Relationship between Affective Dimensions and Physiological

Responses Induced by Emotional Stimuli

Base on Affective Dimensions: Arousal, Valence, Intensity and Approach

Eun-Hye Jang

, Mi-Sook Park

, Byoung-Jun Park

, Sang-Hyeob Kim

Myung-Ae Chung

and Jin-Hun Sohn

Biohealth IT Convergence Technology Research Department, Electronics and Telecommunications Research Institute,

Gajeongno, Yuseong-gu, Daejeon, Republic of Korea

Future Technology Research Department, Electronics and Telecommunications Research Institute,

Gajeongno, Yuseong-gu, Daejeon, Republic of Korea

Department of Psychology, Brain Research Institute, Chungnam National University,

Daehak-ro, Yuseong-gu, Daejeon, Republic of Korea

Keywords: Emotion, Affective Dimension, Physiological Responses.

Abstract: In HCI, emotion recognition using physiological signals have been noticed lately because physiological

signals can be simply acquired with some sensors and are less sensitive to social and cultural difference, in

particular, there is strong correlation between human emotional states and physiological reactions. We have

investigated the relation between affective dimensions, i.e., arousal, valence, intensity and approach, and

physiological responses such as electrocardiograph (ECG), electrodermal activity (EDA), skin temperature

(SKT), and photoplethysmograph (PPG). Three hundred college students participated in the experiment. To

successfully provoke basic emotions (anger, fear, sadness, boredom, interest, surprise, joy, pain, and

neutral), emotion-provoking film clips were excerpted for each target emotion. Physiological signals (ECG,

EDA, PPG and SKT) as emotional responses were measured during participants’ exposure to emotional

stimuli and participants were asked to rate the specific emotions they had experienced on four affective

dimensions, valence, arousal, intensity and approach. The result showed that there are correlations between

affective dimensions and physiological responses. Contrary to valence and approach, arousal and intensity

were positively related to heart rate (HR), skin conductance level (SCL) and skin conductance response

(SCR), and showed negative relation to BVP and PTT. Our result suggests an availability of physiological

signals for emotion recognition in HCI and can be helpful to provide the basis for the emotion recognition

technique in HCI.

1 INTRODUCTION

Emotions are complex processes involving multiple

response channels, including physiological systems,

facial expressions and voices (Barrett and Bliss-

Moreau, 2009; Keltner and Lerner, 2010; Larsen and

Fredrickson, 1999). Since emotional events trigger

sequences of neural activity, which result in changes

in autonomic and neuroendocrine systems (Lovallo

and Thomas, 2000; Larsen et al., 2008),

physiological responses induced by emotions have

been applied for emotion recognition in human

computer interaction (HCI). Recently, the

recognition of human’s feeling or emotion using

multi-channel physiological signals is one of the

main topics in emotional intelligence in HCI

(Wagner, Kim and Andre, 2005). Therefore, it is

very important to investigate relations between

human emotion and physiological responses. This

relation has already supported by previous studies

and may have a major influence on examination and

application of emotion recognition in HCI (Eom,

Park, Noh and Sohn, 2011).

Most emotion studies have focused on relation

between basic emotions such as happiness, sadness,

fear, etc. or affective dimension (e.g., arousal or

valence) and physiological responses (Kreibig, 2010;

Greenwald, Cook and Lang, 1989; Lang et al.,

1998). In studies using a range of affective stimuli

(perceptual or image), skin conductance increases

254

Jang E., Park M., Park B., Kim S., Chung M. and Sohn J..

Relationship between Affective Dimensions and Physiological Responses Induced by Emotional Stimuli - Base on Affective Dimensions: Arousal,

Valence, Intensity and Approach.

DOI: 10.5220/0004728302540259

In Proceedings of the International Conference on Physiological Computing Systems (PhyCS-2014), pages 254-259

ISBN: 978-989-758-006-2

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

vary directly with reports of arousal, independent of

whether the experience is reported as pleasant or

unpleasant (Bradley, Cuthbert and Lang, 1990; Cook,

Hawk, Davis and Stevenson, 1991; Greenwald,

Cook and Lang, 1989). Sequeira and colleagues

(2009) have showed a positive correlation between

arousal reports and conductance response amplitudes.

