A facial electromyographic investigation of affective contrast.
ABSTRACT Affective contrast refers to the tendency for stimuli to be judged as less evocative when preceded by more evocative same-valence stimuli. The authors used facial electromyographic (EMG) activity over corrugator supercilii, which is inversely related to affective valence, to determine if context influences underlying affective reactions. In Experiment 1, moderately pleasant pictures elicited less activity over corrugator supercilii when they were embedded among mildly pleasant, as opposed to extremely pleasant, pictures. In Experiment 2, moderately pleasant pictures elicited less activity over corrugator supercilii when they were embedded among mildly valent (i.e., pleasant and unpleasant), as opposed to extremely valent, pictures; moderately unpleasant pictures elicited comparable EMG activity regardless of context. Results indicate that context can influence affective reactions underlying affective judgments of moderately pleasant stimuli.
Conference Proceeding: Facial Electromyograhy (fEMG) activity in Response to Affective Visual Stimulation[show abstract] [hide abstract]
ABSTRACT: Recently, affective computing findings demonstrated that emotion processing and recognition is important in improving the quality of human computer interaction (HCI). In the present study, new data for a robust discrimination of three emotional states (negative, neutral and positive) employing twochannel facial electromyography (EMG) over zygomaticus major and corrugator supercilii will be presented. The facial EMG activities evoked upon viewing a standard set of pictures selected from the International Affective Picture System (IAPS) and additional self selected pictures revealed that positive pictures led to increased facial EMG activities over zygomaticus major (F (2, 471) = 4.23, p < 0.05), whereas negative pictures elicited greater facial EMG activities over corrugator supercilii (F (2, 476) = 3.06, p < 0.05). In addition, the correlation between facial EMG activities over these two sites and participants’ ratings of stimuli pictures in dimension of valence measured by Self-Assessment Manikin (SAM) was significant (r = - 0.63, p < 0.001, corrugators supercilii, r = 0.51, p < 0.05, zygomaticus major, respectively) . Our results suggest that emotion inducing pictures elicit the intended emotions and that corrugator and zygomaticus EMG can effectively and reliably differentiate negative and positive emotions, respectively.Workshop on Affective Computational Intelligence (WACI) ,IEEE Symposium Series on Computational Intelligence 2011; 04/2011
Affective Contrast 1
Larsen, J. T., & Norris, J. I. (in press). A facial electromyographic investigation of affective contrast.
A facial electromyographic investigation of affective contrast
Jeff T. Larsen
Texas Tech University
J. Ian Norris
Murray State University
Affective contrast refers to the tendency for stimuli to be judged as less evocative when preceded by more
evocative same-valence stimuli. The authors used facial electromyographic (EMG) activity over corrugator
supercilii, which is inversely related to affective valence, to determine if context influences underlying affective
reactions. In Experiment 1, moderately pleasant pictures elicited less activity over corrugator supercilii when
they were embedded among mildly pleasant, as opposed to extremely pleasant, pictures. In Experiment 2,
moderately pleasant pictures elicited less activity over corrugator supercilii when they were embedded among
mildly valent (i.e., pleasant and unpleasant), as opposed to extremely valent, pictures; moderately unpleasant
pictures elicited comparable EMG activity regardless of context. Results indicate that context can influence
affective reactions underlying affective judgments of moderately pleasant stimuli.
ADDRESS CORRESPONDENCE TO:
Jeff T. Larsen
Department of Psychology
Lubbock, TX 79409-2051
Phone: 806-742-3711 x234
Affective Contrast 2
A facial electromyographic investigation of affective contrast
People typically adapt to favorable life events and thereby derive less pleasure from them over time.
This hedonic treadmill (Brickman & Campbell, 1971) is generally thought to operate in the long term.
Brickman, Coates, and Janoff-Bulman (1978), for instance, found that lottery winners were less happy with their
lives one year after winning the lottery than they had been immediately after the event. Other research suggests
that adaptation can occur in the space of minutes rather than years. Participants in an early study by Sandusky
and Parducci (1965) rated relatively neutral odors more pleasant when they were preceded by unpleasant odors
as opposed to pleasant odors. In another study, men who were exposed to erotic pictures of especially attractive
women subsequently rated their own wives as less sexually attractive than other men rated their own wives
(Kenrick, Gutierres, & Goldberg, 1989).
