1-s2.0-S0167876001001271-main

Ž .International Journal of Psychophysiology 41 2001 155�168

The specificity of infant emotional expression foremotion perception

Gottfried Spanglera,�, Susanne Emlinger b, Joerg Meinhardtb,Alfons Hammc

aDepartment of Psychology, Uni�ersity of Giessen, Otto-Behaghel-Strasse 10�F, Giessen D-35394, GermanybInstitute of Psychology, Uni�ersity of Regensburg, Regensburg, GermanycInstitute of Psychology, Uni�ersity of Greifswald, Greifswald, Germany

Received 9 February 2000; received in revised form 3 December 2000; accepted 21 December 2000

Abstract

Attachment theory emphasizes the role of negative emotional expression in infancy for establishing proximity toŽ .and care of the caregiver. According to Lang’s biphasic model of emotions protective reflexes e.g. startle response

are primed if a defensive motivational set is activated. The aim of the study was to examine whether the perceptionof an infant emotional expression can prime such defensive behavior. The sample consisted of 48 university students.Startle reflex, corrugator and zygomatic EMG activity and subjective ratings of valence and arousal were assessed asa response to presentation of pictures of different emotional valence. Affective startle modulation was obtained whenprobes were presented during pictures of the International Affective Pictures System replicating previous findings. Bycontrast, negative infant emotion pictures did not prompt an augmentation of the startle response, although both thesubjective ratings and the mimic EMG activity indicated a clear differentiation between negative and positive infantpictures. This pattern of findings was found only in a between-subject design, but not when the two picture sets werepresented in the same session, indicating an interference of contrasting content of pictures. � 2001 Elsevier ScienceB.V. All rights reserved.

Keywords: Perception of infant emotion; IAPS; Startle response; Mimic responses

� Corresponding author. Tel.: �49-641-99-26060; fax: �49-641-99-26079.Ž .E-mail address: [email protected] G. Spangler .

0167-8760�01�$ - see front matter � 2001 Elsevier Science B.V. All rights reserved.Ž .PII: S 0 1 6 7 - 8 7 6 0 0 1 0 0 1 2 7 - 1

( )G. Spangler et al. � International Journal of Psychophysiology 41 2001 155�168156

1. Introduction

Infant emotional expression plays an importantrole for development, because during the firstyear before the onset of speech and even moreduring the first months before the up-coming ofthe first gestures, emotional expression and mostof all negative emotional expression is the onlyway for an infant to communicate his�her needsto the caregiver. Thus, infant emotional expres-sion is a substantial component of the infant-caregiver interaction. In the same way, substantialare the caregiver’s responses to infant emotionalexpression, in particular the ability of the care-giver to perceive these emotional signals and toadequately interpret them, which should enablehim�her for an appropriate caregiver behavior.

ŽAccording to attachment theory Bowlby, 1969;.Ainsworth et al., 1978 negative emotional expres-

sion or crying in infants is part of the infantattachment behavior system, which would be acti-vated when the infant is distressed or anxious.The function of attachment behaviors is the acti-vation of the complementary caregiving system ofthe mother or father in order to gain their sup-

Žport and comfort for emotional regulation e.g..Spangler et al., 1994 .

There is plenty of empirical evidence regardingthe function and organization of the infant at-

Žtachment system e.g. Ainsworth and Wittig, 1969;Main and Solomon, 1990; Spangler and Gross-

.mann, 1993 . Moreover, there is strong empiricalsupport for the important role of the caregiver’sbehavior on the quality of infant-caregiver attach-

Žment in terms of maternal sensitivity Ainsworthet al., 1978; Grossmann et al., 1985; Spangler et

.al., 1996 . Regarding the internal regulation ofthe caregiving behavior, attachment research fo-cused on the role of attachment representationŽ .Main et al., 1985; van Ijzendoorn, 1995 . In con-trast, little is known about basic motivationalprocesses regulating the caregiver’s responses toinfant attachment behavior.

According to the biphasic emotion theory ofŽ .Lang Lang, 1995; Lang et al., 1990 emotions are

defined as action dispositions, and emotional re-sponses are assumed to be organized on the basisof two opponent motive systems, an appetitive

and an aversive system grounded on neurophysio-logical circuits. According to this approach, thevalence of a certain stimulus or a given contextdefines the general direction of behavior. Whileunpleasant stimuli activate the aversive motivatio-nal system priming defensive behavior to the levelof protective reflexes, pleasant stimuli are ex-pected to activate the appetitive behavioral setpriming approach behavior. A crucial componentof Lang’s biphasic theory is the modulatory func-tion meaning that new affective responses aremodulated by the ongoing affective valence ofbehavior or state and that this modulatory effectcan be observed on different levels of organiza-

�tion level of expressive and evaluative language,Žbehavioral level and physiological level Lang,

.�1995 . From this perspective, affective responsesare most prominent if there is a match betweenthe emotional foreground and the valence of thenew stimulus. Thus, protective reflexes are en-hanced if the organism is in an aversive motiva-tional state and are inhibited if the actual valenceof the context is positive. Similarly, positive sti-muli would activate the appetitive behavioral sys-tem stronger given a positive valence of the on-going context.

Empirical support for Lang’s biphasic theorycomes from a series of studies adopting the star-tle paradigm, in which the eye-blink response to a

Ž .probe stimulus e.g. a sudden loud noise is inves-tigated, while different affective states are evoked.From their theoretical point of view Lang andcoworkers expected that the magnitude of thestartle response would be bigger if the startle

Žoccurred in a negative context i.e. if there is a.match between stimulus and context , as com-

Ž .pared to a positive context mismatch . In theirexperiments they presented acoustic or visualstartle probes to the subjects during ongoing pre-sentation of picture slides with different emotio-nal valence. In accordance to the theoretical ex-pectations the magnitude of the eye-blink re-sponse to the startle probe was larger duringpresentation of unpleasant slides than duringneutral and pleasant slides, the latter leading to

Žthe smallest responses Vrana et al., 1988; Bradley.et al., 1988 . These findings have been replicated

