Content uploaded by Robert F. Potter
Author content
All content in this area was uploaded by Robert F. Potter
Content may be subject to copyright.
COMMUNICATION RESEARCH • October 2001Bolls et al. • The Effects of Message Valence
PAUL D. BOLLS
ANNIE LANG
ROBERT F. POTTER
The Effects of Message Valence
and Listener Arousal on Attention,
Memory, and Facial Muscular
Responses to Radio Advertisements
This study tested the validity of using facial electromyography (EMG) as a
physiological measure of the valence of radio listeners’ emotional responses to
advertisements and explored the effects of message valence and listener
arousal on attention and memory. A within-subjects experiment was con-
ducted in which participants listened to ten 60-second radio advertisements
that had been coded in a pretest as having either a positive or negative emo-
tional tone. Facial EMG, heart rate, and skin conductance data were collected
during exposure to the advertisements. Following exposure, participants com-
pleted free recall and recognition memory tests. Results demonstrated the
validity of using facial EMG to assess the valence of emotional response to
media messages. Heart rate data suggest that negative messages receive more
attention than positive ones. Furthermore, how arousing a message is appears
to be a better predictor of memory than message valence.
The media expose audiences to a wide range of emotional experiences. Media
messages contain content explicitly designed to evoke discrete emotions such
as fear, anger, joy, and sadness. Producers are aware of the intimate relation
-
ship emotion has with three primary goals of media messages: to attract
attention, to entertain, and to persuade. Given the prevalence of emotionally
charged media messages, it is not surprising that much attention has been
directed toward investigating the relationship between emotional media con
-
tent and audience members’ emotional and cognitive responses to those mes
-
sages (A. Lang & Bolls, 1995; A. Lang & Friestad, 1993; A. Lang, Newhagen, &
627
COMMUNICATION RESEARCH, Vol. 28 No. 5, October 2001 627-651
© 2001 Sage Publications
Reeves, 1996; Newhagen & Reeves, 1992; Reeves, Newhagen, Maibach, Basil, &
Kurz, 1989; Thorson & Friestad, 1989).
This study addresses three primary conceptual, methodological, and theo
-
retical questions in understanding the relationship between emotional
media messages and audience members’ responses to those messages. First,
this article attempts to clearly differentiate between the concepts of emo
-
tional tone (a property of the media message) and emotional response (a prop
-
erty of the media audience). Second, this article explores the usefulness and
validity of facial electromyography (EMG) as a real-time, physiological mea
-
sure of the valence of radio listeners’ emotional responses to messages.
Finally, this article attempts to further investigate the question of whether
arousal or valence is a better predictor of memory for message content.
A major conceptual problem encountered by researchers studying emo
-
tional responses to emotional messages is the conceptualization of emotion
as both a feature of media content and an audience member’s response to that
emotional content (A. Lang, Bolls, Potter, & Kawahara, 1999). It is imperative
to distinguish the conceptualization of emotion as a feature of the message
from emotion as an audience response (elicited by the message). Media mes-
sages can be classified and described in terms of their emotional tone. Simi-
larly, these same emotional media messages can induce the experience of
emotion in audience members. The challenge facing researchers is to concep-
tualize clearly the difference between a message’s emotional tone and an
audience member’s emotional experience and, at same time, to develop clear
operational definitions of each.
Emotion as Experience
Considering emotion as a person’s experience requires a theory of emo
-
tion. Many researchers have studied emotion and attempted to define, cate
-
gorize, and better understand how people experience emotion. Two primary
approaches to studying emotion dominate this research area. One is to search
for primary emotions and study emotions as discrete or categorical entities.
The other looks for the underlying dimensions of emotional experience. This
study uses a dimensional theory of emotion. But either theoretical approach
could be used to develop a conceptualization of emotional experience.
A Dimensional Theory of Emotion
A significant portion of research on emotion has used a dimensional the
-
ory of emotion (Osgood, Suci, & Tannenbaum, 1957; P. J. Lang, Greenwald,
628
COMMUNICATION RESEARCH • October 2001
Bradley, & Hamm, 1993). P. J. Lang et al. proposed a dimensional model of
emotion that is founded on a biological organization of appetitive and defen
-
sive motivational connections in the human cognitive system. Lang and col
-
leagues believed that the biological organization underlying their model
accounts for much of the variance in evaluative judgments behind emotional
responses. This model proposes three underlying dimensions to emotional
response. The first dimension, valence, reflects the degree to which an emo
-
tional response is positive or negative. The second dimension, arousal, indi
-
cates the level of activation associated with the emotional response and
ranges from very excited or energized at one extreme to very calm or sleepy at
the other. A third dimension of emotional response is dominance. Measures of
dominance range from a feeling of being in control during an emotional expe
-
rience to a feeling of being controlled by the emotion.
Factor analyses of behavioral,self-report, and physiological data have con
-
sistently supported a strong two-factor solution consisting of arousal and
valence as dimensions of emotional response (P. J. Lang et al., 1993). Domi
-
nance has been found to be a less stable dimension of emotional response and
less reliably measured (Bradley & Lang, 1994). Early in the development of a
three-dimensional model of emotion, it was recognized that dominance
accounts for very little of the variance in emotional response (Mehrabian &
Russell, 1974). Russell (1980) argued that dominance should maintain some
theoretical importance but relegated dominance to a secondary factor. It has
been proposed that dominance may be a content feature of emotional stimuli
but has few motivational or behavioral consequences in determining emo-
tional response (Bradley & Lang, 2000). For this reason, a significant amount
of research has theoretically acknowledged a three-dimensional model of
emotional response but has primarily focused on manipulating and measur
-
ing arousal and valence (i.e., A. Lang, Dhillon, & Dong, 1995; P. J. Lang et al.,
1993).
The conceptualization of valence and arousal as the primary dimensions
of emotional experience is quite common (Cacioppo, Petty, Losch, & Kim,
1986; Dimberg, 1990; Hietanen, Surakka, & Linnankoski, 1998; Jancke, Vogt,
Musial, Lutz, & Kalveram, 1996; A. Lang et al., 1995; A. Lang et al., 1996; P. J.
Lang et al., 1993; Newhagen & Reeves, 1992; Wexler, Warrenburg, Schwartz, &
Janer, 1992). This research suggests that human emotional experience can be
mapped onto a two-dimensional space with valence and arousal as its axes,
with a focus on emotion as it is being experienced by a person. Various tech
-
niques are used to elicit emotion in experimental participants, but the theory
focuses on what happens within a person who is experiencing an emotion.