Also, emotional stimuli such as recall of both

pleasant and unpleasant memories prompts heart rate

acceleration, implying that arousal primarily

determines heart rate change (Cuthbert, Bradley,

York and Lang, 1990) and other researches have

suggested that positive emotions are associated with

greater HR than negative emotions, thus HR is

sensitive to hedonic valence (Lang et al., 1998).

A dimensional approach has focused on

identifying emotions based on a small number of

underlying dimensions (Russell, 1980) and four

dimensions such as activation, valence, potency and

intensity have been obtained in many studies to

describe subjective feeling states (Smith and

Ellsworth, 1985). The valence dimension relates to

how well one is doing at the level of subjective

experience, and ranges from unpleasant to pleasant

feelings. The activation dimension, in turn, relates to

a subjective sense of energy, and ranges from

relaxed to exciting (Rusell and Feldman Barrett,

1999). Also, approach/withdrawal dimension are

managed by two partially distinct self-regulatory

systems (Gray, 1994). Also, another dimension of

emotion is intensity which is of great importance for

the behaviour and physiological responses of an

emotion (Brehm, 1999; Sonnemans and Frijda,

1994). However, relation between physiological and

two dimensions, i.e., arousal and emotion intensity,

requires more attention, and relationships among

them and other affective dimensions are not well

understood (e.g., Laukka, Juslin and Bresin, 2010).

Researchers suggested additional dimensions to

describe a variety of emotions, since some emotions

are located closely on the dimensions.

Approach/withdrawal dimension is useful to

distinguish ‘anger’ and ‘fear’ which are next to each

other on arousal and valence axis. It has been

suggested that approach/withdrawal dimension.

Recently, motivational process is known to be

associated with prefrontal asymmetries (e.g.,

Spielberg, Stewart, Levin, Miller, Heller, 2008). In

other words, left prefrontal seemed to be linked to

approach behaviour and right prefrontal appeared to

be withdrawal behaviour from aversive stimuli.

However, there are only a few studies on approach

dimension and physiological responses. Christie &

Friedman (2004) reported that arousal dimension is

associated with skin conductance (SC) and

mean

arterial blood pressures (

MAP) and approach

dimension is linked to cardiovascular measures such

heart period (HP), diastolic blood pressure

(

DBP), and mean arterial blood pressures (MAP).

However, it is rarely examined the relationship

between emotion dimensions and physiological

signals such as photoplethysmograph (PPG) or skin

temperature (SKT) that is became famous for

emotion recognition in HCI. The study is to

investigate the relationship between affective

dimensions (i.e., arousal, valence, intensity and

approach) and various physiological responses such

as SKT and PPG.

2 EXPERIMENTAL METHODS

2.1 Participant

300 college students (140 male, mean age: 19.2±1.5)

participated in this study. None of them reported any

history of medical illness or psychotropic

medication and denied use of any medication that

would affect the cardiovascular, respiratory, or

central nervous system. Participants were

administered a hearing test, which all of them fell

within the normal hearing range. A written consent

was obtained at the beginning of the study when

they were introduced to the experimental

procedures, and they were also paid $36 USD per

session to compensate for their participation.

2.2 Emotion-provoking Stimuli

To successfully provoke target emotions, one-to

three-minute long emotion-provoking film clips,

captured originally from a variety of movies and TV

shows, were excerpted for each different emotion

(i.e., anger, fear, and sadness). The rest of the stimuli

were created to induce the target emotion

effectively. (i.e., boredom, interest, joy, neutral,

pain, and surprise). Table 1 shows the summarized

emotion induction protocols. Stimuli were

counterbalanced to minimize the effect of the order

and intensity of each stimulus.

By administering the nine stimuli to the entire

participants, the appropriateness and effectiveness of

each stimulus was tested. The appropriateness was

defined as the percentage of whether the given

stimulus properly induced the intended emotion or

not. Effectiveness was determined from the self-

report results, where the participants were asked to

rate intensity of the emotion that he or she felt for

RelationshipbetweenAffectiveDimensionsandPhysiologicalResponsesInducedbyEmotionalStimuli-Baseon

AffectiveDimensions:Arousal,Valence,IntensityandApproach

255

the stimulus on four discrete ranks. It was rated by

the participants on a 9-point Likert-type scale

ranging from -4 being “least” to +4 being “most”) on

each trial. The result of psychological responses

showed that the appropriateness was 81.81% and the

effectiveness was 81.80% in average. It means that

more than 80% participants (240 out of total 300

participants) felt the targeted emotional states and

the intensity of the emotion that they experienced

was +1.38 in average. Evidence to support the

effectiveness and appropriateness of the stimuli is

presented in the result section.