Žin various laboratories Cook et al., 1992; Hamm

( )G. Spangler et al. � International Journal of Psychophysiology 41 2001 155�168 157

.and Vaitl, 1996 using aversive conditioned sti-Ž . Žmuli Hamm et al., 1993 , imagery Vrana and

.Lang, 1990 , film scenes stimuli instead of pic-Ž . Žtures Jansen and Frijda, 1994 , or odors Ehrlich-

.man et al., 1995 .The majority of studies were using picture slides

stemming from the international affective pictureŽsystem IAPS, Center for the Psychophysiological

.Study of Emotion and Attention, 1994 . Thesepictures vary widely in content and affective tone

Ž .and are organized in three affective classes: 1Žunpleasant e.g. pictures of violent death, aimed

. Ž .guns, snakes, angry or starving people ; 2 neu-Ž . Ž .tral e.g. common household objects ; and 3

Žpleasant e.g. attractive nudes, romantic couples,cuddling animals, appetizing foods and happy

.babies .Usually, broad representative selections of the

IAPS were used in most studies adopting thestartle paradigm. Moreover, in clinical studies,pictures were selected with respect to specificemotions under study. For example, in a study

Žwith high animal fearful subjects Hamm et al.,.1997 included a special set of snake and spider

Ž .pictures. According to Lang 1995 , motivationalmodulation of the startle reflex is generally robustdespite variations in specific foreground contentŽ .pg. 373 . However, in the same paper he ad-dressed the issue of dual motives, meaning thatcertain stimuli may activate both the appetitiveand the aversive motive system simultaneously.

Ž .Findings of Balaban and Taussig 1994 indicatethat within different groups of negative pictures

Žthose provoking pity e.g. pictures of babies un-dergoing extreme medical treatments or starving

.people produce somewhat smaller startle respon-ses. Perceiving such pity pictures may lead toaversive responses and at the same time mayactivate an approach disposition to help and care

Ž .for others see Lang, 1995 . These findings indi-cate that in addition to the valence of picturestimuli their specific content may play an impor-tant role, or at least that there may be specificgroups of pictures producing different patterns ofstartle responses.

The current study used this elaborated pictureperception paradigm to investigate the basicprocesses of the regulation of the caregiving sys-

Ž .tem in adults. Either pleasant e.g. smiling orŽ .unpleasant e.g. crying infant emotional expres-

sions were used as picture stimuli to activate thecaregivers’ emotional systems. As mentionedabove, it is assumed that the magnitude of thestartle response depends on whether an aversiveor appetitive motivational response system is acti-vated. While positive stimuli would activate theappetitive behavioral set, negative stimuli wouldactivate the defensive behavioral set. On the otherhand, pictures which are rated as unpleasant, donot necessarily activate a defensive response setŽ .Lang, 1995 . For example, negative emotionalexpression or crying of a baby may be rated as ahighly unpleasant stimulus. According to attach-ment theory, however, infant crying as part of theinfant attachment behavior system deserves thefunction to activate the complementary caregivingsystem of the mother or father in order to gaintheir support for emotional regulation. From thisperspective, the function of infant crying is toelicit approach behavior in the caregiver and notto activate defensive behavior.

The aim of our study was to investigate theusefulness of the startle paradigm for the study ofcaregiving behavior within an attachment theoryframework. In particular, we wanted to investi-gate whether a sample of pictures including infantemotion pictures of different emotional expres-sions would specifically modulate the startle eye-blink response. Assuming a phylogenetic base of

Ž .the caregiving system Bowlby, 1969 we expectthat specific response patterns may be identifiedin adults in general rather than only in parents.Thus, in the first step of our investigation weincluded adults not familiar with infant care, inorder to identify general response patterns. Ofcourse, there may be specific eye-blink responsepatterns in adults depending on caregiving experi-

Ž .ence parents vs. non-parents or in parents de-pending on the familiarity with the child shown inthe picture. Moreover, from an attachment per-spective, there may be different response patternsin parents with a secure attachment representa-tion as compared to those with an insecure at-tachment relationship, as these two groups wouldexhibit different sensitivity to the child’s signals

Žon the behavioral level Grossmann et al., 1988;


.van Ijzendoorn, 1995 . This issue will be ad-Ždressed in a different study Spangler et al., sub-

.mitted .

2. Method

2.1. Sample

ŽThe sample consisted of 48 students 24 male,.24 female of the University of Regensburg. The

mean age was 23.0 years ranging from 19 to 36Ž .only four subjects were older than 30 years .Subjects were randomly assigned to each of fourexperimental groups of 12 subjects each. Twogroups were used for a between-subjects compar-ison and the other two were used for a within-subjects comparison. Gender was balanced acrossthe experimental groups yielding six females andsix males for each group.

2.2. Stimulus material

The stimulus material consisted of two differ-ent sets of 60 pictures each. One set was chosenfrom the international affective picture systemŽIAPS, Center for the Psychophysiological Study

.of Emotion and Attention, 1994 based on theGerman normative valence and arousal ratings

Ž .reported by Hamm and Vaitl 1993 . The picturesselected varied widely in content and affective

Žtone regarding valence from unpleasant to pleas-. Ž .ant as well as arousal calm to arousing . Accord-

ing to the normative ratings of the IAPS therewere 20 pleasant, 20 neutral and 20 unpleasantpictures1.

The other set consisted of pictures of 9-month-Ž .old infants exhibiting a positive e.g. smile , neu-

Ž .tral or negative e.g. crying emotional expression.ŽThe pictures stem from 10 different infants five

.boys, five girls . For each infant there were twopositive, two neutral and two negative pictures.The pictures were extracted from videotapesrecorded during everyday activities at the infants’homes. The classification of the pictures as posi-tive, neutral and negative was accomplished bytwo different observers. Criteria for positive va-lence were happy expression or smiling�laughingindicated by open mouth and laughter linesaround the eyes. Crying, tears, contorted mouth,screwed-up face or reddish eyes and face werecriteria for negative emotional expression. Crite-rion for neutral pictures was the absence of anypositive and negative signs. Only such pictureswere included in which both observers chose thesame category. To draw the subjects attention to

1Pleasant�positive pictures used were: IAPS �159; �161;�200; �215; �250; �422; �461; �560; �575; �812; �162;�167; �201; �203; �251; �452; �465; �576; �582; and�813. Neutral pictures used were: IPAS �103; �123; �166;�221; �551; �700; �701; �706; �708; �751; �108; �124;�156; �220; �552; �705; �709; �715; �717; and �750.Unpleasant�negative pictures used were: IAPS �120; �212;�300; �312; �620; �661; �901; �905; �914; �916; �122;�213; �301; �313; �621; �680; �690; �900; �904; and�918.