629
Bolls et al. • The Effects of Message Valence
Measuring Emotional Experience
Within this research tradition, a great deal of effort has gone into develop
-
ing measures of emotional experience. According to P. J. Lang (1979), there
are three types of data that can be used to measure emotional experience:
behavioral, self-report, and physiological. Behavioral data include observa
-
tions of facial responses and other emotional behaviors such as laughing or
crying. Self-report data include verbal reports of how an emotion feels, made
by the person experiencing it. Physiological data include measures of heart
rate, skin conductance, and facial electromyography collected during the
experience of an emotion.
As with any measure, each of these operational definitions of emotional
response has inherent weaknesses. Self-report measures are subject to social
response bias, particularly during the investigation of media messages cover
-
ing sensitive topics. Furthermore, there is a concern that self-report mea
-
sures provide measurement of responses to extreme points in message con
-
tent rather than responses to the message as a whole.
Coding behavioral responses also suffers from methodological weak-
nesses. Researchers have recognized that it is possible for an emotional
response to occur without a visible change in facial expression or behavior
(Ekman, 1993; Hazlett & Hazlett, 1999). Not only that, but the human brain
is structured so that emotional response can occur before an individual
engages in a behavioral response such as smiling or frowning (LeDoux, 1989).
In other words, behaviors often lag behind emotional response. Also, the cod-
ing of overt behavior cannot detect preconscious emotional responses. A final
drawback of behavioral observation is that fine distinctions in level of emo
-
tional response over time are difficult to observe using limited human senses.
Physiological measures can be used to assess both the arousal dimension
of emotion (by measuring skin conductance) and the valence dimension of
emotion (by measuring facial EMG and heart rate). However, these measures
also have their drawbacks, primarily related to the data collection methods,
which result in a loss of external validity (P. J. Lang et al., 1993).
Nonetheless, all three types of measures have been shown to yield valid
data about how humans experience emotion. Thus, emotional experience, in
this article, is conceptualized as feelings experienced by people that vary
along the dimensions of arousal and valence and can be measured by obtain
-
ing behavioral, self-report, or physiological measures from the person who is
experiencing the emotion.
The second task for researchers studying emotion is to differentiate emo
-
tional experience, located within human beings, from emotion as a media
message feature.
630
COMMUNICATION RESEARCH • October 2001
Emotional Tone—Emotion as a Message Feature
What does it mean to say that a media message is emotional? Probably
what is usually meant is that the message focuses on a topic about which peo
-
ple usually feel emotional. There is a great deal of research in media that
attempts to differentiate messages in terms of their emotional content.
Thorson and Friestad (1989), for example, have investigated the effective
-
ness of advertisements as a result of their emotional tone (positive, negative,
neutral, or poignant). Research on broadcast news has focused on the emo
-
tional tone of news stories as a predictor of processing and memory
(Newhagen & Reeves, 1992; Shoemaker, 1996). A. Lang and her colleagues
(1999) have investigated how messages that are more or less arousing affect
online processing of the messages and memory for them.
All of these approaches assume that messages can be categorized in terms
of their emotional content. That is, a message can contain emotional topics,
displays of emotion, and other types of emotional information, and the degree
to which a message contains this type of information can be assessed through
measurement.
Measuring Emotional Tone
There are three primary methods that have been used to assess emotional
tone in messages. One is to define a message as emotional if it contains dis-
plays of emotion or content that is typically considered to be emotional. This
type of definition was used, for example, in studies done by Reeves and his col
-
leagues looking at the effects of emotional media on attention and hemi
-
spheric asymmetry (Reeves et al., 1985; Reeves, A. Lang, Thorson, &
Rothchild, 1988).
A second way to measure the level of emotion in a media message is to have
people rate the messages in terms of some emotional variable, such as
valence, arousal, fear, disgust, and so on. When this method is used, research
-
ers typically select messages that they believe vary in terms of emotional
tone and then have groups of participants rate the messages on various emo
-
tional dimensions. It should be noted that when using this method, partici
-
pants can be asked either to rate the emotional content of the message or to
rate their own emotional response to the message. The resulting data are
then used to indicate the emotional tone of the message.
The third primary method for evaluating emotional tone in media mes
-
sages is somewhat different in that it involves varying relative emotional
tone through manipulation of the message. This occurs when researchers
631
Bolls et al. • The Effects of Message Valence
wish to examine two versions of the same message that vary in terms of
intensity of emotional content. In this case, a message is altered, usually
through the addition of more emotional content, so that the resultant mes
-
sage is more emotional than the original message. For example, Newhagen
and Reeves (1992) added negative emotional video to negative news stories.
They assumed the addition of the negative visual material would render the
later version of the news story more negative than the original version.
Thus, this article conceptualizes emotional tone as a property of the mes
-
sage, which can be measured through definitions of content or subject rat
-
ings. By clearly separating emotional tone from emotional experience, media
researchers will be better able to investigate the relationship between the
two variables without confusion.
Using EMG to Measure Emotional Response
The second goal of this article is to explore the usefulness and validity of
facial EMG as a physiological measure of the valence of radio listeners’ emo-
tional responses to messages. As discussed above, emotional experience can
be assessed in a variety of ways. This study was designed to test the validity of
using a physiological response as an indicator of the emotional valence expe-
rienced by radio listeners during exposure. More specifically, this experiment
assessed the validity of using facial EMG as a physiological measure of the
valence of emotional response to radio advertisements.
The vast majority of research examining emotional responses to media
has relied on self-reports of emotional experience as the primary
operationalization of emotional experience. Certainly, our intent is not to
diminish the contributions made by these studies. However, it is our conten
-
tion that research in the area of emotional communication can be advanced
by using facial EMG to provide a real-time, precise measure of how positive or
negative an audience member feels while processing the message.
Valence and Facial EMG
The theoretical connection between muscle activity and emotional experi
-
ence dates back to William James’s (1884) peripheral theory of emotion.
James proposed that the contraction of facial muscles was not only a form of
emotional expression but greatly responsible for the experience of emotions
themselves. Facial EMG is the modern measurement of these facial muscle
contractions.This physiological method measures the electrical signal gener
-
ated by the occurrence of action potentials across a group of muscles dedi
-
cated to moving particular parts of the face (Cacioppo, Tassinary, & Fridlund,
632
COMMUNICATION RESEARCH • October 2001
1990). These electrical signals can be detected by the careful placement of
electrodes over specific muscle groups on the surface of the skin. Facial EMG
is commonly obtained from the zygomatic and corrugator muscle groups. The
zygomatic muscle is located along the cheek just above the corner of the lips.
Contraction of the zygomatic muscle draws the angle of the mouth backward
and upward, leading the zygomatic muscle to become known as the smile
muscle (Fridlund & Izard, 1983). The corrugator muscle is located on the
brow just off the bridge of the nose. Contraction of the corrugator muscle
moves the brow downward and inward, associating the corrugator muscle
with frowning (Fridlund & Izard, 1983).
