Risk Analysis, Vol. 26, No. 3, 2006 DOI: 10.1111/j.1539-6924.2006.00773.

The Role of the Affect and Availability Heuristics

in Risk Communication

Carmen Keller,1 ∗ Michael Siegrist,2 and Heinz Gutscher1

Results of past research suggest that affect plays an important role in risk perception. Because
affect may also increase the availability of risks, affect and availability are closely related
concepts. Three studies tested the hypothesis that evoking negative affect (fear), either through
past experience or through experimental manipulation, results in greater perceived risk. The
present research focused on perception of flooding risk. Study 1 and Study 2 showed that
participants who received risk information concerning a longer time period (e.g., 30 years)
perceived more danger compared with participants who received risk information for one year.
Study 2 showed that the interpretation of risk information was influenced by participants’ own
experiences with flooding. In Study 3, affect was experimentally manipulated. After looking at
photographs depicting houses in a flooded region, participants perceived greater risk compared
with participants in a control group. Taken together, the results of these three studies suggest
that affect is important for successful risk communication. Results of the present research are
in line with the affect heuristic proposed by Slovic and colleagues.

KEY WORDS: Affect heuristic; availability heuristic; flood risk; risk communication

1. INTRODUCTION years was relatively low. Europe suffered large eco-

nomic losses from flood disasters in the past decade,
The way in which information about hazards is
however (Linnerooth-Bayer & Amendola, 2003). In
presented influences how the hazards will be per-
order to reduce costs caused by flooding, the pro-
ceived. The question of how lay people should be in-
motion of residents’ responsibility for taking precau-
formed about the magnitude of risks is, therefore, an
tionary measures will become more important in the
important question in the field of risk management
future. A first step in this direction is to gain peo-
(Fischhoff, 1995). Many of the risks we face have low
ple’s attention on flood risks by increasing their risk
probabilities. Communicating about low-probability
perception.
risks is a big challenge, however.
In the present research, the affect heuristic pro-
The present work focuses on communication con-
posed by Slovic and colleagues (Finucane et al., 2000;
cerning flooding risks. For most residents, this is a low-
Slovic et al., 2002, 2004) is used as a framework. Work
probability hazard. In Europe, the death toll in recent
by these researchers suggests that strong emotional
experiences with hazards may be important for in-
1 Department of Psychology, Social Psychology, University of creasing perceived risks. However, the affect heuristic
Zurich, Plattenstrasse 14, 8032 Zurich, Switzerland. is linked to the availability heuristic by Tversky and
2 Swiss Federal Institute of Technology, Zurich, Switzerland.
∗ Address correspondence to Carmen Keller, Department of Psy-
Kahneman (1982). Slovic et al. (2004) suggested that
chology, Social Psychology, University of Zurich, Plattenstrasse
the availability heuristic might not only work through
14, 8032 Zurich, Switzerland; tel.: +41 44 634 21 15; fax: +41 44 ease of recall or imaginability, but also because re-
634 49 31; [email protected]. membered images are connected with affect.

631 0272-4332/06/0100-0631$22.00/1 
C 2006 Society for Risk Analysis
632 Keller, Siegrist, and Gutscher