Table 1Valence and arousal judgements for IAPS and infant pictures

Picture category IAPS-pictures Infant emotional expressions

Normative Ratings in the Ratings from Ratings in theratings current study previous study current study

ValenceŽ . Ž . Ž . Ž .Unpleasant�negative 2.45 0.86 2.68 0.81 3.38 1.24 3.40 0.45Ž . Ž . Ž . Ž .Neutral 5.14 0.46 4.71 0.56 5.95 0.96 5.74 0.42Ž . Ž . Ž . Ž .Pleasant�positive 7.19 0.36 6.46 0.73 7.00 1.19 6.83 0.37

ArousalŽ . Ž . Ž . Ž .Unpleasant�negative 6.71 0.92 5.92 1.42 5.35 1.46 5.28 0.33Ž . Ž . Ž . Ž .Neutral 4.09 1.17 3.26 1.27 3.71 1.27 3.79 0.39Ž . Ž . Ž . Ž .Pleasant�positive 4.29 0.91 4.35 0.87 4.86 1.47 4.68 0.35


the infant’s emotional expression the central fo-cus of the pictures was the child’s face in frontview or at least half-profile with the face coveringa significant part of the whole picture area. Theseinfant emotional expressions prompted the sameaffective evaluations as the standard materials

Ž .taken from the IAPS see Table 1 .

2.3. Between-subject and within-subject design

Two different designs were used for the com-parison each including two experimental groups.In the between-subject design, only infant pic-tures were presented for one of the two groupsand only IAPS pictures were presented for theother one. For the within-subject design, bothinfant and IAPS pictures were divided into twohalves, which were matched with respect to va-lence. One group received one half of the infantpictures and one half of the IAPS pictures, theother group received the complementary halves.

2.4. Procedure

Subjects were tested individually in one session.After arriving at the laboratory, the subjects wereinformed that their physiological reactions to dif-ferent pictures would be determined and thatnoise bursts, which should be ignored wouldsometimes be presented through the headphones.All subjects consented to the procedure. Thenthey were seated in a recliner in a dimly lit roomand the recording electrodes were attached. Thecomputer monitor was a distance of 1.5 m in frontof the subject’s eyes. The size of the visible pic-tures was 40�30 cm. The acoustic startle stimu-lus, which was built with the help a computer

Ž .sound editor, was a burst of broadband 95 dB Awhite noise presented binaurally through stereo-headphones for 50 ms with 0 ms rise and falltimes.

Four pictures including startle probes were thenpresented to make the subjects familiar with theprocedure. Then the experimenter left the roomand 3 min resting baseline was initiated. After-wards, the computer program was started to con-trol the physiological assessment. For each sub-ject, pictures were presented twice. During the

first phase the physiological responses were as-sessed. The second phase was for the affectivejudgements. For both designs the pictures werepresented in a balanced order with respect tovalence and � additionally for the within-sub-jects design � with respect to picture type. Threedifferent orders together with their respective re-versed orders were used resulting in six differentorders. Each was administered for two subjectsŽ .one male, one female per group. During physio-logical assessment each picture was presented for6 s. Between the pictures there was a variable

Ž .interval of between 10 and 14 s black screen . Atotal of 60 startle probes was presented during

Ž .the experiment. Two thirds of them 40 wereadministered during picture presentation with thetime point selected by chance after the first 500ms and before the last 500 ms of presentationŽ .see also Globisch et al., 1999 . One third of the

Ž .startle probes 20 were given between the pictureviewing intervals.

After the experiment which lasted approxi-mately 21 min, the electrodes were removed andthe second phase for the assessment of the rat-ings was started. For affective judgements, thesame order of pictures was used. Each picturewas presented for 6 s. Immediately after pictureoffset the subjects were asked to provide valenceand arousal ratings. These ratings were obtainedusing an interactive computer display. In the va-lence rating the subjects should rate the pleasant-ness of the picture on a 9-point scale ranging

Ž .from very unpleasant�negative �1 to veryŽ .pleasant�positive �9 . In the arousal rating they

should rate the perceived intensity on a 9-point-Ž .scale ranging from very low �1 to very high

Ž .�9 . The subjects had to mark the selectedvalue by use of a keyboard. The rating procedurelasted between 25 and 40 min.

2.5. Apparatus and physiological responsemeasurement

The experiment was controlled by two personalŽ .computers 90 MHz Pentium I . One of them

controlled picture presentation, rating assessmentand the startle stimulus, the other controlled thephysiological assessment. The pictures were pre-


Žsented via a 20-inch color computer monitor Eizo.T662-T . Physiological assessment was conducted

by using the Neuroscan SYNAMPS amplifiercombined with scan software. Startle stimuli wereproduced by a sound blaster and presented by aconnected audio amplifier via a closed stereo

Ž .headphone Quart Phone 75X .The startle reflex was measured by bipolar

recording of the activity of the orbicularis oculimuscle underneath the left eye. EMG activity wasalso recorded at the zygomatic and corrugator

Ž .regions smile and frown muscle, respectively .All facial EMG data were collected using elec-trode placements suggested by Fridlund and Ca-

Ž .cioppo 1986 . Facial skin was cleaned withrubbing alcohol and Hellige miniature AG�AgClelectrodes were applied. In addition, electroder-mal activity and heart rates were assessed which,were not included in this paper. Physiologicalsignals were continuously sampled with a samplerate of 2000 Hz. Prior to further computation theEMG-signals were bandpass-filtered offline from90 to 250 Hz.

