.24***
R2 change
.00
.00
.06***
.00
.00
.07***
.00
.02*
.05***
Individual Characteristics
Political Conservatism
-.07
-.07*
-.04
-.08
-.09
-.04
-.07
-.04
-.04
Demographic/
Sociocultural
Community
-.01
-.05
-.03
Education
-.12*
-.07
.02
-.14*
-.02
-.01
-.07
-.11*
.04
Income
-.06
-.05
.01
-.09
-.01
-.01
-.05
-.07
-.01
Age
.07
-.02
-.08*
.07
.03
-.05
.06
-.06
-.10
Female Gender
.06
.11**
.07
.09
.10*
.11*
.03
.12*
.03
Racial Minority
.08
.17***
.06
-.02
.15**
.13*
.21***
.20***
.01
Hazard Experience
Food Poisoning
.02
.05
.07
-.04
.00
.09
.09
.10
.05
Waterborne Parasite
.10*
.05
-.07
.11
.07
-.05
.07
-.01
-.11*
R2 change
.06***
.15***
.05***
.08**
.12***
.07***
.08*
.20***
.05
Informational
Subjective Norms
.08*
.26***
.15***
.12*
.28***
.07
.03
.25***
.23***
R2 change
.01
.07***
.04***
.01*
.10***
.02*
.00
.06***
.08***
Perceived Hazard Characteristics
Institutional Trust
-.16***
-.14***
.02
-.14*
-.16**
.00
-.19**
-.11*
.05
Personal Control
-.05
-.03
-.05
-.07
-.05
-.08
-.01
-.01
-.03
R2 change
.03***
.04***
.00
.02*
.05***
.01
.04**
.03**
.00
Risk Judgment
.32***
-.09*
.35***
-.06
.29***
-.10
R2 change
.10***
.00
.11***
.00
.08***
.00
Affective Response
.29***
.26***
.31***
R2 change
.05***
.04***
.06***
Multiple R
.31***
.61***
.46***
.34***
.61***
.46***
.34***
.62***
.50***
Adjusted R2
.07
.35
.19
.08
.35
.18
.08
.35
. 21
N(n)
585
585
585
305
305
305
280
280
280
Key: * p .05 **p .01 ***p .001
Table 4:
Regression of Risk Judgment, Affective Response and Information Sufficiency for
Tap Water Risks, by Community
Tap Water Risks:
Both Communities
Tap Water Risks:
Milwaukee
Tap Water Risks:
Cleveland
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Risk Judg-ment
Affec-tive
Re-sponse
Suffi-
ciency Thresh- old
Current Knowledge
-.02
.13*
.35***
-.08
.16*
.40***
.09
.09
.29**
R2 change
.00
.00
.11***
.00
.00
.10***
.00
.00
.11***
Individual Characteristics
Political Conservatism
.04
-.16**
-.02
.12
-.02
-.02
-.15
-.31**
.08
Demographic/
Sociocultural
Community
-.13
-.08
.05
Education
-.08
-.15*
-.06
.09
-.12
-.12
-.25*
-.20
.06
Income
-.06
-.06
-.05
-.05
-.10
-.09
-.09
-.01
.00
Age
.03
-.14*
-.16*
.06
-.28**
-.25*
-.09
-.01
.00
Female Gender
.09
.05
.06
.04
.10
.06
.18
-.07
.10
Racial Minority
-.01
.07
.04
.14
.07
.12
-.23*
.11
-.07
Hazard Experience
Food Poisoning
.04
-.02
.01
.02
-.11
-.07
.03
.03
.06
Waterborne Parasite
.25***
-.10
-.02
.34***
-.06
.03
-.03
-.10
-.15
R2 change
.14***
.15***
.08*
.16*
.21***
.16**
.15
.18*
.07
Informational
Subjective Norms
.10
.17**
.17**
.08
.15*
.18*
.10
.24*
.12
R2 change
.01
.03**
.04**
.01
.03*
.04*
.01
.06*
.04
Perceived Hazard Characteristics
Institutional Trust
-.19**
-.23***
.00
-.28**
-.31***
-.11
-.04
-.15
.06
Personal Control
-.12
-.01
-.03
-.16*
-.04
-.05
-.08
-.03
-.04
R2 change
.05**
.09***
.01
.11***
.15***
.03
.01
.03
.00
Risk Judgment
.36***
.13
.28**
.12
.39***
.08
R2 change
.10***
.03**
.06**
.01
.13***
.03
Affective Response
.17*
.03
.28*
R2 change
.02**
.00
.05*
Multiple R
.45***
.61***
.53***
.52***
.66***
.59***
.41
.63***
.55**
Adjusted R2
.15
.33
.23
.19
.37
.26
.05
.31
.18
N(n)
216
216
216
122
122
122
94
94
94
Key: * p .05 **p .01 ***p .001
Milwaukee in the aftermath of their battle with cryptosporidiosis. The measure
might also be too broad in attempting to encompass both personal efficacy and
response efficacy. This item will be supplemented in the second and third waves
of the survey with more comprehensive measures of perceived behavioral control,
to represent personal efficacy, and of behavioral beliefs about the
effectiveness of actions persons can take to protect themselves from risk, to
represent response efficacy, based on Ajzen's (1988) formulations.