Table 1: Summary of emotion-induction protocols.

Emotion Stimulus protocol

Anger

The scenes involving repeated violence

(Korean film, 2004)

Fear

A part of a Korean horror movie, ‘A Tale

of Two Sisters (2003)’

Sadness

Scenes to address themes of death of father

from Ruler of Your Own World (Korean

Drama, 2002)

Boredom

The auditory presentation of repeating

continuously the sequence of numbers

from 1 to 10 over ten times

Interest Pictures that can create the illusion

Surprise

The presentation of the sudden noise while

participants were attended to moving visual

stimulus

Joy

The roulette playing in which participants

could earn 2~5 dollars

Pain

A standard blood pressure cuff was applied

to participant’s non-dominant arm and was

inflated to a maximum pressure of 250

mmHg

Neutral Repeated scenes involving chair

2.3 Experimental Settings

The experiment was done in a sound-proof (blocked

from noise 35dB and lower) room of 5m x 2.5m

size. A chair for a participant was located in the

center of the room and 19-inch computer monitor

was placed 50 cm ahead from the chair for

presentation of emotion-provoking stimuli. Closed

circuit television right next to the computer monitor

was installed to observe a participant’s behaviour.

There were a computer to present the stimuli to a

participant and TV and VCR to monitor and record a

participant’s behaviour outside the laboratory. Also,

a device (MP100) to measure the activities of

autonomic nervous system (ANS), and another

computer to receive all the signals from MP100

were equipped in the laboratory.

Experimental procedures were as follows: Prior

to the experiment, participants were allowed to take

time to feel comfortable in the laboratory setting and

instruction to experiment was carefully explained to

the participants. Then, electrodes for acquisition of

physiological signals were placed on their wrists,

fingers, and ankle. They had 60 seconds before the

stimulus presentation as baseline state during which

their physiological responses were measured without

any emotional stimulus. Then, they were presented

the emotion-provoking stimuli for 1~3 minutes. At

the end of stimulus presentation, participants were

asked to rate the emotions they had experienced

during exposure to emotional stimuli on four

affective dimensions, valence, arousal, approach,

and intensity. Specifically, participants were asked

the following questions to rate each dimension: 1)

did you feel ‘unpleasant’ or ‘pleasant’ during

exposure to the stimuli, 2) did you feel ‘relaxed’ or

‘aroused’ 3) did you feel like you wanted to

‘withdraw’ from or ‘approach’ the scene (or

situations you went through in boredom, surprise,

joy, and pain conditions) 4) how strong was the

emotion you experienced. The ratings were based on

scales ranging from -4 (, negatively valenced,

relaxed, withdrawal, and weak, respectively) to +4

(positively valenced, aroused, approach, strong,

respectively). After the ratings, they were given 2

minutes to get debriefed and recovered from the

emotional state. The stimulus order was randomized

for each participant and the experiment took roughly

for an hour and a half, including short breaks.

2.4 Physiological Measures and Data

Analysis

Biopac Systems Inc. (California, USA) was used to

measure physiological responses and MP100WS and

AcqKnowledge (version 3.7.1) were used to acquire

the data and analyse them, respectively. The

sampling rate was fixed at 250 Hz for all channels,

and appropriate amplification and band-pass filtering

were performed. ECG electrodes were placed on

both wrists and the left ankle using two kinds of

electrodes, sputtered and AgCl. The electrode on left

ankle was used as a reference. EDA signal was

measured with 8 mm AgCl electrodes placed on the

volar surface of the distal phalanges of the index and

middle fingers of the non-dominant hand. The

electrodes were filled with a 0.05 molar isotonic

NaCl paste to provide a continuous connection

between the electrodes and the skin. PPG sensor was

attached to the first joint of the non-dominant thumb

PhyCS2014-InternationalConferenceonPhysiologicalComputingSystems

256

and SKT signals were acquired by an SKT electrode

attached to the first joint of the non-dominant ring

finger.