Facial EMG has been used extensively by psychophysiologists to measure
the valence of emotional response to environmental stimuli. Fridlund and
Izard (1983) did a broad review of much of the early work using this physio
-
logical measure.The valence of a wide range of visual stimuli has been shown
to elicit differential responses from muscles in the face (Cacioppo et al., 1986).
Much of this work used still slides of environmental stimuli that had been
coded for valence of emotional tone (P. J. Lang et al., 1993). P. J. Lang and col-
leagues exposed participants to slides that had been rated as having positive
or negative emotional tone and reported increased activity in the zygomatic
(smile) muscle group for positive pictures and increased activity in the corru-
gator (frown) muscle group for negative pictures. Facial EMG has also been
measured in response to audio tones of different intensity and valence
(Dimberg, 1990), environmental sounds (Jancke et al., 1996), semantically
positive and negative spoken words (Wexler, Warrenburg, Schwartz, & Janer,
1992), and vocal affective cues (Hietanen et al., 1998). As with visual stimuli,
all of these studies report increased zygomatic activity for positive stimuli
and increased corrugator activity for negative stimuli.
Little research has been done to assess the validity of using facial EMG to
measure emotional responses to more complex stimuli such as broadcast
advertisements. Broadcast advertisements are significantly longer and con
-
tain many more emotional cues than the simple stimuli historically used
by psychophysiologists. Recently, Hazlett and Hazlett (1999) compared
facial EMG to self-report as a measure of emotional responses to television
advertisements. Their results showed that facial EMG more effectively dis
-
criminated emotional response to specific commercials than self-reported
measures. Furthermore, peaks in facial EMG response were found to be
temporally related to specific emotional points in the commercial. The authors
concluded that facial EMG has the potential to be a very useful tool for esti
-
mating emotional response to advertisements that vary in emotional tone.
Upon review of the literature, Hazlett and Hazlett’s (1999) study appears
to be the only study that assesses the validity of using facial EMG to identify
633
Bolls et al. • The Effects of Message Valence
emotional response to broadcast messages. Although their study demon
-
strates that facial musculature response has a temporal relationship to spe
-
cific emotional cues in television advertisements, the authors did not directly
assess the validity of facial EMG as a measurement of emotional valence
because their stimuli were not chosen as explicit exemplars of messages with
specific emotional tone. Furthermore, no studies have been done that assess
facial EMG responses to radio advertisements. Therefore, this study is
designed to directly assess the validity of using facial EMG to measure the
valence of emotional response elicited by radio advertisements that vary in
terms of the valence of their emotional tone.
Despite broadcast advertisements being longer and containing more emo
-
tional cues than still slides, there is no evidence suggesting these messages
should not elicit a measurable emotional experience in listeners. Nor is there
reason to expect emotional auditory messages affect the facial muscles differ
-
ently than visual ones. In other words, the valence (positive or negative) of
the emotional tone of a media message—in this case, a radio message—
should lead to a global increase in zygomatic (smile) muscle activity for posi-
tive messages and a global increase in corrugator (frown) muscle activity for
negative messages. This leads to the first two hypotheses.
Hypothesis 1: Zygomatic muscle activity will be greater for radio adver-
tisements with a positive emotional tone compared with radio adver-
tisements with a negative emotional tone.
Hypothesis 2: Corrugator muscle activity will be greater for radio adver-
tisements with a negative emotional tone compared with radio adver-
tisements with a positive emotional tone.
Message Valence and Cognitive Processing
Message Valence and Attention
The final question addressed by this study concerns relationships among
emotional experience, attention, and memory for media messages. A growing
body of research has investigated the effects of message valence on attention
and memory (Geiger & Reeves, 1993; Gunter, 1987; A. Lang et al., 1995, 1996;
Newhagen & Reeves, 1992). Much of this research has been conducted under
the framework of limited capacity theory (A. Lang, 2000). Limited capacity
theory proposes that attention is under the dual control of the audience mem
-
ber and characteristics of the message. Audience members can purposefully
allocate attention based on goals. Attention can also be reflexively elicited
from audience members by features of the message. One way of indexing this
634
COMMUNICATION RESEARCH • October 2001
reflexive elicitation of attention is to measure the heart rate of the message
recipient. Previous research has found that as the amount of attention being
paid to a media message increases, heart rate decreases (A. Lang, 1995). In
addition, this research has shown that the emotional tone of a message alters
resource allocation, attention, and heart rate. It has frequently been sug
-
gested that organisms are hardwired to process negative stimuli to promote
survival (Bradley, 1994; Shoemaker, 1996; Zajonc, 1984), and research inves
-
tigating exposure to media messages with negative emotional tone has con
-
sistently shown decreases in heart rate associated with media messages with
greater negative emotional tone (A. Lang, et al., 1995, 1996). For example, A.
Lang et al. (1996) measured heart rate during television news stories that
included either negative or nonnegative video footage. Results showed slower
heart rate during exposure to news stories that included the presence of neg
-
ative video footage compared with stories with nonnegative video footage.
They suggested negative video is a production feature capable of reflexively
eliciting attention to a television message. If humans are hardwired to reflex
-
ively allocate attention to negative information, then heart rate should be
slower during exposure to negative stimuli regardless of whether that stim-
uli is primarily visual (such as with television) or exclusively auditory (such
as with radio). Thus, Hypothesis 3 is as follows:
Hypothesis 3: The heart rate will be slower during exposure to radio adver-
tisements with a negative emotional tone compared with radio adver-
tisements with a positive emotional tone.
Message Valence and Memory
The question of whether messages with a positive or negative emotional
tone are more memorable has bewildered media researchers. Researchers
who have studied message valence and memory have produced a somewhat
mixed set of results. Some studies suggest that negative messages are
remembered better than positive messages (A. Lang & Friestad, 1993;
Newhagen & Reeves, 1992; Reeves et al., 1989; Thorson & Friestad, 1985).
Other work suggests that positive messages are more memorable (A. Lang
et al., 1995; Shapiro & Reiger, 1989).
Cognitive psychology provides two contradictory theoretical perspectives
on whether negative or positive stimuli should be more memorable. One per
-
spective, already mentioned, is that humans are hardwired to allocate more
attention to negative stimuli (Zajonc, 1984). This view has been used to argue
that negative stimuli reflexively attract attention from people, resulting in
better memory for bad things. A second theory is that people have a natural
635
Bolls et al. • The Effects of Message Valence
tendency to maximize pleasantness (Fiske & Taylor, 1984). Under this per
-
spective, people exhibit an approach response to positive stimuli and an
avoidance response to negative stimuli. Theoretically, approaching some
-
thing pleasurable results in higher levels of attentional resources being allo
-
cated, and avoiding something dreadful does just the opposite.