We conducted two studies to test the hypothesis mating the probability of a hazard. Slovic et al. (2004)
that evoking negative affect (fear) results in an in- used this framework to explain why people are sen-
creased level of perceived risks. In the third study, sitive to different forms of risk communication. Fre-
we tested the hypothesis that availability of images quency formats produce affect-laden imagery, leading
related to negative affect increases perceived risks. participants to rate a disease that kills 1,286 people out
of 10,000 as being more dangerous than a disease that
kills 24.14% of the population (Yamagishi, 1997).
1.1. Perception of Low-Risk Probabilities
In past research, the availability heuristic has also
Low probabilities are difficult for people to un- been proposed as an explanation for biases in proba-
derstand (Camerer & Kunreuther, 1989). Accord- bility judgments (Tversky & Kahneman, 1982). Ac-
ing to prospect theory (Kahneman & Tversky, 1979), cording to this heuristic, people use the ease with
small probabilities are often underestimated. Vari- which examples of a hazard can be brought to mind as
ous approaches for communicating low-probability a cue for estimating the probability of a hazard. Slovic
risks have been proposed and empirically tested. Risk et al. (2004) suggested that the availability heuristic
comparisons is one such approach (Freudenburg & might work because remembered images are tagged
Rursch, 1994; Johnson, 2003, 2004; Kunreuther et al., with affect, proposing that the availability and affect
2001; Roth et al., 1990; Slovic et al., 1990). Here, the heuristics are closely connected.
risks associated with new hazards are compared with The experiential system uses past experience to
risks that are more familiar to the public. Past research assess risks (Slovic et al., 2004). Research suggests
has shown that judgment under uncertainty improves that past experience is an important factor influenc-
when problems were formulated in terms of frequen- ing people’s perception of hazards (Weinstein, 1989).
cies rather than probabilities (Gigerenzer & Hoffrage, Jackson (1981) found that past experience with earth-
1995). The time period for which a risk is given also quakes influenced the adoption of precautions. Af-
has an influence on how a particular risk is perceived ter a damaging event has occurred, precautions are
(Linville et al., 1993; Shaklee & Fischhoff, 1990; Slovic taken more readily. Similar findings have been re-
et al., 1978; Svenson, 1984). The study by Slovic et al. ported for the purchase of flood insurance. The ex-
(1978) showed that presenting the risks faced during perience of damage was the most important factor
a lifetime of driving induced more people to wear seat that led to the purchase of insurance (Baumann &
belts compared with the presentation of the risks in- Sims, 1978; Zaleskiewicz et al., 2002). O’Connor and
volved in a single trip. The study by Rottenstreich colleagues (2005) found that managers of community
and Hsee (2001) indicated that affect-rich outcomes water systems who experienced a flood emergency in
resulted in more pronounced overweighting of small the last five years perceived a higher flood risk. In
probabilities than affect-poor outcomes. Thus, in the other words, they expected to experience more prob-
case of affect-laden events, small probabilities may lems from flood in the next decade than managers
not be underestimated. The reviewed research sug- who did not experience problems from flood in the
gests that the manner in which risks are presented can last five years. In a similar vein, Siegrist and Gutscher
either decrease or increase the level of risks perceived. (2005) found that past experience with flooding was
the most important factor in predicting people’s risk
perception.
1.2. The Affect Heuristic
Past experiences with a natural hazard are asso-
In recent years, several authors have suggested ciated with negative affect, and that might increase
that affect might play an important role in risk per- the level of risks perceived. However, it should be
ception. Loewenstein and colleagues introduced the emphasized that hazards are not always associated
“risk as feelings” model (Loewenstein et al., 2001). with negative affect. Especially with voluntary haz-
Similarly, Slovic and colleagues proposed the affect ards (e.g., smoking, skiing), people might associate
heuristic (Peters et al., 2004; Slovic et al., 2002, 2004). positive emotions.
In this latter approach, two modes of thinking are dis-
tinguished, the experiential system and the analytical
1.3. Rationale of the Present Studies
system. The analytical system relies on probabilities,
logical reasoning, and evidence. The experiential sys- In risk communication research, a number of dif-
tem relies on images, metaphors, and narratives. Peo- ferent ways of presenting information about risks and
ple use the affect related to a hazard as a cue for esti- probabilities have been examined. Utilizing the affect
Affect, Availability, and Risk Communication 633