Table 2: Physiological features analysed in this study.

ANS features

Abbreviatio

Perceptual attribute

Heart rate (M) HR (M)

the number of heartbeats

per minute

Blood Volume

Pulse (M)

BVP (M)

the phasic change in

blood volume

Pulse Transit

Time (M)

PTT (M)

the time it takes a pulse

wave to travel between

two arterial sites

Skin

Conductance

Level (M)

SCL (M)

the tonic levels ranging

from 10-50uS

Skin

Conductance

Response (M)

SCR (M)

the phasic change of skin

conductance or “peaks”

in the skin conductance

mean of Skin

Temperature

(M)

SKT (M)

the average temperature

of skin surface

Data for 30 sec from the baseline and another 30 sec

from emotional state by emotion-provoking stimulus

were used for the data analysis. Heart rate (HR) was

extracted from ECG measuring heart activity. Skin

conductance level (SCL) and skin conductance

response (SCR) were analysed as EDA indicators,

which represents changes in the electrical properties

of the skin due to the activity of sweat glands and is

physically interpreted as conductance. Blood volume

pulse (BVP) was extracted from PPG and pulse

transit time (PTT) was from ECG and PPG by

measuring the elapsed time between the R-wave of

the ECG and the arrival of the pulse wave at the

finger. Also, the mean SKT was obtained from the

skin temperature measurements. Table 2 shows the

autonomic nervous system (ANS) features included

in the present study. Difference between baseline

state and emotional state was used for the data

analysis. Correlation analysis was done to find the

relationship between the physiological signals and

the four affective dimensions of arousal, valence,

intensity and approach.

3 RESULTS

3.1 Participants’ Perception

of Affective Dimensions

The participants’ psychological ratings of intended

emotion were shown as descriptive statistics. Table

3 shows participants’ mean ratings of intended

emotion intensity ranged from +1.41 to +2.35 (on 9-

point scale). This means that intended emotions are

induced effectively.

Table 3: Participants’ mean ratings of intensity for each

intended emotion condition.

ANG BOR FEA INT JOY NEU PAI SAD SUR

M 2.3 1.9 2.2 1.4 1.8 0 1.7 1.6 2.4

SD 0.7 0.8 0.8 0.6 0.8 0 0.7 0.7 0.7

Abbreviations of each emotion are as follows. ANG: anger, BOR:

boredom, FEA: fear, INT: interest, JOY: joy, NEU: neutral, PAI:

pain, SAD: sadness, SUR: surprise

Figure 1: Participants’ ratings of emotion portrays on

affective dimensions.

‐4

‐2

ANG BOR FEA INT JOY NEU PAI SAD SUR

Arousal

‐4

‐2

ANG BOR FEA INT JOY NEU PAI SAD SUR

Valence

‐4

‐2

ANG BOR FEA INT JOY NEU PAI SAD SUR

Approach

RelationshipbetweenAffectiveDimensionsandPhysiologicalResponsesInducedbyEmotionalStimuli-Baseon

AffectiveDimensions:Arousal,Valence,IntensityandApproach

257

Figure 1 shows how the participants’ ratings on

each affective dimension were varied according to

the intended emotion. Participants rated that they felt

relaxed during the boredom and neutral emotion,

and aroused during anger, fear and surprise emotion.

Across levels of arousal, fear, anger and surprise

portrayed higher, and boredom and neutral lower on

arousal than the other emotions. For the valence and

approach dimensions, joy, interest, and neutral

emotions were portrayed more positively valenced

and approach and the rest of other emotions were

more negatively valenced and withdrawal.

3.2 Inter-correlation among Affective

Dimensions

Table 4 presents the correlation among the mean

ratings of each portrayal on the four affective

dimensions, i.e., arousal, valence, intensity and

approach. The results showed there were positive

correlations between arousal and intensity, and

between valence and approach. Also, arousal and

intensity were negatively related to valence and

approach. In other words, the data suggest that

emotions high in arousal and intensity tended to be

unpleasant and withdrawal.

Table 4: Inter-correlation among affective dimensions

using participants’ ratings (** p<.01).