Both of these theoretical perspectives make the assumption that
increased attention allocated to a stimulus translates into better memory,
which may not be true under all circumstances. A. Lang (2000) proposed that
it is possible for an audience member to allocate a large amount of attentional
resources toward processing a message and later have poor memory for its
content. If this is the case, then valence-driven increases in attentional
resources, regardless of whether they are caused by positive or negative emo
-
tional tone, may not always lead to accurate predictions about the effects of
message valence on memory.
A possible explanation for conflicting theories and results concerning the
effects of valence on memory could be that researchers fail to account for the
effects of another dimension of emotional response, that is, arousal. Arousal,
along with valence, has long been recognized as an important characteristic
of a media message’s emotional tone (A. Lang & Bolls, 1995). A. Lang et al.
(1995) suggested that one reason several studies have found better memory
for negative media messages is that commonly, media messages with a nega-
tive emotional tone tend to be more arousing than positive ones. To test this
hypothesis, they conducted an experiment whereby the emotional tone of
media messages was manipulated both in terms of valence and arousal.
Results showed that when arousal was controlled, positive messages were
remembered slightly better than negative messages. They concluded that the
arousal dimension of an emotional tone may play a larger role in determining
memory than the valence dimension. However, when arousal is controlled,
positive messages may be remembered better than negative messages. This
leads to our final set of hypotheses:
Hypothesis 4a: If positive and negative radio advertisements result in
experienced arousal levels that are not significantly different, positive
messages will be better remembered than negative ones.
Hypothesis 4b: If positive and negative radio advertisements result in
experienced arousal levels that are significantly different, the more
arousing group of messages, whether positive or negative, will be more
memorable.
636
COMMUNICATION RESEARCH • October 2001
Method
Design
The design for this experiment was a mixed 2 (valence) ´ 5 (message) ´ 2
(order of presentation) repeated-measures factorial design. In this design,
arousal was controlled during stimulus selection. Given the purpose of this
experiment, a strong manipulation of valence rather than arousal was
desired. The only between-subjects factor in this design was Order of Presen
-
tation, with two levels representing the two different orders in which the
radio advertisements were presented. Participants were randomly assigned
to one of the presentation orders. The two within-subjects factors were
Valence, with two levels representing radio advertisements with positive or
negative emotional tone, and Message, with five levels representing the five
different advertisements in each level of the Valence factor. Each participant
heard all 10 advertisements during the experiment.
Stimulus Messages
Stimulus messages were selected from a pool of the Mercury Award–
winning radio commercials from the years 1993 to 1995, which were provided
on compact disc by the Radio Advertising Bureau. The Mercury Awards are
given annually by the Radio Creative Fund to honor the best radio commer-
cials of the calendar year. The use of award-winning advertisements served
as a control for professional quality. Furthermore, because the advertise
-
ments were at least 4 years old, participants had not been recently exposed to
them.To choose the stimuli for the experiment, 60 advertisements were cho
-
sen based on the researchers’ evaluation of whether they had a positive or
negative emotional tone. These 60 advertisements were pretested using par
-
ticipants (n = 17) similar in age to those who would participate in the experi
-
ment. The pretest procedure consisted of participants listening to the 60 ads
in one of two different orders of presentation and rating their emotional
responses to each using the Self Assessment Mannequin (SAM) (P. J. Lang
et al., 1993). SAM is a three-dimensional, 9-point pictorial scale that mea
-
sures emotional response on the three dimensions of arousal, valence, and
dominance. SAM has proven to be a reliable and valid measure of emotional
response to advertisements (Morris, 1995) and emotional auditory stimuli
(Potter & Callison, 2000). Mean arousal and valence levels for each
637
Bolls et al. • The Effects of Message Valence
advertisement were calculated. Then, two groups of five ads were selectively
chosen from both the extreme positive and extreme negative poles of the
mean valence scores such that the groups were significantly different in their
level of perceived valence,F(1,15) = 352.62, p < .001, e
2
= .9521, but not signifi
-
cantly different on the arousal dimension (p > .05). Positive messages were
typified by features such as upbeat music, upbeat vocal delivery, and humor.
Negative messages tended to incorporate features such as dark-sounding
music, voice effects (i.e., raspy voice in an antismoking message), and fear
appeal. As a repeated-measures experiment, this study used five messages at
each level of Valence to assess global emotional response to Valence rather
than emotional response to specific production features such as humor or
fear appeal.
The 10 stimulus messages were dubbed into two orders of presentation on
the audio track of a time-coded VHS videotape. One order began with a posi
-
tive advertisement, the other with a negative one. Both orders then alter
-
nated between positive and negative advertisements. No two advertisements
were heard sequentially in both orders. To prevent possible primacy and
recency effects, no advertisement appeared exclusively in the first two or last
two positions of both orders. The five positive radio advertisements consisted
of ads for The Simpsons, Bud Light, Sega, Solaris Yard Basics, and Sheik con-
doms. The five negative radio advertisements consisted of ads for The Boston
Globe, Partnership for a Drug Free America, No Guns in the Street, Abbott
Northwestern Hospital, and the Massachusetts Department of Public
Health.
1
Apparatus
This experiment was controlled by a 386 computer using the Slimy
Recognition/Reaction Time program (Newhagen, 1993) as the master con
-
troller. The stimulus tapes were played by a Panasonic videocassette
recorder, with the time code being read by a Horita TRG-50 PC Time Code
Reader/Generator. The time code reader interfaced with the Slimy program,
which instructed the physiology computer to start and stop data collection at
exact time code points corresponding to the beginning and end of each radio
advertisement. The physiological data were collected and stored by a 386
computer with a LabMaster AD/DA board installed.
Skin conductance was measured by placing two Beckman standard
AG/AGCL electrodes on the participant’s palmar surface after it had been
wiped with distilled water to control hydration levels across participants. To
help make a connection with the participant’s skin, the skin conductance
638
COMMUNICATION RESEARCH • October 2001
electrodes were filled with an electrically neutral medium as suggested by
Dawson, Schell, and Filion (1990). The signal from these electrodes was
passed through a Coulbourn skin conductance coupler and was sampled 20
times per second during exposure to each message.
Heart rate and zygomatic and corrugator activity were all collected using
bipolar electrode placement with a single ground electrode as reference
(Marshall-Goodell, Tassinary, & Cacioppo, 1990). Prior to electrode placement,
the participant used distilled water and a clean towel to remove makeup and
other substances from cheeks and above eyebrows. The experimenter pre
-
pared the participant’s forearms by wiping them with distilled water and a
clean towel.