heuristic as a framework, three experiments were de- of being involved in an automobile accident in a single
signed to test the hypothesis that utilizing risk com- trip is miniscule. However, the probability of a fatal
munication formats that produce more affect-laden accident within a 50-year period of driving must not
imagery would induce a higher level of perceived risks be neglected. Participants who received information
than risk communication formats that evoke no affect. about a 50-year period favored more strong, manda-
Furthermore, it was postulated that evoking negative tory protection than participants exposed to the risk
affect, either through past experience or through ex- of a single trip (Slovic et al., 1978). This result can
perimental manipulation, results in greater perceived be interpreted within the framework of the affect
risks. Because such affect may also lead to an in- heuristic. Small probabilities do not evoke emotional
creased availability of a hazard, the availability heuris- images; relatively high probabilities, however, may
tic is also employed as a theoretical framework. evoke vivid and affect-laden images associated with a
hazard. Because affect may also lead to an increased
availability of a hazard (Slovic et al., 2004), the avail-
2. STUDY 1
ability heuristic is an equally viable explanation for
It is well known that people find probabilities result of the study by Slovic et al. (1978).
hard to interpret (Gigerenzer et al., 2005). Interpre- In Study 1, we examined the impact on perceived
tation improved when problems were formulated in risks of various presentation formats, all of which con-
terms of frequencies rather than probabilities (Cos- veyed identical information. A frequency format can
mides & Tooby, 1996; Gigerenzer & Hoffrage, 1995). be used to summarize the risk as, “on an average, there
One reason for communicating risk to the public is is a flood every 100 years.” The same information can
to improve the correspondence between the assessed be given as a probability for one year: each year, there
magnitude of a risk and people’s responses to this risk is a 1% probability of flood. Instead of giving the risk
(Weinstein & Sandman, 1993). In the domain of flood for one year, a time period for a generation or two liv-
risks, participants often underestimate the risks they ing in a house can be provided: within 40 years, there
are faced with (Siegrist & Gutscher, 2005). Therefore, is a 33% probability of flood; within 80 years, there is
raising risk awareness and precaution behavior is usu- a 55% probability of flood. It was expected that, for
ally the goal of risk communication about natural haz- the risks expressed for one year, the frequency for-
ards. In the present study, we focus on risk awareness. mat would produce higher perceived risks than the
Whether people will in turn act in accordance with probability format. Based on earlier studies reviewed
their awareness and in fact take precautionary mea- above, it was hypothesized that presenting probabil-
sures depends on further psychological factors. Self- ities for a longer time period would result in higher
efficacy is one important factor for understanding and perceived risks compared with the probability of an
changing precaution behavior (Murray-Johnson et al., adverse event within one year.
2004; Witte, 1992). Even though not directly examined
in the present study, changing behavior is the most im-
2.1. Method
portant goal of risk communication.
Results of a study by Siegrist (1997) suggest that 2.1.1. Participants
a frequency format emphasizes the threat of a risk.
A total of 170 psychology students from the
Participants confronted with high risk in a frequency
University of Zürich participated in the study (144
format were willing to pay a higher price for improved
women, 33 men). Participants’ mean age was 25.7
medication compared with participants who received
years (SD = 5.4).
this information in a probability format. Similar re-
sults were reported by Slovic et al. (2000). Clinicians
who received risk information about a person’s vio- 2.1.2. Material and Procedure
lence in a frequency format judged the patient to be
All participants were asked to imagine that they
more dangerous than did clinicians who received the
were planning to buy a house and that they had re-
identical information in a probability format. Accord-
ceived some information about the probability of a
ing to the authors of this study, the frequency format
flood. The information about the risk was described
evoked frightening images, while the probability for-
in four different ways:
mat failed to evoke such emotional images.
Probabilities for a particular risk depend on the r Version 1: “On an average, there is a flood
time period for which a risk is given. The probability every hundred years.”
634 Keller, Siegrist, and Gutscher

r Version 2: “Each year, there is a 1% probability In the present study, much higher probabilities were
of flood.” utilized. Nevertheless, presenting information for a
r Version 3: “Within 40 years, there is a 33% longer time period resulted in higher risk ratings. Re-
probability of flood.” sults of Study 1 suggest that the specific length of the
r Version 4: “Within 80 years, there is a 55% longer time period is of minor importance. The risk
probability of flood.” ratings did not differ for a time period of 40 years to
a time period of 80 years.
In the scenario it was emphasized that “this flood
Results of Study 1 are in line with the hypothe-
causes severe damage, which is only partly covered
ses deduced from the affect heuristic. For probabil-
by insurance.” After reading one version of the short
ities presented for a longer time period, the longer
scenario, participants were asked: “How risky would
time period strongly emphasized the threat of a risk.
you consider that living in a place like this is?” Par-
According to the affect heuristic, it is plausible to as-
ticipants assessed the risks using a number between 1
sume that as a result, such risk communication for-
(“not risky at all”) and 6 (“very risky”). Participants
mats evoke more affect-laden images, which increase
were tested in class, and they completed the ques-
perceived risks. Low probabilities, on the other hand,
tionnaires anonymously. Participants were randomly
are rounded to zero and do not evoke any affect.
assigned to one out of the four conditions.
Because affect could also have an effect indirectly
through availability, these results could also be in-
2.2. Results and Discussion terpreted within the availability heuristic. The threat
evoked by probabilities presented for a longer time
The means and standard deviations of the per-
period made flooding events easily available, thereby
ceived risks for all four versions are shown in
increasing risk perception.
Table I. Data were submitted to a one-way analysis
of variance. The analysis yielded a significant effect,
F(3,173) = 7.73, p < 0.01. Results of the Tukey-test 3. STUDY 2
show that the group receiving the probability infor-
Past experience is an important factor influencing
mation for one year showed significantly lower risk
how risky hazards are perceived (Siegrist & Gutscher,
ratings than the other three groups. The other three
2005). And experiences with hazards strongly influ-
groups did not differ significantly from each other.
ence the purchase of insurance (Baumann & Sims,
Results of Study 1 are in line with other studies in
1978; Zaleskiewicz et al., 2002). However, we are not
which frequency and probability formats were used
aware of any studies that have experimentally exam-
(Siegrist, 1997; Slovic et al., 2000). The frequency for-
ined the influence of past experience on the interpre-
mat emphasized the threat of the risk and resulted,
tation of probability or frequency information about
therefore, in higher risk ratings. In addition, the find-
natural hazards. It seems plausible that statistical in-
ings of Slovic et al. (1978) were successfully replicated.
formation about flood risks may evoke concrete im-
In the Slovic study, the probability of being involved
ages in people who have had experiences with floods.
in a serious accident in one’s lifetime was about 0.01.
However, less affect would be evoked in people with
no such experience. Therefore, it was hypothesized
Table I. Study 1: Means and Standard Deviations of Responses to that past experience would increase perceived risks,
“How Risky Would You Consider Living in a Place Like This Is?”
independent of the information provided.
Format of Communicating Graphical displays are often used in risk commu-
Flood Probability M (SD) nication, though it is still an open question whether the
addition of graphics improves comprehension sub-
1. On an average, a flood every 3.59 (1.17) (n = 51)
stantially (Lipkus & Hollands, 1999). It has been sug-
100 years
2. Each year, there is a 1% 2.72 (1.21) (n = 39) gested that graphical displays may be a useful tech-
probability of flood nique for decreasing risk-taking behavior (Stone et al.,
3. Within 40 years, there is a 33% 3.91 (1.21) (n = 46) 1997). Other studies have suggested that the graphi-
probability of flood cal display may increase risk avoidance only when the
4. Within 80 years, there is a 55% 3.71 (1.25) (n = 41)
graphics highlighted the number of people harmed
probability of flood
at the expense of the total number of people at risk
Note: Ratings made on a six-point scale: 1 = not risky at all and (Stone et al., 2003), and that the graphical presenta-
6 = very risky. tion elicits the same decision-making mechanism as
Affect, Availability, and Risk Communication 635

