Childhood predictors of states of anxiety

Dialogues Clin Neurosci. 2002;4(3):287-293.

Development of the characteristics of social phobia often requires a diathesis in the form of a temperamental bias. A behavioral profile marked by vigorous motor activity and crying to unfamiliar stimuli at 4 months of age - called high reactivity- is characteristic of about 20% of healthy, Caucasian infants. This pattern predicts shy behavior in preschool children and symptoms of social anxiety at age 7, and, at age 11, a subdued personality and biological features that are consonant with a hypothesis of amygdalar excitability. The biological variables that best characterize the children who had been high-reactive infants are right-hemisphere activity in the electroencephalogram (EEC), a larger evoked potential from the inferior colliculus, higher sympathetic tone in the cardiovascular system, and larger event-related potentials to discrepant stimuli. About a quarter of 11-year-olds who had been high reactives displayed behavioral and biological characteristics that are in theoretical accord with the hypothesis of amygdalar excitability, while only 1 of 20 displayed a profile characterized by features in opposition to their temperament. The evidence points to a modest temperamental contribution to the development of symptoms currently regarded as diagnostic of social phobia.

Author Affiliations: 
Department of Psychology, Harvard University, Cambridge, Mass, USA (Jerome Kagan) 
Address for correspondence: 

There is now a consensus that chronic possession of any one of the categories of anxiety disorder is most likely for individuals who inherit a temperamental diathesis. [1] The evidence used to infer a state of anxiety in humans can include verbal report, observed behaviors, or physiology. These three categories of evidence are not highly correlated and, therefore, the meaning of “anxiety” inferred from one source of information is not equivalent to the meaning inferred from a different, source. It is important, therefore, to distinguish among four different concepts. [2]

Judged anxiety refers to verbal statements, on questionnaires or interviews, describing tension, uncertainty, or worry. However, had physiological measures been gathered on these individuals, they would not show the expected physiological accompaniments to their verbal statements. Constructed anxiety refers to a verbal report of anxiety that is accompanied by a physiological profile, but not the profile scientists assume to be theoretically appropriate. For example, an individual with an infection might, feel tense and, in an attempt to understand this feeling tone, might, decided that he or she is worried. Physiological, anxiety refers to activation of the amygdala and its projections in individuals who do not report, conscious feelings of anxiety. The fourth construct is the one most, clinicians and scientists seek to measure. The individual reports feeling worried, tense, or anxious and, in addition, displays the physiological features that. should accompany those feelings, including asymmetry of activation in the electroencephalogram (EEG) or high sympathetic tone.

Some individuals inherit a temperament that renders them especially vulnerable to the latter state of anxiety. This temperamental bias is regarded as a diathesis for the development of one or more of the psychiatric anxiety disorders. [3] It is assumed that these temperamental biases are influenced, in part, by heritable variation in the complex ncurochcmistry of the central and autonomic nervous systems. The relevant neurochemistry could include variation in λ-aminobutyric acid (GABA),corticotropinreleasing hormone, opioids, norepinephrine, and other molecules. [4]

Study design

Evaluation from age 4 months to 7 years

My laboratory has been studying longitudinally a large group of healthy, Caucasian children from middle-class families who have been followed from 4 months to 11 years: [4] Each infant was classified at 4 months of age into one of four temperamental groups based on their behavior to a standard battery of visual, auditory, and olfactory stimuli. Infants who showed a combination of frequent, vigorous motor activity combined with frequent crying were classified as high reactive (22% of the sample). Infants who showed the opposite profile of infrequent motor activity and minimal crying were classified as low reactive (40%). Infants who showed infrequent, motor activity, but frequent, crying were classified as distressed (25%), and infants who showed frequent, motor activity, but minimal distress were classified as aroused (10%). It is assumed, but not yet proven, that the high- and lowreactive groups inherit different profiles of excitability in the amygdala and/or bed nucleus and their projections. These temperamental groups are regarded as categories rather than a continuum of reactivity.