Risk judgment with affective response. Hypothesis H1e proposed that risk
judgment will be positively related to affective response. The hypothesis is
readily supported when data are combined across communities and across risks
(beta = .36, p<.001 in Table 1) and the relationship remains approximately equal
in magnitude in all other comparisons in all tables. Consistent with the path
model, risk judgment is much more strongly related to affective response than to
information sufficiency (see column under "sufficiency threshold" in all tables)
in a pattern that is quite similar across all comparisons.
Affective response, informational subjective norms with information
sufficiency. The next two hypotheses reflect the model's proposition that
affective response (H1f) and/or informational subjective norms (H1g) will be
positively related to information sufficiency. When data are combined across
communities and across risks, information sufficiency does indeed bear positive
relationships with informational subjective norms (beta = .15, p<.001 in Table
1) and affective response (beta = .26, p<.001 in Table 1). With the exception
of current knowledge (which was entered first as part of the process of creating
the information sufficiency variable), both variables bear stronger
relationships with information sufficiency than any other variable in that
analysis. That result is consistent with the model's expectations.
Across comparisons in the other tables, the relationship of affective response
to information sufficiency is generally more robust and consistent than is the
relationship of informational subjective norms. The relationship of affective
response to information sufficiency remains significant in all but one
comparison D tap water risks in Milwaukee as shown in Table 4. Similarly, the
relationship of informational subjective norms information sufficiency remains
significant in all but two comparisons D fish risks in Milwaukee as shown in
Table 3 and tap water risks in Cleveland as shown in Table 4.
Not anticipated by the model, however, are the somewhat stronger relationships
that informational subjective norms usually has with affective response (beta =
.25, p<.001, for the combined overall data in Table 1 and significant in all
comparisons across all tables). There is also a small relationship between
informational subjective norms and risk judgment (beta = .09, p<.01, for the
combined overall data in Table 1) which can be traced primarily to the
relationship between those variables as related to fish risks in Milwaukee
(Table 3). It is possible that the wording of the item that measures
informational subjective norms ("people who are important to me would expect me
to stay on top of information about...")may contain the kind of ambiguity which
could lead to these patterns.8 A better phrasing would be "...think that I
should stay on top of information about..." This item will be altered for the
second and third waves of the survey, and a companion item will be added,
representing Ajzen's (1988) formulation for measuring overall subjective norms.
In general, H1f (the relationship of affective response) is supported in
all but one comparison. Support for H1g (the relationship of informational
subjective norms) is somewhat tentative, perhaps because of the upshot of what
may be faulty wording. Most notably, however, and consistent with
expectations, whenever one of these two variables does not bear a significant
relationship with information sufficiency, the other one does. (Analysis of the
reasons for risk-by-community differences in prediction of information
sufficiency by affective response or informational subjective norms goes beyond
the scope of this paper.)
