Psychoneuroendocrinology, Vol. 23, No. 7, pp. 741–747, 1998© 1998 Elsevier Science Ltd. All rights reserved

Printed in Great Britain0306-4530/98 $19.99+0.00

PII: S0306-4530(98)00047-X

Short Communication

BASELINE SALIVARY CORTISOL LEVELS AND PRECONSCIOUSSELECTIVE ATTENTION FOR THREAT

A Pilot Study

Jack van Honk1, Adriaan Tuiten1, Marcel van den Hout2, Hans Koppeschaar3,Jos Thijssen3, Edward de Haan1 and Rien Verbaten1,4

1 Department of Psychonomics, Psychological Laboratory, Utrecht University, Heidelberglaan 2,3584 CS Utrecht, The Netherlands

2 Department of Experimental Abnormal Psychology, Universiteitssingel 50, Maastricht University,Maastricht, The Netherlands

3 Department of Endocrinology, University Hospital, Utrecht, Heidelberglaan 100, 3584 CX, The Netherlands4 Department of Psychopharmacology, Utrecht University. Sorbonnelaan 16, 3584 CA, The Netherlands

(Recei6ed 16 April 1998; in final form 24 July 1998)

SUMMARY

This study was conducted to examine the relationship between baseline salivary cortisol (CORT) levels andselective attention for displays of angry faces. Selective attention was investigated using a pictorial emotionalStroop task, comparing colournaming-speed of angry and neutral faces. The task was assessed in supraliminal(unmasked) and subliminal (masked) conditions to 28 non-clinical subjects (14 male and 14 female). Repeatedmeasures analysis of variance revealed a significant interaction between median split CORT levels (low vs. high)and masked face valence (angry vs. neutral). The latter effect was mainly due to significant facilitation in the highCORT subject-group; these subjects seemed to allocate their attention away from the masked angry face. Arelation between baseline CORT levels and fast withdrawal behavior is suggested. © 1998 Elsevier Science Ltd.All rights reserved.

Keywords—Emotional Stroop task; Cortisol; Selective attention; Angry faces; Submissive behavior; Preconsciousprocessing.

INTRODUCTION

A large amount of evidence has shown that anxiety increases selective attention tothreat-related material. This automatic vigilance or preoccupation with threat is suggestedto play a vital role in the causation and maintenance of anxiety disorders (Eysenck, 1997).The most frequently used paradigm for demonstrating selective attention or attentionalbias for threat is the emotional Stroop task. In this task, subjects have to name the colourof the ink in which words are printed while ignoring the meaning. The mean colournaming

Address correspondence and reprint requests to: J. van Honk, Department of Psychonomics,Psychological Laboratory, Utrecht University, Heidelberglaan 2, 3584 CS Utrecht, The Netherlands(Tel: +31 30 2533409; Fax: +31 30 2534511; E-mail: J.vanHonk@fss.uu.nl).

741

J. van Honk et al.742

latencies on threat words minus the mean colournaming latencies on neutral words, theso called interference-scores, are elevated in anxious individuals. Attention is suggestedto be allocated automatically towards the threat-value of the word. When these interfer-ence-scores turn negative the response-pattern is called facilitation. It is suggested thatsubjects then avoid the processing of threat cues or automatically orient attention awayfrom threat (Williams et al., 1996).

Selective attention for threat words has also been shown in pre-conscious conditions(Mogg et al., 1993). Pre-conscious or subliminal presentation is achieved through fastpresentation of the threatening word and imminent replacement by a pattern-mask (ameaningless string of reversed and rotated letters) in the same color (Marcel, 1983). Itseems that interference-scores in the supraliminal (unmasked) Stroop task can to someextent be (consciously) controlled by subjects. The (subliminal) masked Stroop taskprecludes this kind of control. Evidence indicates that high anxious normals showmasked interference, but can control unmasked interference. Clinically anxious subjects,however, show interference in the masked and the unmasked task. Inability to controlinterference in the unmasked task is therefore a possible indication for psychopathology(Williams et al., 1996).

The linguistic material used in emotional Stroop studies, however, has only symbolicthreat-value. A more ecologically valid type of material, giving a more direct transfer ofthreat, might be found in the angry facial expression (Bradley et al., 1997). Further-more, in most emotional Stroop studies questionnaires are used for assessing emotionalstates and traits. Given the inherent demand characteristics in these self-reports and thesuggestion that endocrinological measures may assess facets of motivational states whichcan not be revealed by self-reports (Gray et al., 1991), the level of the stress hormonecortisol (CORT) could be considered as a measure of emotional vulnerability. More-over, the recognition of emotional expressions and emotional responses to these expres-sions are assumed to be mediated by the amygdala (Rolls, 1992), a brain structurehighly sensitive to CORT. It is therefore that we investigated, in the present study, apossible relationship between baseline salivary (CORT) levels and selective attention forthreatening (angry) faces. Selective attention was measured using a masked and anunmasked version of a pictorial emotional Stroop task, comparing colournaming laten-cies of angry (masked) faces and neutral (masked) faces.