The children from these temperamental groups were evaluated twice in the second year for their reaction to unfamiliar people, situations, and procedures. The 14and 21 -month-old children who had been categorized as high reactive as infants were more likely than the low reactives to display high levels of fear to unfamiliar people, rooms, and events. [4] This relationship has been verified by Fox and colleagues, [5] who also found that 1-yearolds who had been classified as high-reactive infants at 4 months were more fearful than others when they encountered unfamiliar events. These children were observed when they were four and a half years old in a play session with two other unfamiliar children of the same sex and age, while the three mothers sat on a couch in the playroom. Each child was classified reliably as inhibited, uninhibited, or neither, based on their behavior with the other children and their reactions to two unfamiliar events that occurred after the play session. Significantly more high than low reactives were classified as inhibited. They were quiet, spent long times close to their mother, and did not initiate social interaction with the other children.

When the children were seven and a half years old, we evaluated the prevalence of anxious symptoms in 51 high réactives, 60 low reactives, and 53 children from the other two temperamental groups. The classification of anxious symptoms, which included extreme shyness, worry about, the future, fear of thunderstorms, animals, or loud noises, recurrent, nightmares, and occasional reluctance to go to school, was based on questionnaire and interview data with the mother and the child's teacher. A total of 43 of the 164 children met criteria for possession of anxious symptoms. Forty-five percent, of the children who had been high-reactive infants, compared with 15% of low réactives, had anxious symptoms (chi-square=12.8, P<0.01) [6] High reactives who had anxious symptoms were more fearful in the second year and had higher sitting diastolic blood pressures and a greater magnitude of cooling of the temperature of the fingertips across a series of digit, recall problems, compared with other high reactives. Cooling of the fingertips is the result of sympathetic inervation of the arteriovenous anastomoses under the surface of the skin.

Evaluation at age 11 years

These children were evaluated most recently when they were 11 years old. The 3-hour battery consisted of both behavioral and biological assessments. The behavioral data included the number of spontaneous comments and smiles displayed toward the examiner during the first 18 min of interaction, a reliable rating (4-point scale) of the degree of uncertainty, tension, and anxiety displayed by the child in this setting, and a maternal Qsort of 28 items describing the child's behavior.

Four different, classes of biological variables, each under the potential influence of the amygdala, were also quantified. These biological variables were: (i) asymmetry in the magnitude of dcsynchronization of alpha frequencies in the EEG; (ii) magnitude of the evoked potential from the inferior colliculus to a series of clicks; (iii) sympathetic tone in the cardiovascular system; and (iv) the magnitude of the wave form at 400 ms in the eventrelated potential to discrepant, visual scenes.

Most children and adults have less alpha power in the left than in the right, frontal area when at. rest, suggesting greater activation of cortical pyramidal neurons in the left, frontal lobe. Further, individuals with this EEG profile report more sanguine moods and fewer signs of anxiety than the smaller proportion, who show greater activation on the right, side. [7] The amygdala sends ipsilateral projections to the frontal lobe through the basal nucleus of Meynert and it is likely that these projections contribute to the asymmetry in the alpha band of the EEG. A child who had greater activation in the right amygdala should show greater desynchronization of alpha frequencies in the right, hemisphere and would be classified as right hemisphere active.