Individual characteristics. The second research question (RQ2) is: What are the
relationships between individual characteristics and risk judgment, affective
response, and information sufficiency?
As illustrated in Table 1 and developed in Tables 2 through 4, the separate
variables representing individual characteristics tend to have usually small and
sporadic direct relationships with risk judgment, affective response, and
sufficiency threshold. This result is consistent with expectations and with the
model which proposed that other variables should intervene. A few noteworthy
patterns do reveal themselves, however.
Women and minorities, as illustrated in Tables 1 and 3, do tend to respond
affectively to risks, in particular from eating Great Lakes fish, a little more
than do men and non-minorities (beta = .11, p<.01, for female gender and beta =
.17, p<.001, for racial minority in Table 3). This relationship appears not to
be mitigated by feelings of personal control. T-tests of the relationship
between personal control and these two individual characteristics (not shown)
are non-significant. Men do self-report higher levels of current knowledge
about the risks, however (t853=4.30, p<.001). As noted earlier, however, a more
comprehensive measure of personal control might reveal patterns that our
single-item measure is not responding to.
Hazard experience with becoming ill from a waterborne parasite does indeed
affect tap water risk judgments in Milwaukee (beta = .34, p<.001, in Table 4),
as might certainly be expected. It also bears a slight relationship with fish
risk judgments (beta = .10, p<.05) when data are combined across both
communities (Table 3). It does not appear, however, that this kind of
experience with a parasite provides a strong referent criterion for dealing with
risks from eating fish. Likewise, experience with food poisoning is unrelated
to risk judgment, affective response, or sufficiency threshold for either fish
or tap water risks. However, those who had suffered food poisoning do report
slightly higher levels of current
knowledge about fish risks (t619=2.83, p<.01) and tap water risks (t232=2.21,
p<.05) and those who have had a bout with a parasite do sense higher levels of
current knowledge about tap water risks (t232=3.97, p<.001). Thus, at least
food poisoning might bear some perceived similarity to fish and tap water risks,
but at a rather basic cognitive level.
Age and conservatism reveal two of the stronger relationships with affective
response to tap water risks (Table 4) among the individual characteristics
variables. Political conservatives in Cleveland (beta = -.31, p<.01) and older
people in Milwaukee (beta=-.28, p<.01) are less likely to respond affectively to
tap water risks. Older Milwaukeeans are also less likely to desire more
information to deal with tap water risks (beta= -.25, p<.05). These patterns
may be based on local conditions that are beyond the analysis scope of this
paper.
Education, as an indicator of social status, is related positively to
self-reported current knowledge (r=.15, p<.001), as hypothesized from the model
(H2) and the knowledge-gap model. However, the relationship is not as strong as
might be expected. Income bears a similar but weaker relationship with
current knowledge (r=.11, p<.01) and is generally inactive as a variable in
Tables 1-4.
Conclusion
The results indicate that much can be learned from an audience-based approach
to understanding risk communication.
In this analysis, the strongest results seem to suggest a path of influence
from lower institutional trust to higher risk judgments to stronger affective
responses to the risk (specifically, worry, anger, and uncertainty) to a
perception that more information is needed to allow oneself to deal adequately
with a health risk. Institutional trust also has some apparent influence on
how a person responds affectively to a risk. That seems appropriate, since
institutional trust itself has overtones both cognitive and affective.
Consistent with the model, information sufficiency was found to be affected in
all comparisons by affective response and/or informational subjective norms.
However, informational subjective norms did not relate to information
sufficiency as strongly as might be expected from the model and related
unexpectedly to affective response and, to a much lesser degree, risk judgment.
It is likely that the problem is one of measurement rather than theory, and so
an improved measure will be adopted and tested out in the future. Since
subjective norms can predict other forms of behavior (Ajzen, 1988), they might
be valuable predictors of communication behavior as well.
Personal control was relatively inactive as a variable in this analysis
although it is normally considered to be an important predictor of risk
judgments and risk-related behaviors. Again, it is probable that the problem is
one of measurement rather than theory and improved measurement will be tested
out in the future.