A Beckman standard AG/AGCL electrode was placed on the participant’s
nondominant forearm to serve as the ground connection. A Beckman stan
-
dard AG/AGCL electrode filled with amplification jelly was placed on each of
the participant’s forearms to obtain heart rate data. Heart rate was recorded
as milliseconds between heart beats and later converted to beats per minute
for analysis.
Zygomatic and corrugator muscle activity was recorded using two pairs of
Beckman mini-AG/AGCL electrodes filled with amplification jelly. Two elec-
trodes were placed on the skin surface of the cheek (zygomaticus major mus-
cle), and two electrodes were placed on the skin surface along the brow (corru-
gator supercilii muscle). The signals from these electrodes were sampled 20
times per second during exposure to the messages. The signals were passed
from participants into a bioamplifier attached to high and low pass filters
and then sent through a contour follower for smoothing.
The radio advertisements were heard by the participant through the
speakers of a 19-inch television set. The dimensions of the speakers were
approximately 2 inches by 6 inches and were comparable to the size of the
speakers in many portable stereos. Participants sat in a comfortable chair
approximately 5 feet from the speakers. No video signal was sent to the set
during the presentation of the radio advertisements. The videocassette
player and the experimenter were located in a separate room from the
participant.
Participants
Experiment participants were 41 undergraduate students, 20 males and
21 females, recruited from telecommunication courses at a large Midwest
university. All participants received course credit for their participation.
639
Bolls et al. • The Effects of Message Valence
Experimental Procedures
Participants completed the experiment individually. Upon arrival at the
lab, the experimenter greeted each participant and explained that the pur
-
pose of the study was to explore how people cognitively process media mes
-
sages. Participants were also informed of the procedures and small risks
involved in the collection of physiological data. Informed consent was
obtained from all participants.
After obtaining informed consent, the radio advertisements were played.
In between each advertisement, the stimulus tape was paused to allow the
participant to complete a two-page written questionnaire. The questionnaire
contained a measure of attitude toward the advertisement and the SAM
scale.
Following the presentation of all radio advertisements, participants com
-
pleted a second experiment designed to measure responses to different styles
of television news stories. This served as a distraction task for the radio
experiment.
After completing the distraction task, participants completed two memory
tests for the radio advertisements. The first was a free-recall measure in
which participants were handed a piece of paper with two columns of answer
blanks. The columns were labeled “Brand/Organization” and “Product /
Issue.” Participants were instructed to “write down the brand name or orga-
nization sponsoring the commercial and the specific product or issue that was
advertised for as many of the radio advertisements as you can remember.”
Participants were verbally instructed to inform the experimenter when they
were finished or when they could not remember any more commercials. No
participant was allowed to spend more than 5 minutes completing the
free-recall measure.
The second memory measure was a recognition test. Participants were
presented with a checklist of 42 different brands and organizations. All 10 of
the actual sponsors were included, with the remaining options being foils.
Participants were instructed to “place an X on the line next to the
Brand/Organization names you believe were advertised in one of the radio
commercials you listened to earlier.”
Analysis
All physiological data (heart rate, skin conductance, and facial EMG)
were averaged over 5-second intervals yielding 11 time points per message
for analysis. Physiological data were analyzed by submitting the data to a
640
COMMUNICATION RESEARCH • October 2001
2 (valence) ´ 11 (time) ´ 5 (message) ´ 2 (order) repeated-measures
MANOVA. Time was included as a variable in the analysis because physio
-
logical response occurs across time beginning at the onset of a stimulus
(Cacioppo, Tassinary, & Berntson, 2000). Furthermore, responses to emo
-
tional media content take time to occur. A message is not immediately posi
-
tive or negative, but rather, emotional responses occur as the content of the
message develops over time. All memory data were scored either as 1 (hit) or
0 (miss) and submitted to a 2 (valence) ´ 5 (message) ´ 2 (order)
repeated-measures MANOVA for analysis. The design of this experiment
includes numerous effects that could be tested to control for family wise error;
however, only the predicted effects were tested (Cohen & Cohen, 1983). Only
statistically significant results are reported. Order did not significantly
interact with any of the predicted effects.
Results
Manipulation Check
A manipulation check was performed on perceived Valence of the radio
advertisements by submitting the SAM valence scores obtained from experi-
ment participants to a mixed 2 (valence) ´ 5 (message) ´ 2 (order of presenta-
tion) repeated-measures MANOVA. Results showed a successful manipula-
tion of participants’ emotional experience. The positive and negative
messages were rated significantly different by experiment participants on
the valence dimension of the SAM scale, F(1, 39) = 209.86, p < .001. Positive
messages were rated significantly more positive (M = 6.59, SD = 1.10) than
negative messages (M = 3.15, SD = .98).
Hypothesis 1
This hypothesis predicted that zygomatic muscle activity would be signifi
-
cantly greater during positive radio advertisements compared with negative
advertisements. The main effect for Valence on zygomatic muscle activity
was significant, F(1, 33) = 19.01, p < .001, e
2
= .341. There was also a signifi
-
cant Valence ´ Time interaction, F(1, 330) = 8.16, p < .001, e
2
= .193. This indi
-
cates that the effect of Valence on zygomatic muscle activity is significantly
different across the 60 seconds of the message. Figure 1 shows that zygomatic
activity significantly increases during the 60 seconds of exposure for positive
messages—presumably as the emotion builds—but is significantly lower and
slightly decreases over the course of negative messages.
641
Bolls et al. • The Effects of Message Valence
Hypothesis 2
This hypothesis predicted that corrugator muscle activity would be
greater during negative advertisements compared to positive advertisements.
The main effect of Valence on corrugator muscle activity was signifi-
cant, F(1, 33) = 26.63, p < .001, e
2
= .4291. There was also a significant Valence
´ Time interaction, F(1, 330) = 13.83, p < .001, e
2
= .291. This interaction is
shown in Figure 2. Corrugator muscle activity increases and then levels off
over the course of negative messages but is significantly lower and slightly
decreases over the course of positive messages.
Hypothesis 3
The third hypothesis predicted that heart rate would be slower for nega
-
tive radio advertisements compared to positive radio advertisements. There
was a significant main effect for Valence on heart rate, F(1, 22) = 5.34, p <
.031, e
2
= .158, which is shown in Figure 3. Heart rate was significantly slower
during negative messages compared with positive messages. There was no
Valence ´ Time interaction on heart rate.
