Specific Life Events and Chronic Experiences Differentially Associated with Depression and Anxiety in Young Twins

Specific Life Events and Chronic Experiences Differentially Associated with Depression and Anxiety in Young Twins

Thalia C. Eley

Thalia C. Eley [1,3]

Jim Stevenson [2]

Behavioral genetic analyses indicate that environmental influences associated with depression and anxiety are specific to each symptom type; however, this has not been tested specifically in children. Sixty-one (61) child twin pairs in which at least one twin had a very high anxiety or depression score, and 29 nonanxious, nondepressed pairs were interviewed about life events and chronic stressors in the previous 12 months. Loss events, schoolwork stressors, family relationship problems, and friendship problems were all significantly associated with depression but not anxiety. Threat events were significantly associated with anxiety but not depression. Loss events and schoolwork stressors appeared to act as shared environment influences in that they made twin pairs resemble one another. Threat events, friendship problems, and family relationship problems were individual specific and accounted for differences within the pairs. These results clarify the associations between life events and depressive and anxious symptoms in children and adolescents and reveal specific associations previously unidentified in this age range.

KEY WORDS: Anxiety; depression; environmental influences.


There is a growing body of evidence from behavioral genetic studies that suggests that in adults and children, the high levels of trait correlation between depression and anxiety and the high levels of comorbidity between these two types of disorder are largely due to genetic factors that influence both (Eley, 1997b; Eley & Stevenson, 1999; Kendler, Heath, Martin, & Eaves, 1987; Kendler, Neale, Kessler, Heath, & Eaves, 1992; Thapar & McGuffin, 1997). By contrast, these studies reveal that the influence of environmental factors is largely symptom or disorder specific. In trying to understand the etiology of depression and anxiety in children an adolescents, it is therefore crucial to identify environmental stressors related specifically to either anxiety or depression. Although several decades of work have demonstrated the association between life events and depression with both adult (e.g., Brown & Harris, 1978a; Finlay-Jones & Brown, 1981; Kendler & Karkowski-Shuman, 1997; Paykel et al., 1969) and child sam ples (e.g., Goodyer & Cooper, 1993; Loss, Beck, & Wallace, 1995; Tisher, Tonge, & Home, 1994), there are very few studies that have considered subjects with anxiety as a specific and separate group from those with depression, and none using child samples. However, there have been some studies that produce results relevant to the hypothesis that there are environmental stressors related to depression in children that are different from stressors associated with anxiety.

A study of adult women found that there was a specific relationship between events characterized by loss and the onset of a depressive disorder and events characterized by threat of future danger, and the onset of an anxiety disorder (Finlay-Jones & Brown, 1981). The relationship between loss events, usually parental bereavement, and emotional symptoms has also been demonstrated in several studies of children (e.g., Goodyer & Aitham, 1991; Kranzler, Shaffer, Wasserman, & Davies, 1989; Reinherz et al., 1989; Weller, Weller, Fristad, & Bowes, 1991). However, the association between anxiety and threat events has not been replicated or demonstrated in a child population.

In addition to acute life events, three broad categories of chronic stressors have been found to be associated with high levels of depressive and anxious symptoms in children and adolescents. The first is dysfunctional family relationships, including high “Expressed Emotion” within the family (Rutter & Brown, 1966) and insecure parental attachment (Armsden, McCauley, Greenburg, Burke, & Mitchell, 1990; Asamow, Goldstein, Tompson, & Guthrie, 1993; Puig-Antich et al., 1993; Schwartz, Dorer, Beardslee, Lavori, & Keller, 1990; Tejerina-Allen, Wagner, & Cohen, 1994). The second is quality of friendships, support from confidants, and self-ratings of perceived popularity (Armsden et al., 1990; Goodyer, Wright, & Altham, 1989; Parker & Asher, 1993; Reinherz et al., 1989). The third includes poor school performance and academic problems (Carlson & Cantwell, 1983; Peterson, Compas, Brooks-Gunn, Stemmler, & Grant, 1993; PuigAntich et al., 1993; Williams & McGee, 1991). However, the possibility that such stressors may b e specifically associated with one or another symptom type has not been investigated. We use the term symptom specific to refer to such environmental influences.

A second issue to emerge from the behavioral genetic literature on child psychopathology is that it appears to be those defined as nonshared environmental influences–that is, those specific to the individual–rather than shared environment influences–those that make family members resemble one another–that have the greatest influence on almost all types of emotional and behavioral symptoms (e.g., Eley, Deater-Deckard, Fombonne, Fulker, & Plomin, 1998; Pike & Plomin, 1996; Plomin & Daniels, 1987). Note that we use the term nonsha red to refer to influences whose outcome is individual specific in contrast to those whose outcome is symptom specific as described above. Unusually, for anxiety measures, shared environment influence has also been found to be a significant influence both in the normal range of scores, and for those individuals with high anxiety (Eley & Stevenson, 1999; Stevenson, Batten, & Chemer, 1992; Topoiski et al., 1997; for a review, see Eley, 1999). Furthermore, two studies of depression s ymptoms indicated shared environmental influence on high depression scores, which is not seen in the normal range (Eley, 1997a; Rende, Plomin, Reiss, & Hetherington, 1993). The environmental influences traditionally investigated in developmental studies tend to be family-based variables that have been regarded (often incorrectly) as shared risk factors influencing all members of the family. The demonstration from behavioural genetic research of the importance of the nonshared environment as well as shared environment implies that it would be worth finding ways to distinguish between shared and individual specific environmental influences.