Arousal Valence Intensity Approach

arousal 1 -.324** .289** -.237**

valence -.324** 1 -.302** .836**

intensity .289** -.302** 1 -.253**

approach -.237** .836** -.253** 1

3.3 Relationship between Affective

Dimensions and Physiological

Responses

Table 5 presents the correlation between affective

dimensions and physiological responses. Arousal

and intensity dimensions were positively related to

SCL, SCR and HR and negatively toBVP, PTT and

SKT. Contrary to the result, valence and approach

dimensions hadthe negative correlation with HR,

SCL and SCR, and the positive relation to BVP,

PTT and SKT. The negative correlation between

valence and approach dimensions with HR, SCL and

SCR means that the more unpleasant and withdrawal,

the greater changes in HR, SCL and SCRindexes.

Table 5: Correlation between affective dimensions and

physiological responses (** p<.01).

HR BVP PTT SCL SCR SKT

arousal

.17** -.27** -.29** .30** .37** -.12**

valence

-.15** .20** .15** -.23** -.27** .03

intensity

.21** -.03 -.22** .23** .24** .00

approach

-.18** .16** .14** -.24** -.28** .03

4 CONCLUSIONS

We identified the inter-correlation among four

affective dimensions, arousal, valence, intensity and

approach by the participants’ ratings and the relation

between affective dimensions and physiological

responses induced by emotional stimuli. Our results

showed that arousal and intensity have similar

properties, explaining the intensity has great

importance for the behaviour and physiological

responses of an emotion. Also, valence and approach

are affective dimensions indicating the higher scores

on them are pleasant and approach, and lower

unpleasant and withdrawal, respectively. Consistent

with Christie and Friedman (2004), our results

supports that

valence is more accurately described as

approach–withdrawal when applying autonomic

responses during discrete emotion conditions.

Results

of correlation between affective dimensions and

physiological responses indicated that two affective

dimensions of arousal and intensity are positively

related to HR, SCL and SCR, and negatively to BVP

and PTT. These results mean that the higher arousal

and intensity indicate greater sympathetic activation

reflected by HR, SCL and SCR. On the other hand,

there was negative correlation between valence and

approach dimensions with physiological responses

of HR, SCL and SCR, and positive correlation with

BVP and PTT. It shows that unpleasant and

withdrawal emotions are positively related to

sympathetic activation. Although previous

researches have reported that EDA is a marker of

cortical arousal, our results suggested the significant

relations between affective dimensions and BVP and

PTT, which means that the value of changes in BVP

and PTT could be used as indirect markers of

cortical arousal. Also, in SKT index, it was only

related to arousal dimension. Given to strong

relationships between physiological responses and

other affective dimensions, SKT is reliable less than

other physiological signals such as EDA or HR.

This study has a few limitations. Firstly, this

study only used very simple statistical analysis,

PhyCS2014-InternationalConferenceonPhysiologicalComputingSystems

258

primarily correlation method. In the further study,

the results in a more sophisticated way might be

presented instead of just showing graphs and

correlation results. If linear regression model is

applied for the data, strong statistical predictor of the

reported emotional response can be obtained based

on these physiological responses. Secondly, this

study applied relatively loose threshold on the p-

value of 0.01 for the correlation results. If more

strict and robust statistical threshold is applied for

the data, e.g., p-value of 0.002, it is expected to find

noticeable physiological responses strongly

associated with emotional dimensions.

Despite a few limitations above, this study could

examine an availability of physiological signals for

emotion recognition in HCI and our result can be

helpful to provide the basis for the emotion

recognition technique in HCI. Further, we will

expand the approach of examining other affective

dimensions such as coping potential or novelty.

ACKNOWLEDGEMENTS

This research was supported by the Converging

Research Center Program through the Ministry of

Science, ICT and Future Planning, Korea (2013K00

0329 and 2013K000332).

REFERENCES

Barrett, L. F., Bliss-Moreau, E., 2009. Affect as a

psychological primitive, Advances in Experimental

Social Psychology, vol. 41, pp. 167.

Keltner, D., Lerner, J.S., 2010. Emotion, In: Gilbert DT,

Fiske ST, Lindsay G, editors. The handbook of social

psychology, New York: McGraw Hill. pp. 312.

Larsen, R. J., Fredrickson, B. L., 1999. Measurement

issues in emotion research, In: Kahneman D, Diener

E, Schwartz N, editors. Well-being: The foundations

of hedonic psychology. New York: Gilford. pp. 40.