Hypothesis 4
The final hypothesis predicted that if the positive and negative adver
-
tisements were perceived as equally arousing, then the positive radio advertise-
642
COMMUNICATION RESEARCH • October 2001
-0.5
0
0.5
1
1.5
2
2.5
change in microvolts
0 5 10 15 20 25 30 35 40 45 50 55
seconds
positive
negative
Figure 1. The Effect of Message Valence on Zygomatic (smile) Muscle Activity During
Exposure to Radio Advertisements
ments would be more memorable than the negative advertisements. How
-
ever, if arousal were significantly different between the positive and negative
advertisements, the more arousing group of advertisements would be more
memorable regardless of valence.
The first step in testing Hypothesis 4 was to determine if the positive and
negative radio advertisements used in this experiment were equally arous
-
ing to participants.
2
The self-reported arousal scores from the SAM scale
643
Bolls et al. • The Effects of Message Valence
5 10 15 20 25 30 35 40 45 50 55
seconds
positive
negative
-1
-0.5
0
0.5
1
1.5
change in microvolts
Figure 2. The Effect of Message Valence on Corrugator (frown) Muscle Activity During
Exposure to Radio Advertisements
5 10152025303540455055
Figure 3. The Effect of Message Valence on Heart Rate During Exposure to Radio
Advertisements
were submitted to a 2 (valence) ´ 5 (message) ´ 2 (order) repeated-measures
MANOVA. The main effect for Valence on arousal scores was significant, F(1,
39) = 20.62, p < .000, e
2
= .333. The positive messages (M = 6.04) were rated on
the SAM scale as more arousing than the negative messages (M = 4.94). A
skin conductance response analysis was also conducted. This analysis
involved submitting the skin conductance data to a 2 (valence) ´ 5 (message)
´ 2 (order) repeated-measures MANOVA. Again, there was a main effect for
Valence on the number of skin conductance responses, F(1, 34) = 5.29, p <
.028, e
2
= .110. Participants had more spontaneous skin conductance
responses during positive messages (M = 3.37, SD = 2.84) compared with neg
-
ative messages (M = 2.82, SD = 2.83). Both the SAM and skin conductance
data clearly show that the participants’ emotional experience was signifi
-
cantly more arousing during the positive advertisements compared with the
negative advertisements. Because the positive radio advertisements were
more arousing than the negative advertisements,the prediction for Hypothe
-
sis 4 is that the positive radio advertisements will be more memorable than
the negative ones.
To test Hypothesis 4, the recognition and free recall data were submitted
to a 2 (valence) ´ 5 (message) ´ 2 (order) repeated-measures ANOVA. There
were significant main effects of Valence for both the recognition data, F(1,
39) = 25.12, p <.001, e
2
= .375, and for the free-recall data, F(1, 39) = 11.68,
p <.001, e
2
= .224. Participants recognized the positive messages (M = .90,
SD = .14) better than the negative messages (M = .71, SD = .22). Free recall
was also better for the positive messages (M = .70, SD = .22) compared with
the negative messages (M = .56, SD = .20). The results of this analysis are dis
-
played in Figure 4.
Discussion
Results of this experiment demonstrate the validity of using facial EMG
as a physiological measure of the valence of emotional response to radio
advertisements. As predicted, activity in muscles involved in smiling and
frowning clearly differentiated between messages that had been previously
coded as either positive or negative in emotional tone. A comparison of Fig
-
ures 1 and 2 shows zygomatic (smile) and corrugator (frown) muscles
responding differentially according to message valence. During positive mes
-
sages, zygomatic activity increases, whereas corrugator activity slightly
decreases. During negative messages, corrugator activity increases, whereas
zygomatic activity slightly decreases. Valence of the radio messages was
coded using the SAM scale. Therefore,the pattern of results offers convergent
validity between the SAM scale and facial EMG. Given the extensive testing
644
COMMUNICATION RESEARCH • October 2001
of the SAM scale as a self-report measure of valence (A. Lang et al., 1995; P. J.
Lang et al., 1993; Morris, 1995), researchers should feel fairly confident also
using facial EMG as a measure of valence of emotional response. Further-
more,the pattern of results found for facial EMG activity mirror the results of
previous research using facial EMG as a measure of valence during exposure
to other types of stimuli. Stimuli used in this experiment were radio adver
-
tisements. However, we believe this physiological response to message
valence could be robust enough to generalize across mediums. The results of
this study along with work indicating the validity of using facial EMG as a
measure of the valence of emotional response to still slides (P. J. Lang et al.,
1993) should give researchers the confidence to use facial EMG to assess
emotional response in experiments using video and computer-based stimuli.
This is not to suggest other measures such as self-report and facial expres
-
sion coding should be abandoned. Rather, facial EMG should be viewed as a
useful addition to a media researcher’s bag of tools. Facial EMG can allow
researchers to assess the valence of participants’ emotional response without
directly asking them. This advantage may be particularly useful during the
investigation of media content related to sensitive topics. Furthermore, facial
EMG can identify subtle reactions to emotional media that may not be per
-
ceptible using the facial expression coding methodology.
645
Bolls et al. • The Effects of Message Valence
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Percent correct
Free Recall Recognition
Positive
Negative
Figure 4. Free Recall and Recognition Memory Scores for the Radio Advertisements
The measurement of facial EMG also has a level of temporal precision that
could allow researchers to precisely pinpoint highly emotional media content
and investigate responses to messages that are not exclusively positive or
negative. Hazlett and Hazlett (1999) demonstrated that tracking facial EMG
activity over the course of a television advertisement reveals emotional
responses to specific content in the advertisement across time. The present
study was not designed to examine emotional responses to content within a
specific message; however, it is interesting to note the general trend in facial
EMG activity across time even when the data is collapsed across messages.
During exposure to the positive radio advertisements, zygomatic muscle
activity sharply increased over the first 20 seconds of the messages and then
showed both slight increases and decreases for the remaining 40 seconds.
During exposure to the negative radio advertisements, corrugator activity
showed a much more gradual and steady increase in activity over all 60 sec
-
onds of the messages. The differences in zygomatic and corrugator activity
across time may suggest one way that cognitive processing of positive media
content is different than processing of negative content. Positive content may
evoke sharper and more sudden emotional responses compared to negative
content. Alternatively, the difference in zygomatic and corrugator activity
found in this study could be due to differences in the emotional buildup of pos-
itive versus negative radio messages. Regardless, the variation in both
zygomatic and corrugator activity across time offers further evidence that
facial EMG activity responds to the content of the message across time. The
results of this study, taken with the work of Hazlett and Hazlett, should give
researchers the confidence to use facial EMG to measure the valence of emo
-
tional response to content features occurring at specific time points in a mes
-
sage as well as general emotional response across messages.