The brain stem auditory evoked response (BAER), elicited by a series of clicks delivered through earphones, was a relevant, measure because variation in the magnitude or latency of the fifth wave in the BAER response - called “wave 5” - differentiates between personality and clinical categories. [8]-[11] In addition, adults with panic disorder show a larger wave 5 than do controls. [12] The peak of the fifth wave is believed to represent, the termination of the lateral lemniscus on the inferior colliculus. [13] The theoretical relevance of this fact is that, the amygdala projects to the inferior colliculus through both the central gray and the locus cereleus and, therefore, children with a more excitable amygdala should display a larger wave 5 than others. [14]

The rationale behind recording the event-related potential to discrepant, visual stimuli derived from the assumption that the amygdala reacts to discrepant or unexpected events and projects to cortical neurons that mediate the event-related potential. [15] If high reactives possess a low threshold of reactivity in the amygdala and its projections, then they should show a larger eventrelated potential to discrepant events. The usual wave form to a discrepant, event, occurs between 1 50 and 800 ms with a peak between 300 and 400 ms. Préadolescent children most often show a negative wave form that is called Nc (for negative component). [16]

Each child was presented, through goggles, two series of pictures with 169 pictures in each series. In the first scries, 70% of the pictures were of the same item (a fireworks display), 15% were of the same flower, the oddball stimulus, and the remaining 15% were each different, but ecologically valid (a chair or kitchen utensil). These pictures were called novel valid. In the second series, the frequent picture presented 70% of the time was a yellow fire hydrant, the oddball stimulus was a different, flower, and the remaining fifteen percent of the pictures were each different, but ecologically invalid (for example, a chair with three legs). These pictures were called novel invalid.

Finally, we recorded measures of cardiovascular activity as an index of reactivity in the sympathetic nervous system. The amygdala sends varied projections to the sympathetic system and, therefore, we assumed that high reactives would show signs of greater sympathetic reactivity than low reactives. [17]

The two major variables were the ratio of high- to lowfrequency power in the cardiac spectrum while the child was laying supine. A fast Fourier transformation of the distribution of bcat-to-beat differences in the sample of resting heart rate usually reveals two peaks in the distribution. The higher frequency, around 0.2 Hz, represents the parasympathetic influence of respiration on heart ratc-vagal tone. The lower frequency band, from 0.02 to 0.10 Hz, represents both sympathetic and parasympathetic influences on heart, rate, due to cycles of change in blood pressure and body temperature. Higher relative power in the low frequency band is usually correlated with a high resting heart rate and is indicative of higher sympathetic tone. [18]



The high-reactive children were more subdued and anxious at the 11-year evaluation than the low réactives and were rated as more anxious and inhibited during the first 18 min of the interview (Table I).

Twice as many high as low reactives were rated as extremely inhibited (rating of 4; awarded to children who made very few comments and smiles, displayed a great deal of motor tension, spoke in a soft voice, and showed other signs of concern). Twice as many low as high réactives were rated as minimally anxious and uninhibited with a rating of 1, which described a maximally relaxed and spontaneous child (chi-square=11.8, P<0.01 ). Further, more high than low reactives had values for both number of spontaneous comments and smiles in the lowest quartile of the two distributions; more low réactives had values in the highest quartile for both measures (chi-square=4.2,P<0.05). Thus, the infant temperamental profiles predicted, to a modest degree, spontaneity or a subdued style with the unfamiliar adult examiner.

One half of the current group of high and low reactives had been seen when they were four and a half and seven and a half years of age. A similar rating of degree of anxiety/inhibition on a 4-point scale was assigned to each child based on 90 min of interaction with a different, but unfamiliar female examiner. Seventy percent, of the low reactives, but only 13% of the high réactives were uninhibited at all three ages; 38% of high reactives, but only 6% of low reactives were inhibited at all three ages (chisquare=21.3, P<0.001). It was rare for a low-reactive infant to become a consistently inhibited child or for a high-reactive infant, to become a consistently uninhibited child. As expected, the uninhibited profile was better preserved because family and friends encourage sociability and discourage shyness and timidity.

Seven descriptive items on the maternal Q-sort referred to shyness or sociability in the child. We computed the mean ranks the mother assigned to her child for the three shy and the four sociable items. High reactives were described as more shy and less sociable than the low réactives (r(149)=3.91, P<0.01).