the numerical display (Schirillo & Stone, 2005). Re- 1% 26 %

search in health communication, using graphs in com-
bination with time periods, has shown that survival
graphs strongly influence risk perception when iden-
tical risks are displayed over longer periods of time
(Zikmund-Fisher et al., 2005).
Graphical displays may emphasize the threat of
the risk and, therefore, may evoke emotions. In Study
2, we hypothesized that graphical displays in the form
of a pie chart enhance perceived risks when high prob- Fig. 1. Study 2: Graphical displays. Probability of 1% within 1 year
abilities are communicated, but we did not expect (left graphic), probability of 26% within 30 years (right graphic).
such an effect for low probabilities. Highlighting high
probabilities in a pie chart emphasizes the largeness were planning to buy a house, and that they received
of the segment. The impressive largeness may empha- some information about the probability of a flood.
size the threat of a risk, and in turn lead to increased In the scenario it was emphasized that “this flood
risk perception. In contrast, highlighting low proba- causes severe damage, which is only partly covered
bilities emphasizes the smallness of the risk. This, in by insurance.”
turn, may lead to decreased risk perception because Two factors, time period and graphical display,
no negative affect is evoked, and the risk is classified were manipulated. In the version with a graphical
as harmless. display, the risk was depicted using a pie chart (see
A survey experiment was conceived for testing Fig. 1). This resulted in the following four versions:
the hypotheses. Information was presented in a prob- r probability of 1% within one year, without
ability format for various time periods.
graphical display
r probability of 1% within one year, with graph-
3.1. Method ical display
3.1.1. Participants
r probability of 26% within 30 years, without
graphical display
Potential respondents for the survey were se- r probability of 26% within 30 years, with graph-
lected based on existing maps showing the risk of ical display
flooding. About half were selected from high and
medium flood-risk areas; the other half lived in low- After reading one version of the short scenario,
risk regions. The questionnaire contained items for participants were asked: “How risky would you con-
the survey experiment described here, as well as sider that living in a place like this is?” Participants
items designed to address different research questions assessed the risks using a number between 1 (“not
(Siegrist & Gutscher, 2005). risky at all”) and 6 (“very risky”).
During April and May 2004, the survey was con- Past experience with flooding was measured with
ducted by mail. The questionnaire was completed two items (e.g., “Neighbors or acquaintances were
and returned by 1,598 persons. The response rate for harmed by floods,” “The house I am living in/used
German-speaking Switzerland was 53% (n = 784), to live in or my personal property was once damaged
for the French-speaking part 52% (n = 522), and for by a flood”). Six response categories were used. The
the Italian-speaking part 51% (n = 293). The response two extreme responses were verbally anchored (1 =
rate was almost identical for low-, medium-, and high- does not apply at all, 6 = applies absolutely). The
risk areas. scale, consisting of these two items, had an internal
Sixty-one percent (n = 878) of the respondents consistency of α = 0.68.3
were males, 39% females (n = 558). Nine respondents
did not indicate their gender. The participants’ mean
3.2. Results and Discussion
age was 53.39 (SD = 16.64). Questionnaires were ran-
domly assigned to the respondents. Data were submitted to a 2 (time period: 1 year/
30 years) × 2 (graphical display: no/yes) ANCOVA
3.1.2. Materials
3 Results of a pretest suggested that almost all people have seen
The same scenario as in Study 1 was utilized. flooding events on TV. Therefore, this item was not included in
All participants were asked to imagine that they the final questionnaire.
636 Keller, Siegrist, and Gutscher