2.6. Measures and statistical analyses

ŽFor determination of the EMG activity corru-.gator, zygomaticus, orbicularis oculi the root

mean square was calculated for defined intervalsfollowing recommendations of Fridlund and Ca-

Ž .cioppo 1986 . To assess the corrugator and zygo-matic muscle response to picture presentation,change scores were calculated by subtracting themean activity during 1 s preceding picture onsetŽ .baseline from the average response during the 6s picture viewing interval. To assess the magni-tude of the eye-blink response to the startle probechange, scores were calculated by subtracting themean activity during the 50 ms interval before thestartle probe from the mean activity between 20

Žand 100 ms after the probe onset see Globisch etal., 1999, for a detailed description of the scoring

.procedure . After preliminary distribution analy-Žses, the startle values of three subjects all from

.the within-comparison group were skipped fromanalyses as they were identified as outliers orextreme scores.

Statistical analyses were conducted separatelyfor the between-subjects and the within-subjectsdesign. Accordingly, picture type in theMANOVAs was an independent factor in thebetween-subjects design and a repeated measuresfactor in the within-subject design. In addition,valence was included as a repeated measure fac-tor and gender as an independent factor. Tocontrol for heterogeneity of variance, Green-house-Geisser corrections for degrees of freedomwere used. Nominal degrees of freedom werereported along with epsilon values. Only signifi-cant effects of MANOVAs are reported. Effectsnot addressed were not significant.

3. Results

3.1. Reliability of affecti�e judgements for IAPSpictures

To test whether computer displayed IAPS pic-tures elicited the same picture evaluation thanslides, retest correlations were calculated betweenthe German normative valence and arousal rat-

Žings of the 60 IAPS pictures Hamm and Vaitl,.1993 and the mean ratings obtained in the cur-

Ž .rent study. Retest reliability was r�0.94 forŽ .valence and r�0.82 for arousal indicating a

high correspondence. For the infant pictures,retest correlations were calculated between thevalence and arousal ratings of 36 of the 60 pic-

Žtures already used in a former study Spangler et.al., submitted and the mean ratings obtained in

Ž .the current study. Retest reliability was r�0.94Ž .for valence and r�0.83 for arousal indicating a

high correspondence. The mean valence andarousal ratings for the IAPS and infant picturesobtained in this study are presented in Table 1together with the German normative IAPS rat-ings and infant picture ratings from the Spangler

Ž .et al. submitted study. As can be seen, theŽaffective ratings are within the usual range see

.Lang, 1995 and are quite similar for the differentsamples and different picture sets.


3.2. Valence and arousal judgements of IAPS andinfant pictures

For a comparison of valence ratings of IAPSpictures and infant pictures three-way MANOVAs

Ž .with a priori valence positive, neutral, negative ,Ž .picture type IAPS, infant and the subjects’ gen-

Ž .der male, female as factors were conducted forthe valence and arousal ratings. As can be seenfrom Fig. 1 there was a similar pattern of findingsfor the between-subjects and the within-subjectsanalysis. First, with respect to the a priori valenceof pictures the valence ratings of both picture

Žtypes were as expected between-subjects: F �2,40171.45, P�0.01, ��0.67; within-subjects: F �2,44160.4, P�0.01, ��0.71. Positive pictures re-ceived the highest and negative pictures the lowestvalence ratings, neutral pictures were in-betweenŽ .Duncan, P�0.05 . Secondly, the infant pictureswere generally rated as more pleasant than the

ŽIAPS pictures between-subjects: F �11.0, P1,20�0.01; within-subjects: F �37.59, P�0.01. As1,22indicated in Fig. 1 this seems to be specifically thecase for neutral and negative pictures. The inter-action between valence and picture type, how-ever, was significant only for the within-subjectscomparison F �9.95, P�0.01, ��0.71. It2,44should be noted, however, that nevertheless, cleardifferences depending on the a priori valence ofthe pictures were given for the infant pictures.Thirdly, in both analyses an interaction between

Žgender and valence occurred between-subjects:F �3.96, P�0.05, ��0.67; within-subjects:2,40F �2.92 P�0.10, ��0.71. Post-hoc analyses2,44showed that female subjects exhibited a widerrange of valence ratings. Positive pictures wererated more positive and negative pictures morenegative in females as compared to males.

Unpleasant and pleasant pictures were rated asŽ .more arousing than neutral contents see Fig. 1 .

The arousal ratings were larger for the negativepictures and lowest for the neutral ones, theratings for positive pictures being in-between. Thisrating pattern was obtained both in the between-group and within-group comparison. F �52.33,2,40P�0.01, ��0.90 and F �102.74, P�0.001,2,44

Fig. 1. Affective judgements of valence and arousal for posi-tive, neutral and negative infant and IAPS pictures.

��0.72, respectively. As indicated by the signifi-cant interaction between picture type and valencein the within group comparison F �23.79, P�2,440.001, ��0.97 the arousal rating of the infantpictures showed a slightly different pattern thanthose of the IAPS pictures. While unpleasantIAPS pictures ware rated as most arousing, nega-tive infant pictures obtained only slightly largerarousal judgements than positive pictures.

To test whether the exclusive presentation ofinfant or IAPS pictures, respectively, in one ses-sion as compared to a simultaneous presentationof different types in one session affected thesubjective ratings, the valence and arousal ratingswere compared using a two-way MANOVA with

Ž .valence positive, neutral, negative and type ofŽ .session same, mixed . There were no significant

effects regarding type of session indicating thatthe subjective ratings were not affected by pre-senting a different picture type. Thus, differentemotional expressions of infants prompted thesame different evaluations as standard affectivematerials both in the within and between compar-ison.