Kahneman (1999) suggested that such evidence for affective contrast may be analogous to the form of
adaptation known as brightness contrast.1 A gray disk surrounded by a larger white disk is not only rated as
darker than one surrounded by a larger black disk, it is actually perceived as darker due to lateral inhibition in
the retinal ganglion cells. By this account, affective reactions are relative. That is, moderately pleasant stimuli
are judged more pleasant when embedded among less pleasant stimuli because they actually elicit more positive
affect in such contexts. As numerous theorists (e.g., Campbell, Lewis, & Hunt, 1958; Kahneman, 1999) have
pointed out, however, self-report evidence for affective contrast cannot rule out more mundane explanations
involving response biases, which are factors that systematically distort judgment (Colman, 2006). One
possibility is that participants feel compelled to use the entire response scale (e.g., extremely unpleasant to
extremely pleasant) to satisfy the conversational norm to provide informative answers. Consider an individual
presented with a series of exceptionally pleasant stimuli. Rating all of these stimuli as extremely pleasant would
convey no information about how pleasant they are relative to one another, so the individual may choose to rate
the least pleasant of these exceptionally pleasant stimuli as only mildly pleasant. (For discussions of the role of
conversational norms in survey responding, see Schwarz, 1990, 1996.) By this account, moderately pleasant
stimuli elicit comparable levels of positive affect in mildly and extremely pleasant contexts even though they are
judged more pleasant in mildly pleasant contexts.
These competing explanations for affective contrast, which we term the affective relativity and response
bias accounts, respectively, can be conceptualized in terms of Birnbaum’s (1978) model of psychophysical
judgment (see Figure 1). In Birnbaum’s model, the relationship between a physical stimulus (φ) and judgment is
mediated by an intervening psychological representation of the stimulus (ψ). Birnbaum’s model can be extended
to affective processes by treating the affective reaction to a stimulus as its psychological representation. The
affective relativity account contends that context influences what we term the affective function, which relates
1 Contrast occurs when a target stimulus is judged away from some comparison stimulus or stimuli. In a variety of well-
delineated circumstances, judgment is assimilated toward, rather than contrasted away from, the comparison (for reviews,
see Martin, Seta, & Crelia, 1990, Schwarz & Bless, 1992).
Affective Contrast 3
the stimulus to the affective reaction it elicits. Conversely, the response bias account holds that context
influences the response function, which relates the affective reaction to judgment (see Figure 1).
Figure 1. Adaptation of Birnbaum’s (1978) model of psychophysical judgment. φ and represents the physical
stimulus and ψ represents the psychological representation of the stimulus.
Testing the competing hypotheses
Manis (1971) suggested that it may be impossible to provide strong tests of the response bias and
affectively relativity accounts because both can account for contrast effects on ratings. However, some data that
speak to each account’s viability have been collected. For instance, affective contrast effects grow weaker as the
number of response options provided to participants increases from 3 to 7 to 101 (Wedell & Parducci, 1988). To
the extent that affective contrast completely disappears as the number of response options increases even further,
these data would seem to provide evidence for the response bias account.
Another strategy has been to test for contrast effects on subsequent behavior on the assumption that
contrast effects on behavior would provide evidence for affective relativity. Simpson and Ostrom (1976) asked
participants to write a paragraph describing their impression of a neutral target person after rating the target
person in a context of unlikable or likable individuals. As predicted by the affective relativity account,
participants who had rated the target in the context of unlikable individuals wrote more favorable paragraphs.
Sherman, Ahlm, Berman, and Lynn (1978), however, pointed out that context effects on behavior may
be a result of context effects on judgment. They had participants rate the importance of recycling in the context
of trivial social issues (e.g., leash laws for pets) or important social issues (e.g., capital punishment). In addition
to rating recycling more important, participants in the trivial context condition agreed to distribute more pro-
recycling pamphlets for a confederate who posed as a participant and allegedly worked at a recycling center.
Nonetheless, the effect of context on recycling behavior was only obtained among participants whose attention
had been called to their rating of the importance of recycling, which suggests that context effects on behavior
may simply reflect self-perception processes (Bem, 1972).
Evidence from choice tasks provides the most provocative evidence for affective relativity. Cooke and
Mellers (1994; cited in Mellers & Cooke, 1996) asked two groups of undergraduates to make a number of
choices between pairs of apartments in which one apartment was closer to campus but also more expensive. One
group made choices between apartments that had similar rents but varied widely in terms of proximity. In this
context, the marginal cost of being one minute closer to campus was a mere $4. The other group made choices
between apartments that varied widely in monthly rent but were similar in terms of proximity to campus. In this
context, the average marginal cost of being one minute closer to campus exceeded $50. Embedded in both
Affective Contrast 4
contexts was a subset of pairs in which the average marginal cost of being one minute closer to campus was
$12.50. Participants in the latter group preferred more of the nearby apartments in this subset of pairs,
presumably because they construed the added expense of being closer to campus as a relative bargain.