METHODS

Subjects and QuestionnairesSubjects (14 female and 14 male) were student-volunteers from the Utrecht Univer-

sity. They were medication-free, had no history of depression or anxiety-disorders, andnormal or corrected to normal vision. No habitual smokers were involved (22 non-smokers and six occasional smokers, : ten cigarettes a week). Subjects were requestednot to smoke on testing-day. Mean age was 24.6 years (SD=4.4). Subjects completedthe trait version of the (STAI) State Trait Anxiety Inventory (Spielberger, 1988). Theseself-reports were used to assess relations between anxiety-levels, CORT and task-perfor-mance. All subjects gave signed, informed consent and received a payment for participa-tion.

Cortisol and preconscious processing 743

MaterialsThe Ekman and Friesen (1976) pictures of facial affect were used as stimuli. In the

experiment ten different faces were used, each displaying a neutral and an angry expres-sion. Four copies of each model were made and subsequently coloured by placing a red,green, blue or yellow transparent folio for the picture in the slide-frame. A set of extrastimuli was prepared for practice-trails. For constructing slides of masks we randomly cut,resembled and rephotographed pictures of faces. The colours of the masks were congruentto the colours of their targets, through placement of the same transparent red, blue, greenand yellow folio’s for the picture in the slide-frames. The stimuli were back-projectedthrough a milk-coloured glass screen into the experimental room. The screen was locatedat 110 cm from the subject at eye-level. The resulting slide-image was 20×30 cm. ThreeKodak Carousels with Compur high speed shutter were used for presentation. Thepresentation and timing was regulated by Test Point software on a Laser Pentium 135personal computer. In the supraliminal condition, one trial consisted of a slide containinga fixation point (carousel 1) which was shown for 750 ms and subsequently followed by thetarget slide (carousel 2), showing the coloured neutral, or angry, face. Target presentationwas terminated at vocal response initiation. In the subliminal condition the same fixationpoint was also shown for 750 ms, but the target was exposed for only 30 ms andimmediately followed by a mask (carousel 3). Here, target-mask presentation was termi-nated at vocal response initiation. Initiation of vocal response was registered by thecomputer’s clock.

Threshold-controlBy extensive piloting in our laboratory we established an objective threshold for the

recognition of emotional expressions for our displays. These pilots indicated that a 30 msmasking interval effectively precluded recognition of the emotional valence of targets inevery subject. Nevertheless, the masking procedure was checked after finishing the Strooptasks. A two alternative forced choice (2AFC) emotional-neutral recognition procedurewas used. In this 2AFC, a random set of 30 masked faces was shown to the subjects. Inadvance they were explicitly told that 15 neutral and 15 angry faces were in the set, andinstructed to indicate (in doubt by guessing), through pushing a button, if the presentedslide was a neutral or an emotional expression.

ProcedureThe effect of circadian hormonal rhythms were minimized by conducting all sessions

between 1330 and 1730h. Just prior to the Stroop task saliva-samples were collected forCORT measurements. The cortisol level in saliva is a valid measure of unbound hormonefraction, and only unbound cortisol reaches target tissues (Kirchbaum and Hellhammer,1994). The colournaming task was carried in a special experimental room. All apparatuswere placed outside of the room to keep sound disturbance minimal. An intercom allowedsubject and experimenter to communicate. Subjects were instructed to name the colour ofthe target-slide as quick as possible, while ignoring the content. Twenty practice-trails (tenmasked and ten unmasked) had to assure subjects understood the instructions. In the mainexperiment 40 masked and 40 unmasked faces were shown in random order. In bothconditions there were 20 neutral and 20 angry faces. Masked and unmasked condition

J. van Honk et al.744

were presented blocked and these blocks were counterbalanced over subjects. Dimmedlight was used through-out the experiment.

Biochemical MeasurementSalivary CORT levels were determined without extraction at the Department of

Endocrinology of the Utrecht University using an in house competitive radio-im-munoassay (RIA) employing a polyclonal anticortisol-antibody (K7348). Following chro-matotographic verification of its purity, 1,2-3H(N)-Hydrocortisone (NET 185,NEN-Dupont-, Dreiech, Germany) was used as a tracer. Lower limit for detection is 0.5nmol/l and reference values for adults 4–28 nmol/l.