3.3. Startle reflex

For a comparison of the startle response toIAPS and infant pictures, three-way MANOVAs

Ž .with a priori valence positive, neutral, negative ,Ž . Žpicture type IAPS vs. infant and gender male,

.female as factors were conducted. The between-subjects comparison of the magnitude of the star-tle reflex revealed a significant interactionbetween picture type and valence F �4.39,2,44P�0.05, ��0.95. Simple main effect analysesrevealed that the startle response varied in theexpected way only for the IAPS pictures F �2,446.07, P�0.01, ��0.95, but not for the infantpictures F �0.60, N.S.. As can be seen from2,44Fig. 2, for the IAPS pictures the response washighest for the unpleasant ones and lowest for thepleasant ones, the neutral ones lying in-between.Their linear component was significant F �1,227.32, P�0.05. The difference was significantbetween pleasant and unpleasant as well as

Žbetween neutral and unpleasant Duncan, P�.0.05 . The startle responses to both pleasant and

neutral and unpleasant infant pictures were onthe level of the positive IAPS pictures and weresignificantly lower than the startle responses to

Ž .the neutral and negative IAPS pictures P�0.05 .Regarding the within-comparison of the startle

response no significant valence effects could beidentified indicating a lack of startle modulationin this context. Although Fig. 2 indicates a higherlevel of startle response in subjects of the within-comparison, this group difference, which is due toa somewhat higher variance in the within-group,is not significant.

Fig. 2. Magnitude of the startle response for positive, neutraland negative infant and IAPS pictures.

3.4. Facial muscle acti�ity

For the corrugator and zygomatic responseagain three-way MANOVAs with a priori valence,picture type and gender as factors were con-ducted for the between-subject and the within-subject analysis. The between-subject comparisonof corrugator activity for infant and IAPS picturesrevealed a significant main effect for valence F2,40�9.58, P�0.01, ��0.59, which was qualified bytwo-way interactions between valence and picturetype F �3.60, P�0.05, ��0.59 and between2,40valence and gender F �5.89, P�0.01, ��2,400.59. The within-subject comparison of corrugatoractivity resulted in main effects for picture typeF �9.16, P�0.01 and valence F �10.04,1,22 2,44P�0.001, ��0.54.

Valence effects occurred in both designs. AsŽ .can be seen from Fig. 3 upper panel the data

indicate � as expected � a decrease in corruga-tor activity during positive pictures and an in-crease during negative ones. This response pat-tern was obtained for both picture types in thebetween-subject design as well as in the within-subject design. Nevertheless, there are obviousdifferences with respect to picture type and de-sign. For IAPS pictures, the corrugator decreasefor positive pictures is only minimal. In contrast,there is a marked corrugator increase during neg-ative pictures which is significantly higher than

Žduring neutral and positive pictures Duncan, P.�0.05 . This pattern is almost the same for the

two different designs. Quite differently for theŽinfant pictures at least for the within-subject

.design , there is a marked corrugator decrease forpositive and also for neutral pictures which statis-tically differs significantly from the small corruga-tor increase during presentation of negative in-

Ž .fant pictures P � 0.05 . Although for thebetween-subjects design the pattern is similar, thedifferences between the values are too small to besignificant. In sum, as is expressed by the maineffects for picture type, infant pictures have a biasfor corrugator decrease, while IAPS pictures havea bias for increase. Gender effects only occurredin the between-subjects design. Post-hoc analysesindicated that the valence effect is morepronounced for the female subjects.


Fig. 3. Corrugator and zygomatic response for positive, neu-tral and negative infant and IAPS pictures.

The zygomatic responses to positive, neutraland negative infant and IAPS pictures are de-picted in the lower panel of Fig. 3. Althoughzygomatic activity seems to be higher during pre-sentation of positive pictures as compared to neg-ative pictures, both the between-subjects and thewithin-subjects comparison of zygomatic activitydid not reveal any significant effect. It should benoted, however, that there was a tendency for amain effect for picture type in the within-subjectsanalysis F �3.13, P�0.10 indicating higher1,22zygomatic responses during infant pictures ascompared to IAPS pictures which again speaksfor a bias for positive evaluation of infant pic-tures.

4. Discussion

The data of the current experiment replicateand extend earlier findings of emotional respond-ing using the picture perception paradigm. Repli-cating previous findings the startle eye-blink re-sponse was potentiated during viewing of unpleas-ant and relatively inhibited during pleasant pic-

tures. However, the current data also clearly showsthat there is not a one-to-one relationshipbetween picture evaluation and startle response.Although prompting the same pattern of valencejudgements, pictures of infant emotional expres-sions did not produce any affective startle modu-lation. This is interesting because facial muscleactivity was clearly related to the different emo-tional expressions of the infant pictures. Infantpictures elicited less corrugator activity both inthe between and the within comparison and morezygomatic activity in the mixed group. Neverthe-less, the subjects responded with a decrease ofcorrugator activity to positive pictures and withan increase to negative pictures. Thus, infantpictures exhibiting different qualities of emotio-nal expression actually are experienced also indifferent ways. This becomes obvious on the levelof subjective experience as well as on the level offacial mimic responses. Consequently, they seemto provide both positive and negative stimuli andthey cannot be regarded as equivalent stimuli foremotional perception.

There may be several explanations for thesediscrepancies between different response systems.First of all, it could be possible that for methodo-logical reasons, we failed to identify the typicalstartle responses to infant pictures, although theyactually do exist, i.e. the findings may be ex-plained by the occurrence of the type II error.Possible methodological restrictions could be dueto a sample bias or a reduced quality of dataassessment. There are some arguments speakingagainst methodological shortcomings. The sampleused in our study is a normal student samplecomparable to most of the previous studies.Moreover, the comparison of the valence andarousal ratings of the IAPS pictures in the cur-rent study with those in the norm sample revealedno substantial differences indicating the compara-bility of the sample with respect to subjectiveevaluations. Rating agreement with the normsample was very high for valence and quite highfor arousal. Although there were some differ-ences between the norm sample and the currentsample with respect to the level of valence andarousal rated by the subjects, these differenceswere quite small and did not tangle the basic


assignment of the picture groups. Finally, ap-propriate data assessment in the current samplewas indicated by the fact that within the sameexperiment we were able to replicate the basicaffective startle modulation by using IAPS pic-tures.