Rating Low reactive High reactive
1 (relaxed, uninhibited) 50 % 22 %
2 28 % 22 %
3 8 % 22 %
4 (maximally anxious, inhibited) 14 % 34 %
Table I. Percentage of high and low reactives receiving ratings of 1, 2, 3, or 4, while interacting with the examiner at the 1 1 -year-old evaluation.

Biological variables

Figure 1 illustrates the mean standard scores for high and low reactives on the four biological variables at 11 years of age: right parietal activation in the EEG, wave 5 magnitudes, sympathetic tone (a low ratio and a high resting heart rate), and the mean of the integrated voltages from 400 to 1000 ms for the event-related potential to the first oddball flower and the novel invalid scenes. The high réactives had greater EEG activation at the right parietal site (r(152)=2.53, P<0.05). Further, the high reactives who had been highly fearful in the second year showed greater activation in the right frontal area compared with low reactives who were equally fearful in the second year. The high reactives also had significantly larger wave 5 values (r(125)=2.87, P<0.05), and this variable best separated the two temperamental groups. [19]

The high reactives also had greater sympathetic tone in the cardiovascular system; 49% of high reactives, but only 32% of low reactives, combined a low ratio in the spectral analysis with a high resting heart rate; 32% of the low réactives, but only 16% of high réactives, combined a high ratio with a low resting heart, rate (chisquare=4.9, P<0.05). It is also of interest that high sympathetic tone was the best correlate of behavior. The children with high sympathetic tone, compared to those with high vagal tone, spoke less often, were rated as more anxious, and were described by their mother as shy. Although the EEG, wave 5, and event-related potential data also separated the two temperamental groups, these measures were less closely related to the child's behavior.

One explanation is that sympathetic activity is likely to influence the orbitofrontal cortex, which mediates a conscious awareness of feeling tone. A rise in heart, rate and blood pressure and a change in breathing results in information being sent to the brain through the medulla to provoke changes in the orbitofrontal cortex that can evoke an alteration in conscious feeling. A subdued mood and avoidance behavior can be consequences of this altered feeling tone. By contrast, activity in the inferior colliculus and the pyramidal neurons of the cortex are less likely to influence orbitofrontal neurons and, therefore, no change in feeling tone occurs and there should be a minimal relationship to behavior. It is important that among high and low reactives, who were equally subdued in their behavior in the laboratory, only the high réactives showed the biological features of right parietal asymmetry and a large wave 5. The similar behaviors do not necessarily imply similar values on all biological variables. That is why it is important, for investigators and clinicians to gather biological data to supplement, their behavioral observations and interviews. Finally, the high reactives had significantly larger Nc voltages to the first, oddball picture and the novel invalid pictures (r(136)=2.00, P<0.05). Further, the correlation between the voltages and these two classes of pictures across frontal and parietal sites were always positive and significant, for high-reactive children, but not for the low reactives. That is, only high reactives showed coherence in the magnitude of the Nc across disparate cortical sites, implying that the discrepant scenes recruited neurons over a broader cortical area.

There was an interesting asymmetry in the sensitivity of low compared with high values on the four biological measurements. Low values better differentiated low from high reactives than did high values, suggesting that it is easier for low-reactive than for high-reactive children to attain a state of low cortical and autonomic arousal, even though the former can attain, temporarily, a state of higher arousal in a laboratory setting. All animals must be biologically prepared to become aroused to threat or challenge. The psychological advantages of low arousal are less obvious and apparently a smaller proportion of individuals are able to reach a state of relaxation.

Figure 1 Mean standard score for four biological variables for 11-year-old children who were classified as high or low reactive at 4 months.