Further research. Along with testing improved measures of subjective norms and
personal control, the next stages of research in this program will (1) test
hypotheses derived from the remainder of the model (i.e., direct predictors of
risk information seeking and processing behaviors) and (2) test hypotheses about
the relationship of risk information seeking and processing to the performance
and maintenance of preventive behaviors. The latter will rely a lot on Ajzen's
(1988) Theory of Planned Behavior as adapted to the performance of preventive
behaviors and operationalized in a survey research setting. The program will
also examine whether the model of risk information seeking and processing can be
applied to perceptions of risk not to oneself but to the ecosystem. Over time,
research in this program will also seek to include variables that were omitted
from this analysis due to time and budget constraints, such as measures of
personal benefits related to risks and perceived threats to personal values, and
apply the model to a variety of health risks.
Reprise. The means of testing the model through combining data across risks
and communities and then by internal replications across risks and communities
was very valuable. The model predicted results rather well in regard to risks
from eating Great Lakes fish but less so in regard to drinking Great Lakes tap
water. The tap water risk analysis, however, might have been hampered by
relatively small (therefore unstable) subsample sizes and by the somewhat
unrepresentative nature of respondents in that path.
In general, however, the Model of Risk Information Seeking and Processing, at
least as tested so far, seems to offer promise as a research and theoretic tool
to guide inquiry.
Notes
1. Although subjective probability and perceived seriousness need not be
correlated, they are in this study (r=.50, p<.001), yielding Cronbach's alpha of
.66 for the risk judgment scale. A third measure, representing the perceived
immediacy of the onset of the illness, was not used in the development of the
risk judgment scale because it had low correlations with other variables in the
analysis. It was uncorrelated with subjective probability and perceived
seriousness. These conditions might be unique to these communities and this set
of risks, however.
2. Future analyses will parse out these three affective responses.
3. Although current knowledge and sufficiency threshold need not be correlated,
they are in this study (r=.27, p<.001). To help validate the information
sufficiency measure, we ran first-order partial correlations between sufficiency
threshold and four related measures in the questionnaire, with control for
current knowledge. Results reinforce the construct validity of information
sufficiency. Specifically, sufficiency threshold correlates positively with the
item "When the topic of risks from [eating Lake (Michigan) (Erie) fish]
[drinking Lake (Michigan) (Erie) tap water] comes up, I try to learn more about
it" (partial r=.26, p<.001); positively with the item "When it comes to the
risks from [eating Lake (Michigan) (Erie) fish] [drinking Lake (Michigan) (Erie)
tap water], I'm likely to go out of my way to get more information" (partial
r=.30, p<.001); negatively with the item "What I know about this topic is
enough" (partial r=-.35, p<.001); and negatively with the item "Gathering a lot
of information on the risks from [eating Lake (Michigan) (Erie) fish] [drinking
Lake (Michigan) (Erie) tap water] is a waste of time" (partial r=-.31, p<.001).
Parrott et al. (1998) found that self-reported (perceived) current knowledge of
how to adapt to the health risks from skin cancer correlated positively with
objectively tested knowledge of those procedures and with adopting preventive
and detection behaviors.
4. The single-item measure of perceived impact of the risk on future generations
was removed from the regression analysis because of multicollinearity problems.
5. We used the regression approach to be consistent with Cohen and Cohen (1975),
who consider it superior to calculating change/difference scores.
6. To conserve N, missing data in the control variables (the block of
"individual characteristics" variables) were replaced with a suitable measure of
central tendency. Otherwise, missing data were not replaced.
7. There is no significant relationship between institutional trust and
informational subjective norms R = - .02, ns), nor was one expected.
8. The wording "expect me to" could be interpreted in a behaviorally predictive
rather than normative way by a number of people. Those with stronger affective
responses to the risk might presume that others close to them would simply
anticipate (not prescribe) that they would keep on top of information about the
risk.
References
Ajzen, I. (1988) Attitudes, Personality, and Behavior. Milton Keynes, UK: Open
University Press.
Ajzen, I, and M. Fishbein (1980) Understanding Attitudes and Predicting Social
Behavior. Englewood Cliffs, NJ: Prentice Hall.