A second explanation for the pattern of inhib-ited startle responses for the infant pictures couldbe that infant pictures do not really representnegative stimuli for the subjects. If all infantpictures irrespective of the quality of emotionexpressed by an infant, i.e. the negative ones aswell as the positive ones, could be regarded aspositive stimuli, we would not expect differencesin the magnitude of the startle response and theresponse level should be on the usual level ofpositive pictures. Consequently, we would expectthe pattern of response actually observed in thisstudy. This objection was clearly contradicted byboth the subjects’ responses on the evaluative

Ž .level subjective judgements of infant picturesŽand the behavioral level mimic responses to in-

.fant pictures . Although infant pictures in generalwere rated more positive than IAPS pictures,these differences were less than 1 point on the

Žrating scale for positive and negative pictures see.Table 1 and, thus were quite small as compared

to the mean difference between positive and neg-ative pictures being 4�5 times bigger. That is, theinfant picture ratings were within the usual limitsand there were clear-cut and marked differencesbetween positive, neutral and negative infant pic-tures comparable to the differences betweenpleasant, neutral and unpleasant IAPS pictures.

Thirdly, although at the first sight the findingstangle basic assumptions of Lang’s bi-phasic emo-tion theory, they nevertheless, can be explained

Ž .within the theory. As described by Lang 1995 ,there may be stimuli simultaneously activatingthe appetitive and the aversive motivational sys-tem. Specifically, this can be the case in emotio-nal stimuli not only expressing a certain quality ofemotion but additionally possessing specific socialfunction. According to Balaban and TaussigŽ .1994 , stimuli provoking pity actually may have anegative valence and, thus would activate theaversive behavioral set. Moreover, at the sametime they also may activate the approach system,

which leads to somewhat reduced startle respon-Ž .ses. In a recent study Ito et al. 1998 , testing the

pictures of the IAPS with respect to their bivalentnature found that a bipolar structure of valence,although given for most pictures cannot be as-sumed for all of them. While there was alsoevidence for uncoupled positive or negative acti-vation, co-activation of the two motivational stateswas indicated for some pictures but could not beproved statistically. Thus, unipolar valence ratingsmay be recommended for further studies includ-ing infant pictures in order to control for valenceco-activation.

The startle responses to negative infant pic-tures could be explained in a similar way. Byexpressing negative emotions, infants not onlywould signal their negative emotional state, whichmay contribute to an unpleasant feeling of thecaregiver thus leading to an activation of theaversive behavioral set, but at the same time theyalso would communicate their need for comfortand regulation thus eliciting approach behavior intheir interaction partners. While for pictures pro-voking pity, a balanced activation of appetitiveand aversive behavioral set is assumed by LangŽ .1995 , the activation of the appetitive system maybe predominant and the activation of the aversivesystem may be minimized in subjects faced withinfant negative emotional expression.

An additional interpretation may be derivedfrom a more recent and more specific elaboration

Ž .of Lang’s emotion theory Lang et al., 1997 as-suming that the augmentation of the startle re-sponse to negative stimuli depends on the prox-imity of threat and the imminence of attack.Accordingly, negative facial expression though ex-perienced as unpleasant may not necessarily indi-cate imminent threat or danger to the observer,as may be the case, for example, when viewing ahumilated body. Such a difference may be ex-pressed in the less negative subjective evaluationof negative infant pictures as compared to nega-tive IAPS pictures. It should be noted, however,that the startle responses to the negative infantpictures are even lower than the neutral IAPSpictures, although the subjective evaluation of thenegative infant pictures is clearly more negativethan that of the neutral IAPS pictures.


Within attachment theory, which offers an ex-planation of the development of infant-caregiverrelationships on the basis of phylogenetic assump-tions, emotional expression plays an important

Žrole for infant�caregiver interaction Bowlby,.1969 . Infant emotional expression is regarded as

an attachment behavior with the specific functionof activating the parents’ approach and caregivingbehavior in order to gain comfort and support for

Ž .emotional regulation. According to Bowlby 1969 ,the infant attachment system is pre-adapted tothe caregiving system on the side of the parentwhich is also seen as a product of a phylogeneti-cally determined developmental process. If theinfant is crying, the caregiver would approach theinfant to provide support rather than withdraw orflight from the situation. To warrant such a re-sponse, basic information processes for evalua-tion of infant signals are needed. For an ap-propriate function of negative emotional expres-sion within the attachment system informationprocessing of the caregiver leading to a negative

Ževaluation with the consequence of avoiding be-.havior would by very dysfunctional. In contrast, a

positive evaluation of infant pictures even whenexhibiting negative emotions may be regarded asa pre-requisite of appropriate caregiving behav-ior.

In terms of the startle response, one wouldexpect that the startle response is inhibited inthose persons who show appropriate caregivingbehavior. This inhibition might even be strongerduring negative emotional expressions of infantpictures although being rated as more unpleasant.According to attachment theory emotional ex-pression, in addition to its expressive function,has a social or relational function as an emotional

Ž .signal Campos et al., 1989 . Specifically, the neg-ative expression of an infant as a signal for actualneeds or felt distress should activate the caregiv-ing system of his�her attachment figure. Conse-quently, the negative valence should not activateavoiding behavior, but, in contrast, can be seen asan essential component for eliciting parental ap-proach. Thus, parental information processingproviding a positive or avoiding a negative evalua-tion of infant negative emotional behavior maysupport the function of the parents’ caregiving

system. Such processes may be reflected in or maycontribute to an inhibition of the startle responsein persons confronted with negative infant emo-tions. This specific parental responsiveness tonegative expression may have developed duringphylogeny due to its survival value for the off-spring.

While inter-individual differences in the eye-blink response were not considered in this study,attachment theory would suggest such differencesdepending on the subjects’ quality of attachmentrepresentation. Further studies are needed to in-vestigate whether this missing startle response isa general pattern or may be due to subjects with asecure attachment representation as compared tothose with an insecure representation who arecharacterized by a restricted ability for negative

Žemotional regulation Spangler and Zimmer-.mann, 1999 and, thus may experience negative

expression as more alarming and consequently,may show an augmented startle response. That is,it might be that distressed infants elicit aversivemotivations in some persons and appetitive moti-vations in others.