Prediction of states of anxiety

About 1 in 4 children who had been high reactive and 1 in 4 children who had been low reactive developed a behavioral and a biological profile at age 11 that was in accord with theoretical expectations, while only 1 of 20 children developed a profile of social behavior and biology that, violated their expected profile. This result, is of interest in light of the varied social experiences that these children have encountered over the prior 11 years. Most children displayed behavioral and biological patterns that were characteristic of randomly selected children from middle-class, Caucasian populations. Thus, the prediction that a high-reactive infant will not be highly sociable and exuberant, and show low biological arousal at age 11 can be made with much greater confidence than the prediction that this category of child will be extremely subdued and anxious, and show signs of high arousal in cortical and autonomic targets. The suggestion that a temperamental bias constrains development more effectively than it determines particular outcomes applies to environmental conditions as well. If all one knows about a group of 1 00 children is that, they were born to economically secure, well-educated, nurturing parents and must predict the likely psychological adult, outcomes, the most accurate guesses will refer to the profiles that should not occur: criminality, school failure, psychosis, homelessness, drug addiction, and poverty. Predictions concerning the more specific features that, will be part of the adult personality are less likely to be validated. Each temperament eliminates many more possibilities than it determines. This principle holds for the cells of the young embryo. The final fate of a neural crest cell in a 3-weekold embryo, whether sensory ganglion, melanocyte, or a muscle of the heart, is less certain than the fact that this cell will definitely not become connective tissue or part of the reproductive system.


The evidence affirms the view that a temperamental bias for high reactivity in infancy, detectable early in development, is predictive of a personality profile marked by shyness, timidity, and anxiety to unfamiliar events and this behavioral phenotype is accompanied by a select biological pattern that implies amygdalar excitability. The question of greater relevance for clinicians is whether this category of child is at higher risk for any of the current psychiatric anxiety disorders. Preliminary evidence invites an affirmative reply. An independent group of 1 3-year-olds, who had been classified as inhibited or uninhibited in the second year, were interviewed by a psychiatrist who had no knowledge of their initial temperamental classification or later laboratory behavior. More of the adolescents who had been inhibited rather than uninhibited in the second year had symptoms of social anxiety (61% versus 27%). [20] However, these inhibited children were not more likely to have developed specific target phobias or separation anxiety, implying that, inhibited children might be at special risk for the development of social phobia during the adolescent or adult years.

The feared target of the social phobic is concern over the evaluations made by unfamiliar people in unfamiliar situations. By contrast, the feared target of the phobic patient is a very specific object that can harm or contaminate the agent. The feared target of the panic patient is anxiety over an unexpected autonomic surge. The targets and physiological profiles of these three groups are different and probably comprise different psychiatric categories.

Two puzzles remain. First, 20% of the children were high-reactive infants, but the prevalence of social phobia is less that. 10%. This fact suggests that many high reactives find an adaptive niche in their society that allows them to titer unpredictable social encounters. The biography of T. S. Eliot implies that he may have been a high-reactive infant, for he certainly was a shy child. His decision to become a poet permitted a degree of isolation that his temperament required.

The second fact is that more females than males are diagnosed with social phobia, although there is no excess of girls over boys who are classified as high reactive during infancy. This fact suggests that cultural ideals and differential socialization of boys and girls contribute to the sex difference in social phobia. Boys may try much harder to conquer their avoidance behavior and shyness. An excerpt from an essay written by one of the 11-year-old children, who was a high-reactive infant and a fearful toddler supports this claim.

I have always been more of an anxious person than some other people ... it took me a very long time to realize how to cope with this heightened slate of nervousness ... I have also found that the manifestation of my anxiety can be overcome by using simple mind over matter techniques. A good example of this is when I was 8, after learning about what asthma was, [started, to feel like I was having trouble breathing. In a heightened slate of anxiety, I subconsciously forced myself into believing that I had asthma. This has happened many times. Besides just general fears, it was a struggle to overcome this anxiety manifestation. I overcame these problems, though. I know how to deal with them when they occur. Because I now understand my predisposition towards anxiety, I can talk myself out of simple fears.