Ajzen, I. & C. Timko (1986) Correspondence between health attitudes and
behavior.
Journal of Basic and Applied Social Psychology 7: 259-276.
Bandura, A. (1986) Social Foundations of Thought and Action. Englewood Cliffs,
NJ:
Prentice Hall.
Bandura, A. (1977) Self efficacy: Toward a unifying theory of behavioral change.
Psychological Review 84: 191- 215.
Batra, R., & D.M. Stayman (1990) The role of mood in advertising effectiveness.
Journal of Consumer Research 17: 203-214.
Bohner, G., & W. Apostolidou. (1994) Mood and persuasion: Independent effects
of
affect before and after message processing. Journal of Social Psychology
134(5):
707-709.
Bohner, G., S. Chaiken, & P. Hundyadi (1994) The role of mood and message
ambiguity
in the interplay of heuristic and systematic processing. European Journal of
Social
Psychology 24: 207-221.
Chaiken, S., A. Liberman & A.H. Eagly (1989) Heuristic and systematic processing
within and beyond the persuasion context. In J.S. Vleman and J.A. Bargh,
eds.,
Unintended Thought. New York: Guilford, 212- 252.
Cohen, J., & P. Cohen (1975) Applied Multiple Regression/Correlation Analysis
for
the Behavioral Sciences. New York: Wiley and Sons.
Dillard, J. P. (1994) Rethinking the study of fear appeals: An emotional
perspective. Communication Theory 4(4), 295-323.
Dillard, J. P., C.A. Plotnick, L.C. Godbold, V.S. Freimuth, & T. Edgar (1996)
The
multiple affective outcomes of AIDs PSAs: Fear appeals do more than scare
people.
Communication Research 23(1), 44-72.
Dillard, J. P., & B.A. Wilson (1993) Communication and affect. Communication
Research 20(5):637- 646.
Donohew, L. (1990) Public health campaigns: Individual message strategies. In
E.B.
Ray and L. Donohew, eds., Communication and Health: Systems and Applications.
Hillsdale, NJ: Lawrence Erlbaum, 136-152.
Dunwoody, S. & K. Neuwirth (1991) Coming to terms with the impact of
communication
on scientific and technological risk judgments. In L. Wilkins and P.
Patterson,
eds., Risky Business: Communicating Issues of Science, Risk, and Public
Policy.
Westport, CT: Greenwood Press, 11-30.
Eagly, A.H. & S. Chaiken (1993) The Psychology of Attitudes. Fort Worth:
Harcourt
Brace Jovanovich.
Earle, T. C., & G. Cvetkovich (1994) Risk communication: The social
construction
of meaning and trust. In B. Brehmer & N. Sahlin, eds., Future Risks and Risk
Management. Dordrecht, Netherlands: Kluwer Academic Publishers. 141-181.
Earle, T. C., & G. Cvetkovich (1995) Social Trust: Toward a Cosmopolitan
Society.
Westport, CT: Praeger.
Ekman, P. & R.J. Davidson (1994) The Nature of Emotions: Fundamental Questions.
New
York, NY: Oxford University Press.
Fischer, G.W., M.G. Morgan, B. Fischoff, I. Nair & L.B. Lave (1991) What risks
are
people concerned about? Risk Analysis 11(2): 303-314.
Fishbein, M. & I. Ajzen (1975) Belief, Attitude, Intention, and Behavior: An
Introduction to Theory and Research. Reading, MA: Addison-Wesley.
Flynn, J., W. Burns, C.K. Mertz & P. Slovic (1992) Trust as a determinant of
opposition to a high-level radioactive waste respository: An analysis of a
structural model. Risk Analysis 12(3): 417-429.
Frijda, N. H. (1986) The emotions. Cambridge: Cambridge University Press.
Gantz, W., M. Fitzmaurice & E. Fink (1991) Assessing the active component in
information seeking. Journalism Quarterly 68(4): 630-637.
Gould, L.C., G.T. Gardner, D.R. DeLuca, A. R. Tiemann, L.W. Doob & J.A .J.