There are several limitations to using facial EMG that researchers should
keep in mind. Facial EMG has been noted to be susceptible to demand charac
-
teristics of the experiment (Fridlund & Izard, 1983). Placing electrodes on the
face of a participant may sensitize the participant to the fact that the experi
-
ment has something to do with facial expressions. Researchers must take
special care to disguise the purpose of the experiment and distract the partic
-
ipant during electrode placement. The facial EMG signal is also susceptible
to several sources of noise, ranging from electrical interference to movement
artifact from the participant. As a result, care must also be taken to properly
filter the EMG signal and to monitor any extra movements on the part of the
participant. Fridlund and Cacioppo (1986) have provided a thorough discus
-
sion of the limitations and guidelines for the collection of EMG data.
The results of this experiment also support the theoretical position that
people pay more attention to negative stimuli. As predicted, heart rate was
646
COMMUNICATION RESEARCH • October 2001
slower during exposure to negative messages compared with positive mes
-
sages. Because past research shows that decelerations in heart rate indicate
an increase in allocation of cognitive resources to message processing, we
suggest that the current results indicate that participants allocated more
attention to the negative advertisements compared with the positive ones.
However, our results also show that allocating more attention to a group of
messages does not always lead to better memory for message content. Even
though the negative advertisements received more resource allocation, data
show that the positive advertisements were more memorable. We suggest
that this seeming contradiction can be explained not by the amount of atten
-
tion allocated to the advertisements but rather by the levels of arousal expe
-
rienced by participants during exposure. The positive advertisements in this
study resulted in significantly higher levels of self-reported and physiological
arousal compared with the negative ones. So even though greater attention
was paid to the negative advertisements, a sense of arousal was lacking dur
-
ing exposure that kept the content of the advertisement from being soundly
stored in long-term memory. This, in turn, prevented it from being easily
retrieved during later memory tests. During exposure to the positive adver-
tisements, however, higher arousal levels led to an increase in the amount of
information being well stored—even though comparatively less was being
attended to. One practical suggestion resulting from these findings is that if
the goal of a media message is to be remembered, producers should put ample
effort into making it arousing rather than focusing exclusively on the valence
of its appeal.
One limitation to this study is that the data do not address the possible
impact of dominance on the pattern of results found. As stated earlier,
research has shown that dominance accounts for a very small portion of the
variance in emotional response. The model of emotion used in this study is
founded on deep-seeded nervous system pathways believed to be responsible
for the biological expression of emotion (P. J. Lang, Bradley, & Cuthbert,
1997). Given that valence and arousal account for most of the variance in
emotional response, it is likely that these nervous system pathways,
through which facial EMG response occurs, primarily respond to valence
and arousal rather than dominance. Because dominance plays such a small
role in emotional response, any effect of it on the results of this study is proba
-
bly minimal.
One should also use care in interpreting the results based on arousal expe
-
rienced during exposure to the radio messages. The results indicate that
valence predicts facial EMG activity, whereas arousal predicts memory.
Arousal was not manipulated the same as valence was in this study. The pri
-
mary purpose of the study was to investigate how facial EMG responds to the
647
Bolls et al. • The Effects of Message Valence
valence of radio messages. Therefore,a strong manipulation of valence rather
than arousal was desired. However, because arousal was measured, the data
allowed for analysis of the effect of arousal on memory because experiment
participants felt significantly different levels of arousal during exposure to
positive versus negative messages. The hypothesis dealing with the effect of
arousal on memory was based on arousal felt by the participants rather than
a manipulation of arousal in the messages by the researchers.
As a laboratory experiment, this study is susceptible to several limita
-
tions. Conditions in this experiment were not meant to represent a natural
listening environment. As a result, we were unable to assess the impact of
other variables, such as individual goals and tasks done, in conjunction with
radio listening, which would likely affect the cognitive processing of media in
more natural environments. Using controlled laboratory environments to
examine the cognitive and emotional variables of interest in this study
allowed us to discover solid relationships before moving into the messy world
of testing more ecologically valid listening environments. Furthermore, it
seems likely that variables present in more natural environments may medi-
ate the effects discovered in this experiment but would not change their
direction. Therefore, despite limitations, this study makes significant contri-
butions to the research community’s ability to measure emotional responses
to positive and negative media messages and deepens our understanding of
the interrelationship between valence, arousal, attention, and memory.
Notes
1. Interested researchers can obtain the stimulus messages used in this experiment
by contacting Paul Bolls, Edward R. Murrow School of Communication, Washington
State University, P.O. Box 642520, Pullman, WA 99164-2520.
2. This initial analysis was necessary to determine if the actual participants had
arousal responses to the messages consistent with the pretest participants. The origi
-
nal stimulus materials were chosen based on their ability to create a significant
valence manipulation without a significant difference in self-reported arousal levels
between levels of the Valence factor. This group of stimuli did not replicate the
nonsignificant arousal findings, leaving us to test Hypothesis 4b.
References
Bradley, M. M. (1994). Emotional memory: A dimensional analysis. In S. Van
Goozen, N. E. Van de Poll, & J. A. Sergeant (Eds.), Emotions: Essays on
emotion theory (pp. 97-134). Hillsdale, NJ: Lawrence Erlbaum.
648
COMMUNICATION RESEARCH • October 2001
Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The self-assessment
manikin and the semantic differential. Journal of Behavioral Therapy
and Experimental Psychiatry, 25, 49-59.
Bradley, M. M., & Lang, P. J. (2000). Measuring emotion: Behavior feeling, and
physiology. In R. D. Lane & L. Nadel (Eds.), Cognitive neuroscience of emo
-
tion. New York: Oxford University Press.
Cacioppo, J. T., Petty, R. E., Losch, M. E., & Kim, H. S. (1986). Electro-
myographic activity over facial muscle regions can differentiate the
valence and intensity of affective reactions. Journal of Personality and
Social Psychology, 50, 260-268.
Cacioppo, J. T., Tassinary, L. G., & Berntson, G. G. (2000). Psychophysiological
science. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.), Hand
-
book of psychophysiology (pp. 3-26). New York: Cambridge University
Press.
Cacioppo, J. T., Tassinary, L. G., & Fridlund, A. J. (1990). The skeletomotor
system. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of
psychophysiology: Physical, social, and inferential elements (pp. 325-384).
New York: Cambridge University Press.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analy
-
sis for the behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum.
Dawson, M. E., Schell, A. M., & Filion, D. L. (1990). The electrodermal system.
In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of psychophysiology:
Physical, social, and inferential elements (pp. 295-324). New York: Cam-
bridge University Press.
Dimberg, U. (1990). Facial electromyographic reactions and autonomic activ-
ity to auditory stimuli. Biological Psychology, 31, 137-147.
Ekman, P. (1993). Facial expression and emotion. American Psychologist,
48(4), 384-392.
Fiske, S., & Taylor, S. (1984). Social cognition. New York: Random House.