Table II. Study 2: Means and Standard Deviations of low reliability, we might have underestimated the ef-
Responses to “How Risky Would You Consider Living in a fect. Utilizing a scale with a higher reliability might
Place Like This Is?”
reveal a greater effect of past experience. Based on
Time Period these results, we can assume that people who expe-
Graphical Display 1 Year 30 Years
rienced past flooding events had images that were
tagged with affect (Slovic et al., 2004). Persons who
No 3.36 (1.48) n = 393 3.77 (1.44) n = 376 stored images or narratives about floods in their mem-
Yes 3.17 (1.51) n = 384 3.64 (1.44) n = 405 ories, perceived the same probability information dif-
ferently from people without such memories.
Note: Ratings made on a six-point scale: 1 = not risky at all and
6 = very risky. Standard deviations are given in parantheses.
4. STUDY 3
with past experience as a covariate. The analysis Past experience seems to have an important im-
of the dependent variable, perceived risks, yielded pact on how probabilities are interpreted. However,
a significant main effect for the time period factor in Study 2, the availability of past experience was as-
(F(1,1553) = 34.46, p < 0.001). Participants perceived sessed utilizing survey questions. Therefore, it cannot
greater danger when confronted with the risk ex- be ruled out that other factors may be crucial. The goal
pressed for 30 years than with the risk expressed of Study 3 was to directly manipulate the availability
for one year. The main effect for graphical display of flood events. According to the notion of Slovic et al.
was also significant (F(1,1553) = 4.22, p = 0.04). The (2004), availability may work because concrete and
graphical display decreased perceived risks. The in- imagined images come tagged with affect. Manipula-
teraction term was not significant (F(1,1553) = 0.07, tion of such images should evoke emotions in partic-
n.s.). The covariate past experience significantly influ- ipants who should, as a result, perceive greater risks
enced perceived risks (F(1,1553) = 28.67, p < 0.001). than participants who do not receive this treatment.
Means and standard deviations are shown in Table II.
The effect of different time periods on perceived
4.1. Participants
risks observed in Study 1 was successfully replicated
using a representative sample. In Study 1, the risk A total of 92 students from the University of
expressed for one year was compared with the risks Zürich participated in the study (52 women, 40 men).
expressed for 40 years or 80 years. In Study 2, a time Participants’ mean age was 26.2 years (SD = 5.8).
period of 30 years was compared with the risk ex-
pressed for one year. Results suggest, therefore, that
4.2. Materials and Procedure
the specific duration of the longer time period is of
minor importance. Affect was manipulated utilizing photographs.
Graphical display had an unexpected effect. Par- In the experimental group, participants were asked
ticipants who were confronted with the pie chart per- to look at two photographs depicting houses during
ceived fewer risks compared with participants who flood. In the control group, participants were asked
read the verbal information only. It could be that even to look at two photographs depicting houses without
for a time period of 30 years, the probability depicted any reference to flood. Participants looked at the pho-
in a pie chart was not that impressive. Further re- tographs for about 30 seconds.
search needs to examine whether the effect of graph- The same scenario as in Study 2 was utilized. All
ical displays is the same for small as for large proba- participants were asked to imagine that they were
bilities. Additionally, graphical displays may have an planning to buy a house, and that they had received
influence on risk perception when the graphical dis- some information about the probability of a flood.
play emphasizes information elements different from The scenario emphasized that “this flood causes se-
those in a verbal description (Stone et al., 2003). vere damage, which is only partly covered by insur-
Despite the low reliability of the scale measur- ance.” Half of the participants received the risk infor-
ing past experience, our analysis revealed a signifi- mation based on one year; the other half received the
cant effect of experience on risk perception. Past ex- risk information for a time period of 30 years.
perience had the predicted impact on risk perception. After reading one version of the short scenario,
Participants’ risk perception was strongly influenced participants were asked: “How risky would you con-
by their own experiences with flooding. Due to the sider that living in a place like this is?” Participants
Affect, Availability, and Risk Communication 637