Incorporating these findings in Lang’s emotiontheory would require the inclusion of informationabout the content or type of emotional stimuli in

Žaddition to valence and arousal information see.Cuthbert et al., 1996 for an appropriate predic-

tion of approach or avoidance motivation. In con-trast, while the two primary opponent motivesystems may provide an appropriate organizationof emotions in a big variety of contexts, the care-giving system or some kind of social motive sys-tem could additionally be involved in emotionalregulation in interaction contexts. Such additionalmotivational processes may not be independentof the primary motive systems. Rather they maybe based on or may function with the help ofthem.

Of course, before a final evaluation of thespecificity of infant emotions for emotional per-ception, these findings must be replicated in fur-ther studies. Actually, a first successful replicationcomes from a different study in our laboratorywhere we presented infant emotional pictures to

Ž .mothers and fathers Spangler et al., submitted .Comparable to the current findings the parents


did not exhibit different startle responses to posi-tive and negative infant pictures, while the facialresponses showed again the same response pat-tern during viewing negative and positive infantpictures as in the current study. Moreover, moth-ers, particularly when looking at pictures of theirown infants, exhibited a corrugator decrease evenwhen faced with negative emotional expression.This replication may provide some confidence forthe validity of our null results regarding infantemotional expression. Taken together, thesefindings indicate the potential usefulness of thestartle paradigm for the assessment of motivatio-nal states in parents.

Whereas, in our between-subjects analyses thefindings were according to our theoretical as-sumptions, the results were quite different in thewithin-subjects analysis. Including the two differ-

Žent types of pictures both IAPS and infant pic-.tures simultaneously in one session the linear

change pattern of the startle response could notbe identified even for the IAPS pictures. The onlydifference between the between-subjects and thewithin-subjects analysis was that in the first de-sign, subjects received homogeneous sets of pic-

Žtures with respect to picture type either infantpictures of different valence or IAPS pictures of

.different valence , while in the latter one subjectsŽreceived different types of pictures infant pic-

.tures and IAPS pictures during the same session.This finding may be explained by an interfer-

ence effect in the mixed group brought about bythe combined presentation of the two contrastingtypes of pictures. The discrepancy between theinfant pictures and the partly quite terrifying neg-

Žative pictures of the IAPS e.g. pictures of violent.death, starving people etc. may activate serious

consternation at least in some of the subjects orsome of the subjects may draw associationsbetween the infants and the terrifying stimuli.This may lead to a generally quite ambivalentemotional foreground throughout the whole ses-sion, which could have masked differencesbetween pictures of different type and differentvalence. This assumption is supported by the

Ž .somewhat but not statistically proved higheroverall level startle responses in the mixed com-parison groups. Of course, these are post-hoc

interpretations and a final decision for either theperson interpretation or the context interpreta-tion is not possible at the moment. The findingsmust be subject to replication efforts and if repli-cation was successful, systematic variation of con-trasting picture types must be tested.

In this study, gender effects were found for thesubjective ratings and the corrugator activity. Bothfor the within-comparison and the between-com-parison the valence effect was more pronouncedfor females than for males, i.e. females ratednegative stimuli more negative and positive sti-muli more positive. The same was true for thecorrugator response of the between-subjects com-parison groups where females exhibited highercorrugator activity during negative pictures andlower corrugator activity during positive pictures.This is in line with existing empirical evidencesuggesting females to exhibit a broader range ofexpressivity and a higher level of arousal in theirjudgements of emotional stimuli as compared to

Žmales Bradley et al., 1993; Spangler et al., sub-.mitted and accordingly, to show a higher activity

Žof mimic facial muscles Bradley et al., 1993;Schwartz et al., 1980; Spangler et al., submitted;

.see also Schupp et al., 1997 .In sum, the findings indicate an inhibition of

the startle response to pictures of infants exhibit-ing negative emotional expression, although sub-jectively rated as negative. These data suggestthat stimuli although being negatively evaluateddo not always prompt defensive behavior. By con-trast, negative emotional expressions of infantsmight activate an approach system or activate aconflict between appetitive and defensive behav-ioral sets. Accordingly, the startle eye-blink is notpotentiated during negative infant emotional ex-pression. Moreover, the study presents importantfindings demonstrating that this method might beuseful in assessing motivational states in parentsand trying to relate this to processes involved inthe development of parent�infant attachment.

Acknowledgements

We are very indebted to Prof. Reinhard Pekrunfrom the University of Regensburg for providing


the technical resources of his psychophysiologicallaboratory and to Tobias Engelschalk for pro-gramming the software for controlling the dataassessment. We would like to thank our studentsfor their participation.

References

Ainsworth, M.D.S., Wittig, B.A., 1969. Attachment and theexploratory behavior of 1-year-olds in a strange situation.

Ž .In: Foss, B.M. Ed. , . Determinants of Infant Behavior, 55.Methuen, London, pp. 113�136.

Ainsworth, M.D.S., Blehar, M.C., Waters, E., Wall, S., 1978.Patterns of Attachment. A Psychological Study of theStrange Situation. Erlbaum, Hillsdale NJ.

Balaban, M.T., Taussig, H.N., 1994. Salience of fear�threat inthe affective modulation of the human startle blink. Biol.Psychol. 38, 117�132.

Bowlby, J., 1969. Attachment and Loss. Attachment, BasicBooks, New York.

Bradley, M.M., Cuthbert, B.N., Lang, P.J., 1988. Lateral pre-sentation of acoustic stimuli in a varying affective fore-

� �ground Abstract . Psychophysiology 25, 436.Bradley, M.M., Greenwald, M.K., Hamm, A.O., 1993. Affec-

¨tive picture processing. In: Birbaumer, N., Ohman, A.Ž .Eds. , The Structure of Emotion: Psychophysiological,Cognitive and Clinical Aspects. Hogrefe and Huber Publsi-hers, Toronto, pp. 48�65.

Campos, J.J., Campos, R.G., Barrett, K.C., 1989. Emergentthemes in the study of emotional development and emotionregulation. Dev. Psychol. 25, 394�402.

Center for the Psychophysiological Study of Emotion andŽ .Attention. 1994 . The international affective picture sys-

Ž .tem photographic slides . Center for Research in Psy-chophysiology, Gainesville, FL.