Lovallo, W. R., Thomas, T. L., 2000. Stress hormones in

psychophysiological research: Emotional, behavioral,

and cognitive implications. In J. T. Cacioppo, L. G.

Tassinary, & G. G. Berntson (Eds.), Handbook of

psychophysiology (2nd ed., pp. 342–367). Cambridge,

England: Cambridge University Press.

Larsen, R. J., Mercer, K. A., Balota, D. A., Strube, M. J.,

2008. Not all negative words slow down lexical

decision and naming speed: Importance of word

arousal, Emotion, vol. 8, pp. 445.

Wagner, J., Kim, J., Andre, E., 2005. From physiological

signals to emotions: Implementing and comparing

selected methods for feature extraction and

classification, IEEE International Conference on

Multimedia and Expo.

Eom, J. S., Park, H. J., Noh, J. H., Sohn, J. H., 2011.

Cardiovascular response to surprise stimulus, Korean

Journal of the Science of Emotion & Sensibility, vol.

14, pp. 147.

Kreibig, S. D., 2010. Autonomic nervous system activity

in emotion: A review, Biological psychology, vol. 84,

pp. 394.

Greenwald, M. K., Cook, E. W., Lang, P. J., 1989.

Affective judgment and psychophysiological response:

Dimensional covariation in the evaluation of pictorial

stimuli, Journal of Psychophysiology, vol. 3, pp. 51.

Lang, P. J., Bradley, M. M., Cuthbert, B. N., 1998.

Emotion, motivation, and anxiety: Brain mechanisms

and psychophysiology, Biological Psychiatry, vol. 44,

pp.1248.

Bradley, M. M., Cuthbert, B. N., Lang, P. J., 1990. Startle

reflex modification: emotion or attention?,

Psychophysiology, vol. 27, pp. 513.

Cook, E.W., Hawk, L. W., Davis, T. L., Stevenson, V.E.,

1991. Affective individual differences and startle

reflex modulation, Journal of Abnormal Psychology,

vol. 100, pp. 3.

Sequeira, H., Hot, P., Silvert, L., Delplanque, S., 2009.

Electrical autonomic correlates of emotion,

International Journal of Psychophysiology, vol. 71, pp.

50.

Cuthbert, B. N., Bradley, M. M., York, D. J., Lang, P. J.,

1990. Affective imagery and startle modulation,

Psychophysiology, vol. 27, S24.

Russell, J. A., 1980. A circumplex model of affect,

Journal of Personality and Social Psychology, vol. 39,

pp. 1161.

Smith, C. A., Ellsworth, P.C., 1985. Patterns of cognitive

appraisal in emotion, Journal of Personality and

Social Psychology, vol. 48, pp. 813.

Rusell, J.A., Barrett, F., 1999. Core affect, prototypical

emotional episodes, and other things called emotion:

Dissecting the elephant, Journal of Personality and

Social Psychology, vol. 76, pp. 805.

Gray, J. A., 1981. Three fundamental emotion systems. In

P. Ekman & R.J. Davidson (Eds.), The nature of

emotion: Fundamental questions (pp. 243–247). New

York: Oxford University Press.

Brehm, J. W., 1999. The intensity of emotion, Personality

and Social Psychology Review, vol. 3, pp. 2.

Sonnemans, J., Frijda, N. H., 1994. The structure of

subjective emotional intensity, Cognition & Emotion,

vol. 8, pp. 329.

Laukka, P., Juslin, P., Bresin, R., 2010. A dimensional

approach to vocal expression of emotion, Cognition &

Emotion, vol. 19, pp. 633.

Spielberg, J. M, Stewart, J. L., Levin, R. L., Miller, G. A.,

Heller, W., 2008, Prefrontal Cortex, Emotion, and

Approach/Withdrawal Motivation, Social and

Personality Psychology Compass, vol. 2, pp. 135-153.

Christie, I. C., Friedman, B. H., 2004, Autonomic

specificity of discrete emotion and dimensions of

affective space: a multivariate approach, International

Journal of Psychophysiology, vol. 51, pp. 143-153.

RelationshipbetweenAffectiveDimensionsandPhysiologicalResponsesInducedbyEmotionalStimuli-Baseon

AffectiveDimensions:Arousal,Valence,IntensityandApproach

259