Table III. Study 3: Means and Standard Deviations of Results of the present studies clearly supported this
Responses to “How Risky Would You Consider Living in a hypothesis. These results are in line with the affect
Place Like This Is?”
heuristic proposed by Slovic and colleagues (Finucane
Time Period et al., 2000; Slovic et al., 2002, 2004). Because affect
Condition 1 Year 30 Years
may increase the availability of a risk (Slovic et al.,
2004), the availability heuristic provides an equally
Experimental group 2.95 (1.21) n = 22 3.83 (1.07) n = 23 viable explanation for the present results.
Control group 2.22 (1.09) n = 23 3.38 (1.44) n = 24 Study 1 and Study 2 examined the influence of
various presentation formats on perceived risks. Re-
Note: Ratings made on a six-point scale: 1 = not risky at all and
6 = very risky. Standard deviations are given in parantheses. sults suggest that a frequency format and a format
giving probabilities for longer periods of time both
emphasized the threat of a risk. Study 2 showed fur-
assessed the risks using a number between 1 (“not ther that people with flood experience interpreted
risky at all”) and 6 (“very risky”). Participants were the same probability differently from people with no
tested individually. They were randomly assigned to such experience. Experiences with a hazard lead to
one out of the four conditions. higher levels of perceived risk. Study 3 directly ma-
nipulated availability of affect-laden imagery. Partici-
pants who had looked at photographs showing houses
4.3. Results and Discussion
in a flood zone evaluated the risk information as in-
Data were submitted to a 2 (time period: 1 year/ dicating greater danger than participants who looked
30 years) × 2 (affect manipulation: houses in a flooded at neutral photographs.
area/neutral houses) ANOVA. The analysis of the The present studies support the notion that utiliz-
dependent variable, perceived risks, yielded a signifi- ing the affect heuristic is important for successful risk
cant main effect for the time period (F(1,88) = 16.04, communication. One reason for communicating risk
p < 0.01), and a significant main effect for the affect to the public is to improve the correspondence be-
manipulation factor (F(1,88) = 5.50, p = 0.02). The tween the assessed magnitude of a risk and people’s
interaction term was not significant (F(1,88) = 0.32, responses to that risk (Weinstein & Sandman, 1993).
n.s.). Means and standard deviations are shown in Because people often underestimate low-probability
Table III. risks, raising risk awareness is usually one of the goals
For the time period factor, the expected effect of risk communication about natural hazards. Raising
was observed. Utilizing a longer time period increased risk awareness can be achieved when risk informa-
perceived risks. In addition, the hypothesized ef- tion is presented in the form of frequencies or in the
fect for the experimental manipulation was observed. form of probabilities for a longer time period. Both
Availability of affect-laden images about flooding approaches may evoke affect, thereby increasing the
events increased perceived risks. Based on these re- availability of a risk, resulting in greater perceived
sults, we can assume that the images evoked negative risks.
emotions, which in turn lead to higher perceived risks. Participants’ experiences with floods influenced
Due to the link between the affect and availability their interpretation of risk information. Participants
heuristics, we cannot determine which heuristic pro- who could remember images related to flood assessed
vides the better explanation for the present findings. the given risk information as indicating greater risk
However, results of Study 3 support the notion that than participants who could not remember such im-
use of the affect and availability heuristics could be ages. In risk communication, narratives from people
important for successful risk communication. who directly experienced floods could be used. Such
narratives may evoke affect-laden imagery that leads
to an increased level of perceived risk. Study 3 showed
5. GENERAL DISCUSSION
that the presentation of flood photographs had this ef-
Feelings and affect are important factors that in- fect. To increase effectiveness, photographs showing
fluence perceived risks (Slovic et al., 2004). Affect, the outcome of a low-probability risk should be in-
therefore, should be taken into account for successful cluded in risk communication materials.
risk communication. The present research tested the Some limitations of the present research need to
hypothesis that the evocation of negative affect may be discussed. The present research focused on gain-
be an efficient way for increasing perceived risks. ing people’s attention to a risk in order to increase
638 Keller, Siegrist, and Gutscher