Cook III, E.W., Davis, T.L., Hawk Jr., L.W., Spence, E.L.,Gautier, C.H., 1992. Fearfulness and startle potentationduring aversive visual stimuli. Psychophysiology 29,633�645.

Cuthbert, B.N., Bradley, M.M., Lang, P.J., 1996. Probing pic-ture perception: activation and emotion. Psychophysiology33, 103�111.

Ehrlichman, H., Brown, S., Zhu, J., Warrenburg, St., 1995.Startle-reflex modulation during exposure to pleasant andunpleasant odors. Psychophysiology 32, 150�154.

Fridlund, A.J., Cacioppo, J.T., 1986. Guidelines for humanelectromyographic research. Psychophysiology 23, 567�589.

¨Globisch, J., Hamm, A.O., Esteves, F., Ohman, A., 1999. Fearsappears fast: temporal course of startle reflex potentiationin animal fearful subjects. Psychophysiology 36, 66�75.

Grossmann, K., Grossmann, K.E., Spangler, G., Suess, G.,Unzner, L., 1985. Maternal sensitivity and newborns orien-tation responses as related to quality of attachment in

Ž .northern Germany. In: I. Bretherton and E. Waters Eds. ,Growing Points in Attachment Theory and Research. Mon.

Ž .Soc. Res. Child Dev. 50 1�2 , 233�278.

Grossmann, K., Fremmer-Bombik, E., Rudolph, J., Gross-mann, K.E., 1988. Maternal attachment representations asrelated to child-mother attachment patterns and maternalsensitivity and acceptance of her infant. In: Hinde, R.A.,

Ž .Stevenson-Hinde, J. Eds. , Relations within Families. Ox-ford University Press, Oxford, pp. 241�260.

Hamm, A.O., Vaitl, D., 1993. Emotionsinduktion durch vi-suelle Reize: validierung einer stimulationsmethode aufdrei reaktionsebenen. Psychol. Rundschau 44, 143�161.

Hamm, A.O., Vaitl, D., 1996. Affective learning: awarenessand aversion. Psychophysiology 33, 698�710.

Hamm, A.O., Cuthbert, B.N., Globisch, J., Vaitl, D., 1997.Fear and the startle reflex: blink modulation and au-tonomic response patterns in animal and mutilation fearfulsubjects. Psychophysiology 34, 97�107.

Hamm, A.O., Greenwald, M.K., Bradley, M.M., Lang, P.J.,1993. Emotional learning, hedonic change and the startleprobe. J. Abnorm. Psychol. 102, 453�465.

Ito, T.A., Cacioppo, J.T., Lang, P.J., 1998. Eliciting affectusing the international affective picture system: trajectoriesthrough evaluative space. Pers. Soc. Psych. Bull. 24,855�879.

Jansen, D.M., Frijda, N., 1994. Modulation of acoustic startleresponse by film-induced fear and sexual arousal. Psy-chophysiology 31, 565�571.

Lang, P.J., 1995. The emotion probe. Studies of emotion andattention. Am. Psychol. 50, 372�385.

Lang, P.J., Bradley, M.M., Cuthbert, B.N., 1990. Emotion,attention and the startle reflex. Psychol. Rev. 97, 377�395.

Lang, P.J., Simons, R.F., Balaban, M.T., 1997. Motivated at-tention: affect, activation and action. In: Lang, P.J., Bradley,

Ž .M.M., Cuthbert, B.N. Eds. , Attention and Orienting: Sen-sory and Motivational Processes. Lawrence Erlbaum Asso-ciates, Mahwah, NJ, pp. 97�135.

Main, M., Solomon, J., 1990. Procedures for identifying infantsas disorganized�disoriented during the Ainsworth strangesituation. In: Greenberg, M.T., Cicchetti, D., Cummings,

Ž .E.M. Eds. , Attachment in the Pre-school Years. Theory,Research and Intervention. University of Chicago Press,Chicago, pp. 121�160.

Ž .Main, M., Kaplan, N. and Cassidy, J. 1985 . Security ininfancy, childhood, and adulthood: A move to the level of

Ž .representation. In: I. Bretherton and E. Waters Eds. ,Growing Points in Attachment Theory and Research. Mon.

Ž .Soc. Res. Child Dev., 50 1�2 , Ser.No. 209.Schupp, H., Cuthbert, B.N., Bradley, M.M., Birbaumer, N.,

Lang, P.J., 1997. Porbe P3 and blinks: two measures ofaffective startle modulation. Psychophysiology 34, 1�6.

Schwartz, G., Brown, S., Ahern, G., 1980. Facial muscle pat-terning and subjctive experience during affective imagery:sex differences. Psychophysiology 17, 75�82.

Spangler, G., Grossmann, K.E., 1993. Biobehavioral organiza-tion in securely and insecurely attached infants. Child Dev.64, 1439�1450.

Spangler, G., Zimmermann, P., 1999. Attachment representa-tion and emotion regulation in adolescents: a psychobiolog-


ical perspective on internal working models. Attach. Hu-man Dev. 1, 270�290.

Spangler, G., Schieche, M., Ilg, U., Maier, U., Ackermann, C.,1994. Maternal sensitivity as an external organizer forbiobehavioral regulation in infancy. Dev. Psychobiol. 27,425�437.

Spangler, G., Fremmer-Bombik, E., Grossmann, K., 1996.Social and individual determinants of attachment securityand disorganization during the first year. Inf. Ment. HealthJ. 17, 127�139.

Ž .Spangler, G., Geserick, B., and von Wahlert, A. submitted .Perception of infant emotion in mothers and fathers.

van Ijzendoorn, M.H., 1995. Adult attachment representa-tions, parental responsiveness, and infant attachment: ameta-analysis on the predictive validity of the adult attach-ment interview. Psychol. Bull. 117, 387�403.

Vrana, S.R., Lang, P.J., 1990. Fear imagery and the startleprobe reflex. J. Abnorm. Psychol. 99, 189�197.

Vrana, S.R., Spence, E.L., Lang, P.J., 1988. The startle proberesponse: a new measure of emotion? J. Abnorm. Psychol.97, 487�491.

Documents

1-s2.0-S0167876001001271-main