Fridlund, A. J., & Cacioppo, J. T. (1986). Guidelines for human electro-
myographic research. Psychophysiology, 23, 567-589.
Fridlund, A. J., & Izard, C. E. (1983). Electromyographic studies of facial
expressions of emotions and patterns of emotions. In J. T. Cacioppo &
R. E. Petty (Eds.), Social psychophysiology (pp. 243-286). New York:
Guilford.
Geiger, S., & Reeves, B. (1993). The effects of scene changes and semantic
relatedness on attention to television. Communication Research, 20,
155-175.
Gunter, B.(1987). Poor reception: Misunderstanding and forgetting broadcast
news. Hillsdale, NJ: Lawrence Erlbaum.
Hazlett, R. L., & Hazlett, S. Y. (1999). Emotional response to television com
-
mercials: Facial EMG vs. self-report. Journal of Advertising Research,
39(2), 7-23.
Hietanen, J. K., Surakka, V., & Linnankoski, I. (1998). Facial electromyo-
graphic responses to vocal affect expressions. Psychophysiology, 35,
530-536.
James, W. (1884). What is an emotion? Mind, 9, 188-204.
649
Bolls et al. • The Effects of Message Valence
Jancke, L., Vogt, J., Musial, F., Lutz, K., & Kalveram, K. T. (1996). Facial EMG
responses to auditory stimuli. International Journal of Psychophysiology,
22, 85-96.
Lang, A. (1995). What can the heart tell us about thinking? In A. Lang (Ed.),
Measuring psychological responses to media (pp. 99-112). Hillsdale, NJ:
Lawrence Erlbaum.
Lang, A. (2000). The limited capacity model of mediated message processing.
Journal of Communication, 50(1), 46-70.
Lang, A., & Bolls, P. D. (1995). Memory for emotional television messages:
Arousal, valence and capacity. Paper presented at the annual conference
of the International Communication Association, Albuquerque, NM.
Lang, A., Bolls, P. D., Potter, R. F., & Kawahara, K. (1999). The effects of pro
-
duction pacing and arousing content on the information processing of tele
-
vision messages. Journal of Broadcasting and Electronic Media, 43(4),
451-475.
Lang, A., Dhillon, K., & Dong, Q. (1995). The effects of emotional arousal and
valence on television viewers’ cognitive capacity and memory. Journal of
Broadcasting and Electronic Media, 39, 313-327.
Lang, A., & Friestad, M. (1993). Emotion, hemispheric specialization and
visual and verbal memory for television messages. Communication
Research, 20, 647-670.
Lang,A., Newhagen, J. & Reeves, B.(1996). Negative video as structure: Emo-
tion, attention, capacity, and memory. Journal of Broadcasting and Elec-
tronic Media, 40, 460-477.
Lang, P. J. (1979). A bio-informational theory of emotional imagery.
Psychophysiology, 16(6), 495-512.
Lang, P. J., Bradley, M. M., & Cuthbert, M. M. (1997). Motivated attention:
Affect, activation and action. In P. J. Lang, R. F. Simons, & M. T. Balaban
(Eds.), Attention and orienting: Sensory and motivational processes
(pp. 97-136). Hillsdale, NJ: Lawrence Erlbaum.
Lang, P. J., Greenwald, M., Bradley, M. M., & Hamm, A. O. (1993). Looking at
pictures: Evaluative, facial, visceral, and behavioral responses. Psycho-
physiology, 30, 261-273.
LeDoux, J. E. (1989). Cognitive-emotional interactions in the brain. Cogni
-
tion and Emotion, 3(4), 267-289.
Marshall-Goodell, B. S., Tassinary, L. G., & Cacioppo, J. T. (1990). Principles of
bioelectrical measurement. In J. T. Cacioppo & L. G.Tassinary (Eds.),Prin
-
ciples of psychophysiology: Physical, social, and inferential elements
(pp. 113-148). New York: Cambridge University Press.
Mehrabian, A., & Russell, J. A. (1974). An approach to environmental psychol
-
ogy. Cambridge, MA: MIT.
Morris, J. D. (1995). Observations: SAM the self assessment mannequin—An
efficient cross-cultural measurement of emotional response. Journal of
Advertising Research, 35(6), 63-68.
Newhagen, J. (1993). Psychological instrumentation of memory and yes-no
recognition. College Park, MA: University of Maryland.
650
COMMUNICATION RESEARCH • October 2001
Newhagen, J., & Reeves,B. (1992). This evening’s bad news: Effects of compel
-
ling negative television news images on memory. Journal of Communica
-
tion, 42, 25-41.
Osgood, C., Suci, G., & Tannenbaum, P. (1957). The measurement of meaning.
Urbana: University of Illinois.
Potter, R. F., & Callison, C. (2000). Sounds exciting! The effects of audio com
-
plexity on listeners’ attitudes and memory for radio promotional
announcements. Journal of Radio Studies, 7, 29-51.
Reeves, B., Lang, A., Thorson, E., & Rothschild, M. (1988). Hemispheric
lateralization and the processing of emotional television scenes. Human
Communication Research, 15, 493-508.
Reeves, B., Newhagen, J., Maibach, E., Basil, M., & Kurz, K. (1989). Negative
and positive television messages: Effects of message type and message
context on attention and memory. American Behavioral Scientist, 34,
679-694.
Reeves, B., Thorson, E., Rothschild, M., McDonald, D., Hirsch, J., & Goldstein,
R. (1985). Attention to television: Intra stimulus effects of movement and
scene changes on alpha variation over time. International Journal of Neu
-
roscience, 25, 241-255.
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and
Social Psychology, 36, 1161-1178.
Shapiro, M., & Rieger, R. (1989). Comparing positive and negative political
advertising. Paper presented at the annual conference of the International
Communication Association, San Francisco.
Shoemaker, P. J. (1996). Hardwired for news: Using biological and cultural
evolution to explain the surveillance function. Journal of Communication,
46(3), 32-47.
Thorson, E., & Friestad, M. (1989). The effects of emotion on episodic memory
for television commercials. In A. Tybout & P. Cafferata (Eds.), Advertising
and consumer psychology (pp. 305-325). New York: Lexington Press.
Wexler, B. E., Warrenburg, S., Schwartz, G. E., & Janer, L. D. (1992). EEG and
EMG repsonses to emotion-evoking stimuli processed without conscious
awareness. Neuropsychologia, 30, 1065-1079.
Zajonc, R. B. (1984). The interaction of affect and cognition. In K. R. Scherer &
P. Ekman (Eds.), Approaches to emotion (pp. 239-246). Hillsdale, NJ: Law
-
rence Erlbaum.
651
Bolls et al. • The Effects of Message Valence