Stolwijk
(1988) Perceptions of Technological Risks and Benefits. New York: Russell Sage
Foundation.
Gregory, R. & R. Mendelsohn (1993) Perceived risk, dread, and benefits. Risk
Analysis 13(3): 259-264.
Griffin, D. & A. Tversky (1992) The weighing of evidence and the determinants of
confidence. Cognitive Psychology 24: 411-435.
Griffin, R.J. (1990) Energy in the eighties: Education, communication, and the
knowledge gap. Journalism Quarterly 67(3), 554-566.
Griffin, R.J., S. Dunwoody, F. Zabala & M. Kamerick (1994) Public reliance on
risk
communication channels in the wake of a cryptosporidium outbreak. Presented to
the
Society for Risk Analysis, annual convention, Baltimore MD. December.
Griffin, R.J., S. Dunwoody, T. Dybro & F. Zabala (1994) The relationship of
communication to risk perceptions and preventive behavior related to lead in
drinking water. Presented to the Association for Education in Journalism and
Mass
Communication, annual convention, Atlanta GA. August.
Griffin, R.J., K. Neuwirth & S. Dunwoody (1995) Using the theory of reasoned
action
to examine the impact of health risk messages. In B.R. Burleson, ed.,
Communication
Yearbook 18. Thousand Oaks, CA: Sage, 201-228.
Grunig, J.E. (1983) Communication behaviors and attitudes of environmental
publics:
Two studies. Journalism Monographs 81, March.
Jepson, C., & S. Chaiken, S. (1990) Chronic issue-specific fear inhibits
systematic
processing of persuasive communications. Journal of Social Behavior and
Personality 5(2): 61-84.
Johnson, E. J. & A. Tversky (1983) Affect, generalization and the perception of
risk. Journal of Personality and Social Psychology 45(1): 20-31.
Johnson, J.D., & H. Meischke (1993) A comprehensive model of cancer-related
information seeking applied to magazines. Human Communication Research 19(3):
343-367.
Juanillo Jr., N.I. & C.W. Scherer (1995) Attaining a state of informed
judgments:
Toward a dialectical discourse on risk. In B.R. Burleson, ed., Communication
Yearbook 18. Thousand Oaks, CA: Sage, 278-299.
Kasperson, R., S. Klaidman, V. Houk, E. Silbergeld, E.H. Blair, F. Lynn, P.
Sandman, A. Alm & T. Vacor (1987) Trust and credibility: The central issue? In
J.C.
Davies, V.T. Covello and F.W. Allen, eds., Risk Communication. Washington, DC:
The
Conservation Foundation, 43-62.
Kosicki, G.M., & J.M. McLeod (1990) Learning from political news: Effects of
media
images and information processing strategies. In S. Kraus, ed., Mass
Communication
and Political Information Processing. Hillsdale, NJ: Erlbaum: 69-83.
Krimsky, S., & D. Golding. (1992) Social Theories of Risk. Westport, CT:
Praeger.
Krimsky, S. & A. Plough (1988) Environmental Hazards: Communicating Risks as a
Social Process. Dover, MA: Auburn House.
Kuykendall, D., & J.P. Keating (1990) Mood and Persuasion: Evidence for the
differential influence of positive and negative states. Psychology & Marketing
7(1): 1-9.
Lenz, E.R. (1984) Information seeking: A component of client decisions and
health
behavior. Advances in Nursing Science 6: 59-72.
MacGregor, D., P. Slovic, R.G. Mason, J. Detweiler, S.E. Binney & B. Dodd (1994)
Perceived risks of radioactive waste transport through Oregon: Results of a
statewide survey. Risk Analysis 14(1): 5-14.
National Research Council (1989) Improving Risk Communication. Washington, DC:
National Academy Press.
Nisbett, R.E., D.H. Krnatz, S. Jepson & Z. Kunda (1983) The use of statistical
heuristics in everyday inductive reasoning. Psychological Review 90(4):
339-363.
Olien, C., G.A. Donohue & P.J. Tichenor (1983) Structure, communication, and
social
power: Evolution of the knowledge gap hypothesis. In E. Wartella, D.C.