StatisticsInterference-scores were calculated (the individual mean colournaming latencies on

angry faces minus the individual mean colournaming latencies on neutral faces). Todetermine effects of CORT on task-performance, subject were divided in groups using amedian split. A repeated measures multivariate analyses of variance (MANOVA) wasconducted, using unmasked face valence (angry versus neutral) and masked face valence(angry vs. neutral) as within subjects factors. CORT (low vs. high) and gender (women vs.men) were used as between subjects factors. To this design, time (1330–1530h vs.1530–1730h) was added as between subjects factor to control for possible influences oftime of examination. Additionally, paired t-tests were used to investigate if interference orfacilitation in split CORT groups differed from zero (equal colournaming-latencies forneutral and angry faces).

RESULTS

CortisolMedian split on cortisol-levels resulted in a low cortisol subject-group with a mean level

of 6.3 nmol/l (SD 0.8; min 4.5; max 7.1), and a high cortisol subject group with a meanlevel of 10.0 nmol/l (SD 2.5; min 7.4; max 16.9). The difference in CORT levels betweenthe groups was significant (t(26)=5.3; pB .001).

AnxietyThere was no relation whatsoever between anxiety, cortisol and task-performance.

Pearsons correlations: anxiety versus cortisol (r(28)=0.09, ns), anxiety vs. unmaskedinterference (r(28)=0.07, ns) and anxiety versus masked interference (r(28)=0.01, ns).Anxiety was therefore excluded from further analyses.

AwarenessForced choice checks indicated that masked presentation precluded recognition of

emotional valence of targets. Only 50.2% of overall responses were correct (chanceperformance=50%). Upper limit was set to 19 correct scores; none of the subjects scoredabove 18. Non-parametric tests for deviation from the expected value (cut point=15),showed that deviation was not significant (binomial, two-tailed, p= .85), meaning themasking procedure was efficient.

Cortisol and preconscious processing 745

MANOVABecause there were no significant effects or interactions concerning gender and time (all

F ’sB1.2), these variables were excluded from final analyses. Subsequently, there was acrucial significant multivariate interaction between CORT and face valence (F(2.25)=4.55; pB .025). Univariate tests showed that the interaction between CORT and unmaskedface valence was not significant (F(1.26)=0.46; ns). The interaction between CORT andmasked face valence was, on the contrary, highly significant (F(2.20)=6.61; pB .02). Fig.1 shows the pattern of colournaming in both conditions on the basis of interference-scores.

As can be seen from Fig. 1, in the unmasked condition both CORT groups showinterference. Paired t-tests (one-tailed) revealed no significance for the high CORT group(t(13)=0.3; ns) and a non-significant trend for the low CORT group (t(13)=1.2; p= .12).The significant CORT× face valence interaction in the masked condition dependedprimarily on the high CORT group showing significant facilitation (t(13)= −2.9; pB .01),while the minor interference in the low CORT group was not significant (t(13)=0.7; ns).This means, that high CORT subjects were significantly faster when colournaming angrymasked faces (compared with neutral masked faces).

DISCUSSION

The results of our study showed an association between baseline salivary CORT levelsand selective attention for masked angry faces. The significant CORT× face valenceinteraction in the masked condition depended primarily on high CORT subject-groupshowing significant facilitation; they attended away from the masked angry face. How canthis finding be explained?

The use of pictorial material in motivational selective attention tasks is still in it’sinfancy. Recently, in two unpublished studies using pictorial material, the same reversedpattern (attending away rather than attending towards pictures) has been found. AnkeEhlers (personal communication) used a modified dot-probe paradigm and found thatsocially anxious subjects and social phobics direct their attention away from faces(especially from emotional faces) and Jerome Kagan (personal communication) used a

Fig. 1. Bars represent mean interference-scores in median split CORT groups in the unmasked andthe masked condition.

J. van Honk et al.746

pictorial emotional Stroop task and found extremely socially anxious girls facilitatingwhen colournaming threatening pictures.