As Mellers and Cooke (1996) explained, it is unclear how such preference reversals across Cooke and
Mellers’ (1994) context conditions could be the result of response biases. Indeed, if participants in the two
groups had valued proximity equally, they would have made the same choices. The most plausible explanation
for the finding that the groups made different choices is that context affected participants’ underlying
evaluations of proximity (Mellers & Cooke, 1996). Cooke and Mellers’ (1994) results provide provocative
evidence for affective relativity, but it is important to note that they did not measure evaluative reactions to
nearby and distant apartments directly. Rather, they made inferences about whether context affected evaluative
reactions by investigating whether context affected the downstream consequences of those evaluative reactions
A psychophysiological approach
In that research involving self-reports and behavioral measures has provided indirect evidence for and
against the affective relativity account, it may be useful to turn to psychophysiological measures, which can
provide more proximal indices of evaluative and affective reactions than other tasks can (for a review, see
Larsen, Berntson, Ito, Poehlmann, & Cacioppo, in press). An event-related brain potential termed the feedback
error-related negativity (fERN) is sensitive to monetary payoffs such that penalties elicit larger fERNs than
rewards (Miltner, Braun, & Coles, 1997). Holroyd, Larsen, and Cohen (2004) demonstrated that the effect of
payoffs on fERN amplitude is context-dependent. Stimuli that informed participants that they had received no
money elicited larger fERNs when participants expected to win money than when they had expected to lose
money. These findings provide some evidence for affective relativity, but the fERN is only elicited by fairly
simple stimuli and therefore cannot shed light on affective reactions to more complex affective stimuli.
In contrast, decades of research make clear that a wide variety of affective stimuli including emotional
imagery (e.g., Schwartz, Fair, Salt, Mandel, & Klerman, 1976), films (Hess, Banse, & Kappas, 1995), and
pictures (Cacioppo, Bush, & Tassinary, 1992; Cacioppo, Petty, Losch, & Kim, 1986) influence facial
electromyographic (EMG) activity. Relative to neutral stimuli, pleasant stimuli elicit greater activity over
zygomaticus major (Lang, Greenwald, Bradley, & Hamm, 1993; Larsen, Norris, & Cacioppo, 2003), which
traverses the cheek and pulls the corners of the mouth up into a smile. The effect of positive affect tends to be J-
shaped, such that only extremely pleasant stimuli elicit activity over zygomaticus major (e.g., pictures of kittens,
puppies, babies; Lang et al., 1993; Larsen et al., 2003). In contrast, moderately pleasant pictures typically elicit
no more activity than neutral pictures. Thus, activity over zygomaticus major is insensitive to small differences
in moderate levels of positive affect. Our point is not that activity over zygomaticus major is insensitive to
positive affect. Indeed, extremely pleasant pictures elicit much more activity over zygomaticus major than
neutral pictures do. We merely point out that moderately pleasant pictures do not typically elicit much activity
over zygomaticus major. Two studies conducted by Cacioppo et al. (1986, 1992), for example, revealed that
Affective Contrast 5
moderately pleasant pictures elicited no more activity over zygomaticus major than mildly pleasant or even
mildly unpleasant pictures. Unless contextual manipulations have a pronounced effect on affective reactions to
moderately pleasant pictures, which seems unlikely, activity over zygomaticus major is not especially well-
suited for indexing whether context influences affective reactions to moderately pleasant stimuli.
Activity over corrugator supercilii, which furrows the brow into a frown, may be more useful. Activity
over corrugator supercilii is inversely related to valence such that pleasant stimuli elicit less, and unpleasant
stimuli more, activity than do neutral stimuli (Cacioppo et al., 1986; Lang et al., 1993; Larsen et al., 2003).
Moreover, the effect of valence on activity over corrugator supercilii is fairly linear. As a result, activity over
corrugator supercilii is more sensitive to small differences in moderate levels of positive affect than is activity
over zygomaticus major. In two experiments we tested the affective relativity and response bias accounts of
affective contrast by recording activity over corrugator supercilii in response to evocative pictures presented in
different contexts. We also recorded activity over zygomaticus major, which may be sensitive to affective
context in the unlikely event that affective relativity effects are especially pronounced.
Participants viewed a common series of moderately pleasant target pictures embedded in a context of
mildly pleasant or extremely pleasant pictures. In addition to asking participants to rate how pleasant they found
each picture, we measured activity over zygomaticus major and corrugator supercilii as they viewed the
pictures. Both the affective relativity and response bias accounts predict that the moderately pleasant target
pictures would be judged more pleasant when embedded in the mildly pleasant, as opposed to extremely
The two accounts make different predictions about EMG activity. According to the affective relativity
account, moderately pleasant target pictures would be judged more pleasant in the mildly pleasant context
because they actually elicit more positive affect. Thus, the affective relativity account predicts that moderately
pleasant target pictures would elicit less activity over corrugator supercilii when embedded in a context of
mildly pleasant, as opposed to mildly extremely pleasant, pictures. Activity over zygomaticus major is less
sensitive to small differences in moderate levels of positive affect (Larsen et al., 2003), but moderately pleasant
pictures might elicit more activity over zygomaticus major in the mildly pleasant context. According to the
response bias account, however, moderately pleasant target pictures would elicit no more positive affect in the
mildly pleasant context than in the extremely pleasant context. Thus, the response bias account predicts that
moderately pleasant target pictures would elicit comparable levels of activity over corrugator supercilii and
zygomaticus major in the two contexts.
Thirty-eight female psychology students at Texas Tech University participated and were compensated
with course credit. Data from four participants were removed from the analyses due to excessive movement and
EMG artifact in activity over zygomaticus major and corrugator supercilii, yielding a final sample size of 34.