their risk perception. To raise precautionary behavior, Finucane, M. L., Alhakami, A., Slovic, P., & Johnson, S. M. (2000).
additional psychological factors such as self-efficacy The affect heuristic in judgments of risks and benefits. Journal
of Behavioral Decision Making, 13, 1–17.
(Murray-Johnson et al., 2004; Witte, 1992) would have Fischhoff, B. (1995). Risk perception and communicating un-
to be taken into account. plugged: Twenty years of process. Risk Analysis, 15(2), 137–
Risk perception was measured by only one item. 145.
Freudenburg, W. R., & Rursch, J. A. (1994). The risks of putting the
Single-item measures tend to have low reliabilities. numbers in context—A cautionary tale. Risk Analysis, 14(6),
Therefore, we may have underestimated the effect of 949–958.
the affect heuristic. Gigerenzer, G., Hertwig, R., van den Broek, E., Fasolo, B., &
Katsikopoulos, K. V. (2005). “A 30% chance of rain tomor-
Participants’ affect was not measured. However, row”: How does the public understand probabilistic weather
our results suggest that our manipulations evoked af- forecasts? Risk Analysis, 25(3), 623–629.
fect and therefore influenced risk perception. In fu- Gigerenzer, G., & Hoffrage, U. (1995). How to improve Bayesian
reasoning without instruction—Frequency formats. Psycho-
ture studies, specific forms of affect should be assessed logical Review, 102(4), 684–704.
by using survey techniques. Furthermore, participants Jackson, E. L. (1981). Response to earthquake hazard. Environ-
should be asked what kinds of images were evoked in ment and Behavior, 13, 387–416.
Johnson, B. B. (2003). Are some risk comparisons more effective
their minds by the different communication formats. under conflict?: A replication and extension of Roth et al. Risk
This would provide greater insight into the affect- Analysis, 23(4), 767–780.
laden imagery that is an important part of the experi- Johnson, B. B. (2004). Risk comparisons, conflict, and risk accept-
ability claims. Risk Analysis, 24(1), 131–145.
ential system. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis
Results were interpreted within the framework of of decision under risk. Econometrica, 47, 263–291.
the affect heuristic (Slovic et al., 2004) and the avail- Kunreuther, H., Novemsky, N., & Kahneman, D. (2001). Making
low probabilities useful. Journal of Risk and Uncertainty, 23(2),
ability heuristic (Tversky & Kahneman, 1982). The 103–120.
availability heuristic has been proposed as an expla- Linnerooth-Bayer, J., & Amendola, A. (2003). Introduction to spe-
nation for the biases in probability judgments. Slovic cial issue on flood risks in Europe. Risk Analysis, 23(3), 537–
543.
et al. (2004) suggested that the availability heuristic Linville, P. W., Fischer, G. W., & Fischhoff, B. (1993). AIDS risk
may work because remembered images are associated perception and decision biases. In J. B. Pryor & G. D. Reeder
with affect. In the present study, participants’ decision (Eds.), The Social Psychology of HIV Infection (pp. 5–38).
Hillsdale, NJ: Lawrence Erlbaum.
processes were not examined. Whether the availabil- Lipkus, I. M., & Hollands, J. G. (1999). The visual communication
ity heuristic or the affect heuristic provides the better of risk. Journal of the National Cancer Institute Monographs,
explanation for the present findings, therefore, can- 25, 149–163.
Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch, N. (2001).
not be fully answered. In future studies, it would be Risk as feelings. Psychological Bulletin, 127, 267–286.
most important to test ways to disentangle the mech- Murray-Johnson, L., Witte, K., Patel, D., Orrego, V., Zuckerman,
anisms of the affect and availability heuristics. The C., Maxfield, A. M., & Thimons, E. D. (2004). Using the ex-
tended parallel process model to prevent noise-induced hear-
process should be examined in more detail, with the ing loss among coal miners in Appalachia. Health Education
aim of determining whether it is the number of in- & Behavior, 31(6), 741–755.
stances remembered or the affect associated with the O’Connor, R. E., Yarnal, B., Dow, K., Jocoy, C. L., & Car-
bone, G. J. (2005). Feeling at risk matters: Water managers
images that is more important in shaping perceived and the decision to use forecasts. Risk Analysis, 25(5), 1265–
risks. 1275.
Peters, E. M., Burraston, B., & Mertz, C. K. (2004). An emotion-
based model of risk perception and stigma susceptibility: Cog-
nitive appraisals of emotion, affective reactivity, worldviews,
ACKNOWLEDGMENTS and risk perceptions in the generation of technological stigma.
Risk Analysis, 24(5), 1349–1367.
Part of the research was made possible through a Roth, E., Morgan, M. G., Fischhoff, B., Lave, L., & Bostrom, A.
grant by PLANAT. (1990). What do we know about making risk comparisons. Risk
Analysis, 10(3), 375–387.
Rottenstreich, Y., & Hsee, C. K. (2001). Money, kisses, and elec-
tric shocks: On the affective psychology of risk. Psychological
REFERENCES Science, 12(3), 185–190.
Schirillo, J. A., & Stone, E. R. (2005). The greater ability of graphical
Baumann, D. D., & Sims, J. H. (1978). Flood insurance: Some de- versus numerical displays to increase risk avoidance involves
terminants of adoption. Economic Geography, 54, 189–196. a common mechanism. Risk Analysis, 25(3), 555–566.
Camerer, C. F., & Kunreuther, H. (1989). Decision-processes for Shaklee, H., & Fischhoff, B. (1990). The psychology of contracep-
low probability events—Policy implications. Journal of Policy tive surprises: Cumulative risk and contraceptive effectiveness.
Analysis and Management, 8(4), 565–592. Journal of Applied Social Psychology, 20, 385–403.
Cosmides, L., & Tooby, J. (1996). Are humans good intuitive statis- Siegrist, M. (1997). Communicating low risk magnitudes: Incidence
ticians after all? Rethinking some conclusions from the litera- rates expressed as frequency versus rates expressed as proba-
ture on judgment under uncertainty. Cognition, 58, 1–73. bility. Risk Analysis, 17(4), 507–510.
Affect, Availability, and Risk Communication 639