It is also important to note that a high-reactive temperament protects the child from engaging in risky behavior - whether drugs, driving at high speeds, or temptations for delinquent, behavior. Thus, the child with a high-reactive temperament has some advantages in our society and parents of such infants might decide not to change their child's behavior when the next set of pharmacological advances permits them that choice.

1. Regier DA, Narrow WE, Rae BS The epidemiology of anxiety disorder. J Psychiatr Res. 1999;24:3-14 [ Pub Med ]
2. Kagan J, Snidman N, McManis M, Woodward S Temperamental contributions to the affect family of anxiety. Psychiatr Clin North Am. 2001;24:677-688 [ Pub Med ]
3. Torgersen S Genetic factors in anxiety disorders. Arch Gen Psychiatry. 1983;40:1085-1089 [ Pub Med ]
4. Kagan J Galen's Prophecy. 1994
5. Fox NA, Schmidt LA, Calkins SD The role of frontal activation in the regulation and dysregulation of social behavior during the preschool years. Dev Psychopathol. 1996;80:89-102 [ Pub Med ]
6. Kagan J, Snidman N, Zentner M, Peterson E Infant temperament and anxious symptoms in school aged children. Dev Psychopathol. 1999;11:209-224 [ Pub Med ]
7. Davidson RJ Anterior electrophysiological asymmetries, emotion, and depression. Psychophysiology. 1998;35:607-614 [ Pub Med ]
8. Bullock WA, Gilliland K Eysenck's arousal theory of introversion-extroversion. J Pers Soc Psychol. 1 993;64:113-123 [ Pub Med ]
9. Stelmack RM, Wilson KG Extroversion and the effects of frequency of intensity of the auditory brain stem evoked response. Pers Individ Dif. 1982;3:373-380 [ Pub Med ]
10. Swickert RJ, Gilliland K Relationship between the brain stem auditory evoked response and extroversion, impulsivity, and sociability. J Res Pers. 1998;32:314-330 [ Pub Med ]
11. Chiappa KH Evoked Potentials in Clinical Medicine. 1997
12. Iwanami A, Isono H, Okajima Y, Kamaijima K Auditory event-related potentials in panic disorder. Eur Arch Psychiatry Clin Neurosci. 1997;247:107-111 [ Pub Med ]
13. Amaral DG, Price JL, Pitkanen A, Carmichael ST Anatomical Organization of the Primate Amygdaloid Complex in the Amygdala. 1992:1-66
14. Maisonnette SS, Kawasaki MC, Coimbra NC, Brandao ML Effects of lesions of amygdaloid nucleiand substancia nigra on aversive responses incluced by electrical stimulation of the inferior colliculus. Brain Res Bull. 1996;40:93-98 [ Pub Med ]
15. Daffner KR, Scinto LFM, Calvo V, et al. The influence of stimulus deviance on electrophysiologic and behavioral responses to novel events. J Cog Neurosci. 2000;12:393-406 [ Pub Med ]
16. de Haan M, Nelson CA Brain activity differentiates face and object processing in 6-month-old infants. Dev Psychol. 1999;35:1113-1121 [ Pub Med ]
17. Le Doux JE, Iwata J, Cicchetti P, et al. Different projections of the central amygdaloid nucleus mediate autonomic and behavioral correlates of conditioned fear. J Neurosci. 1988;8:2517-2529 [ Pub Med ]
18. Akselrod S, Gordon D, Ubell FA, Shannon DC, Barger DC, Cohen J Power spectrum analysis of heart rate fluctuation. Science. 1981;213:220-222 [ Pub Med ]
19. Woodward SA, McManis MH, Kagan J, et al. Infant temperament and the brainstem auditory evoked response in later childhood. Dev Psychol. 2001;37:1-6 [ Pub Med ]
20. Schwartz CE, Snidman N, Kagan J Adolescent social anxiety as an outcome of inhibited temperament in childhood. J Am Acad Child Adolesc Psychiatry. 1999;38:1008-1015 [ Pub Med ]