Can causality be presumed in our findings? CORT could influence cognitive motiva-tional behavior by binding of CORT molecules on receptors of the amygdala. Interestinglyin this respect, relations have been found between CORT and shyness. High CORT levelsin children were related to cautious and socially avoidant behavior. Schmidt et al. (1997)suggest that high levels of CORT in shy children induce changes in the amygdala,exacerbating their fearfulness. Furthermore, in non-human primate infants conservation ofenergy and withdrawal from danger is called the conservation withdrawal response(Kaufman and Rosenblum, 1969). Animals in this passive mode of survival also show highbaseline CORT levels. Moreover, high baseline CORT levels are shown in primates low indominance hierarchies (Sapolsky, 1992). Danger-withdrawal suits the behavior-pattern ofthe submissive animal. Its defensive reaction to display of anger by a dominant group-member is a quick gaze-aversion (O8 hman, 1986).

We suggest that the response-pattern shown in our high CORT subject-group in maskedconditions might point at an adaptive avoidance-response to angry faces. If the role ofCORT is causal in this mechanism, this role seems highly sensitive, given the fact thatbaseline levels in our high CORT subjects were still in the normal range of values.According to Kling and Brothers (1992), automatic social responses not mediated byneo-cortical processing still form a substrate of social behavior in higher animals. Thismight give a clue to the absence of effects in the unmasked condition. As mentionedearlier, the unmasked condition is vulnerable to conscious strategies, making the maskedcondition more reliable (Williams et al., 1996); see also Kihlstrom (1987) for the notion ofconscious counter-control.

In conclusion, this study suggests that the subjects in our high CORT group orientedtheir attention away from masked angry faces. Preconscious initiated withdrawal tenden-cies or submissive behavior reflexes might be responsible for this effect. More corticalintervention could be responsible for absence of significant effects in the unmaskedcondition. Our research further indicates that salivary CORT measurements can make avaluable contribution in selective attention studies investigating motivational aspects ofbehavior.

Acknowledgements: We thank B.L. Smoor for development of computer-programs and I. Maitimu-Smeele forlaboratory work.

REFERENCES

Bradley, B. P., Mogg, K., Millar, M., Bonham-Carter, C., Ferguson, M., Jenkins, J. and Parr, M.(1997) Attentional biases and emotional faces. Cognition and Emotion 11 (1), 25–42.

Ekman, P. and Friesen, W. (1976) Pictures of Facial Affect. Consulting Psychologists Press, PaloAlto.

Eysenck, M. W. (1997) Anxiety and Cognition. Psychology Press, Hove, East Sussex.Gray, A., Jackson, P. M. and McKinlay, J. B. (1991) The relation between dominance, anger,

hormones in normally aging man: results from the Massachusetts Male Aging Study. Psychoso-matic Medicine 53, 375–385.

Kaufman, J. C. and Rosenblum, R. A. (1969) Effects of separation from the mother on theemotional behavior of infants monkeys. Annals of the New York Academy of Sciences 159,861–895.

Kihlstrom, J. F. (1987) The cognitive unconscious. Science 237, 1445–1452.

Cortisol and preconscious processing 747

Kirchbaum, C. and Hellhammer, D. K. (1994) Salivary cortisol in neuroendocrine research: recentdevelopments and applications. Psychoneuroendocrinology 19 (4), 313–333.

Kling, A. S. and Brothers, L. A. (1992) The amygdala and social behavior. In: Aggleton, J. P. (Ed.).The Amygdala. Wiley-Liss, New York, pp. 353–377.

Marcel, A. J. (1983) Conscious and unconscious perception: experiments on visual masking andword recognition. Cogniti6e Psychology 15, 197–237.

Mogg, K., Bradley, P. B., Williams, R. and Mathews, A. (1993) Subliminal processing of emotionalinformation in anxiety and depression. Journal of Abnormal Psychology 102 (2), 304–311.

O8 hman, A. (1986) Face the beast and fear the face: animal and social fears as prototypes forevolutionary analyses of emotion. Psychophysiology 23, 123–145.

Rolls, E. T. (1992) Neurophysiology and the functions of the primate amygdala. In: Aggleton, J. P.(Ed.). The Amygdala. Wiley-Liss, New York, pp. 485–503.

Sapolsky, R. (1992) Stress, the Aging Brain, and Mechanisms of Neural Death. MIT press,Cambridge.

Schmidt, L. A., Fox, N. A., Rubin, K. H., Sternberg, E. M., Gold, P. W., Schmith, C. C. andSchulkin, J. (1997) Behavioral and neuroendocrine responses in shy children. De6elopmentalPsychobiology 30 (2), 127–140.

Spielberger, C. D. (1988) Manual of the State-Trait Anxiety In6entory (STAXI). PsychologicalAssessment Resources, Odessa, FL.

Williams, J. M. G., Mathews, A. and MacLeod, C. (1996) The emotional Stroop task andpsychopathology. Psychological Bulletin 120, 3–24.

.