Siegrist, M., & Gutscher, H. (In press). Flooding risks: A compar- ior. Journal of Experimental Psychology. Applied, 3(4), 243–
ison of lay people’s perception and expert’s assessments. Risk 256.
Analysis. Svenson, O. (1984). Cognitive-processes in judging cumulative risk
Slovic, P., Finucane, M., Peters, E., & MacGregor, D. G. (2002). over different periods of time. Organizational Behavior and
The affect heuristic. In T. Gilovich, D. Griffin, & D. Kahneman Human Performance, 33(1), 22–41.
(Eds.), Heuristics and Biases: The Psychology of Intuitive Judg- Tversky, A., & Kahneman, D. (1982). Availability: A heuristic for
ment (pp. 397–420). Cambridge: Cambridge University Press. judging frequency and probability. In D. Kahneman, P. Slovic,
Slovic, P., Finucane, M. L., Peters, E., & MacGregor, D. G. (2004). & A. Tversky (Eds.), Judgment Under Uncertainty: Heuristics
Risk as analysis and risk as feelings: Some thoughts about af- and Biases (pp. 163–189). Cambridge: Cambridge University
fect, reason, risk, and rationality. Risk Analysis, 24(2), 311–322. Press.
Slovic, P., Fischhoff, B., & Lichtenstein, S. (1978). Accident proba- Weinstein, N. D. (1989). Effects of personal experience on self-
bilities and seat belt usage: A psychological perspective. Acci- protective behavior. Psychological Bulletin, 105, 31–50.
dent, Analysis and Prevention, 10, 281–285. Weinstein, N. D., & Sandman, P. M. (1993). Some criteria for eval-
Slovic, P., Kraus, N., & Covello, V. T. (1990). What should we know uating risk messages. Risk Analysis, 13(1), 103–114.
about making risk comparisons. Risk Analysis, 10(3), 389–392. Witte, K. (1992). Putting the fear back into the appeals: The ex-
Slovic, P., Monahan, J., & MacGregor, D. G. (2000). Violence risk tended parallel process model. Communication Monographs,
assessment and risk communication: The effects of using ac- 59, 329–349.
tual cases, providing instruction, and employing probability Yamagishi, K. (1997). When a 12.86% mortality is more dangerous
versus frequency formats. Law and Human Behavior, 24, 271– than 24.14%: Implications for risk communication. Applied
296. Cognitive Psychology, 11, 495–506.
Stone, E. R., Siek, R. W., Bull, B. E., Yates, J. F., Parks, S. C., & Rush, Zaleskiewicz, T., Piskorz, Z., & Borkowska, A. (2002). Fear or
C. J. (2003). Foreground:background salience: Explaining the money? Decisions on insuring oneself against flood. Risk De-
effects of graphical displays on risk avoidance. Organizational cision and Policy, 7, 221–233.
Behavior and Human Decision Processes, 90, 19–36. Zikmund-Fisher, B. J., Fagerlin, A., & Ubel, P. A. (2005). What’s
Stone, E. R., Yates, J. F., & Parker, A. M. (1997). Effects of nu- time got to do with it? Inattention to duration in interpretation
merical and graphical displays on professed risk-taking behav- of survival graphs. Risk Analysis, 25(3), 589–595.