Whitney,
and S. Windahl, eds., Mass Communication Review Yearbook, Vol. 4. Beverly
Hills,
CA: Sage, 455-461.
Parrott, R., J. Monahan, S. Ainsworth & C. Steiner (1998) Communicating to
farmers
about skin cancer: The behavior adaptation model. Human Communication Research
24(3): 386-409.
Perkins, D.N., M. Farady & B. Bushey (1991) Everyday reasoning and the roots of
intelligence. In J. Voss, D.N. Perkins and J. Segal, eds., Informal Reasoning
and
Education. Hillsdale, NJ: Lawrence Erlbaum, 83-105.
Perse, E.M. (1990a) Audience selectivity and involvement in the newer media
environment. Communication Research 17(5): 675-697.
Perse, E.M. (1990b) Media involvement and local news effects. Journal of
Broadcasting & Electronic Media 34(1): 17-36.
Perse, E.M. (1990c) Involvement with local television news: Cognitive and
emotional
dimensions. Human Communication Research 16(4): 556-581.
Perse, E.M., & J.A. Courtright (1993) Normative images of communication media:
Mass
and interpersonal channels in the new media environment. Human Communication
Research 19(4): 485-503.
Peters, E. & P. Slovic (1996) The role of affect and worldviews as orienting
dispositions in the perception and acceptance of nuclear power. Journal of
Applied
Social Psychology 26: 1427-1453.
Petty, R.E. & J.T. Cacioppo (1981) Attitudes and Persuasion: Classic and
Contemporary Approaches. Dubuque, Iowa: W.C. Brown.
Plutchik, R. & H. Kellerman (1989) The Measurement of Emotions. San Diego, CA:
Academic Press.
Rayner, S. (1992) Cultural theory and risk analysis. In S. Krimsky & D.
Golding,
eds., Social Theories of Risk. Westport, CT: Praeger, 83-114.
Rayner, S. & R. Cantor (1987) How fair is safe enough? The cultural approach to
societal technology choice. Risk Analysis 7: 3-9.
Rogers, R.W. (1985) Attitude change and information integration in fear appeals.
Psychological Reports 56:179- 182.
Scherer, K. R. (1984) On the nature and function of emotion: A component process
approach. In K. R. Scherer & P. Eckman, eds., Approaches to Emotion.
Hillsdale, New
Jersey: Lawrence Erlbaum Associates, 293-317.
Schwarzer, R. (1992) Self-efficacy in the adoption and maintenance of health
behaviors: Theoretical approaches and a new model. In R. Schwarzer, ed.,
Self-Efficacy: Thought Control of Action. Washington, DC: Hemisphere, 217-243.
Slovic, Paul. (1987) Perception of risk. Science 36:280-285.
Slovic, P. (1992) Perception of risk: Reflections on the Psychometric Paradigm.
In
S. Krimsky and D. Golding, eds., Social Theories of Risk. Westport, CT:
Praeger,
117-152.
Slovic, P. (1994) Scientific versus public perceptions of environmental risks:
Where is the reality? Presentation at the 160th national meeting of the
American
Association for the Advancement of Science, San Francisco CA. February.
Slovic, P. (1993) Perceived risk, trust, and democracy. Risk Analysis 13(6):
675-682.
Tichenor, P., G.A. Donohue, & C. Olien (1970) Mass media flow and differential
growth in knowledge. Public Opinion Quarterly 34: 159-170.
Vaughn, E. & B. Nordenstam (1991) The perception of environmental risks among
ethnically diverse groups. Journal of Cross-Cultural Psychology 22(1): 29-60.
Weinstein, N.D. (1993) Testing four competing theories of health-protective
behavior. Health Psychology 12(4): 324-333.
Witte, K. (1994) Fear control and danger control: A test of the extended
parallel
process model (EPPM). Communication Monographs 61: 113-134.
Wynne, B. (1992) Risk and social learning: Reification to engagement. In S.
Krimsky
and D. Golding, eds., Social Theories of Risk. Westport, CT: Praeger, 275-297.
|