In summary, we propose two approaches to establishing “unique” aspects of the relationship between environmental influences and outcome. First, we predicted that many aspects of the environment would be “symptom specific,” and as such would be associated with either depression or anxiety, but not both. Second, as in the tradition of behavioral genetics, we hypothesized that although many environmental influences would be shared within the family, resulting in within-pair similarity, there would also be significant associations between our environmental measures and outcome that were “nonshared” in their effect, resulting in within-pair differences. Examples of shared environmental influences might include economic variables such as poverty or housing, which would be expected to have a similar impact on different family members. In contrast, examples of nonshared environment would be events or situations that have an impact on only one individual, such as accidents or illness, and events within the context of relationships of importance only to one family member; for example, friendships.

In this study, we therefore set out to test the hypothesis that certain environmental influences would be specifically associated with either anxiety or depression scores in children and adolescents. In particular, we tested the hypothesis that loss events would be associated with depression but not anxiety, and that threat events would be associated with anxiety but not depression. We also tested whether three types of ongoing stressors–family relationship problems, friendship problems, and schoolwork or academic stressors–were specifically associated with either depression or anxiety. In order to ascertain whether such influences were part of the shared environment (i.e., make family members similar to one another) or were features of the nonshared environment (i.e., make family members different from one another), we used between-pair and within-pair comparisons in a sample of twins.

Briefly, the logic for this approach is as follows. The first stage of the analysis was to compare all probands with all controls in order to establish associations between proband status and our measures of the environment. These analyses established whether all those children scoring highly for depression or anxiety had higher mean levels of negative environmental influences than the controls regardless of the status of the co-twins. This between-pair analysis thus revealed both shared and nonshared influences on outcome. In the second stage of the analysis, a within-pair approach was taken. Mean levels of negative environmental influences were calculated for both members of discordant twin pairs. Only nonshared environmental variance would result in within-pair dissimilarity (or discordant status), and as such this analysis was focussed on identifying nonshared influences on Outcome. In summary, we used the between-pair and within-pair analyses to try to elucidate the shared and nonshared nature of the in fluences we were exploring.



The study was in two stages. The initial sample was recruited from the Register for Child Twins. This register began as a population sample of 13-year-old twins, and has been extended by recruitment of families from twins clubs to include volunteers willing to take part in research. The sample includes families from throughout the British Isles, with a large proportion being resident in the southeast of England. Parents of 795 twin pairs age 8 to 16 years were contacted with mailed questionnaires, of whom 529 responded (67%). Of the 266 pairs who did not return the questionnaires, 59 had moved away, giving a corrected response rate of 72%.

As with all volunteer twin samples, there was a bias toward higher socioeconomic status (SES) families. The British SES bands were used to classify the families: I (professional), II (managerial and technical), IIINM (skilled nonmanual), IIIM (skilled manual), IV (semiskilled manual), and V (unskilled manual) (see Jowell, Curtice, Park, Brook, Thomson, & Bryson, 1997). The SES proportions for the families were 26%, 43%, 13%, 13%, 4%, and 1%, respectively, for bands I, II, IIINM, IIIM IV, and V. Comparative data on SES was available from the 1991 survey of the National Child Development Study, a British population cohort study. In 1991, the cohort was age 33, and the SES for men with children under age 16 were 7%, 31%, 12%, 33%, 15%, and 3% (K. Smith, personal communication, January 5, 1998).

One hundred ten (110) same-sex pairs were invited into the second stage of the study on the basis of their stage-1 data (see below), and 90 took part, a response rate of 82%. These 90 pairs consisted of 52 identical twin pairs and 38 nonidentical twin pairs, 41 male and 49 female pairs, with ages ranging from 8 to 16 years (M = 12.0, SD = 2.8). The SES proportions for the families were 28%, 38% 12%, 11%, 2%, and 1%, respectively, for bands I, II, IIINM, IIIM, IV and V.

Procedure and Measures

In the first stage of the study, the children and adolescents were screened for high or low levels of depressive and anxious symptomatology using two self-report measures: the Children’s Depression Inventory (CDI; Kovacs, 1981, 1985) and the State–Trait Anxiety Inventory for Children (STAIC; Spielberger, 1973). The CDI has been shown to have good test–retest reliability and high internal consistency (Costello & Angold, 1988; Fundudis et al., 1991; Kovacs, 1981, 1985). The STAIC consists of two scales, one measuring state anxiety (i.e., current level) and the other measuring trait anxiety (i.e., usual level). The test–retest reliability and internal consistency of this scale have both been demonstrated to be high (Spielberger, 1973). All twins who took part completed an informed consent form, as did their parents. One reminder was sent out approximately 4 months after the initial letter. The zygosity of the twins was determined using a twin similarity questionnaire for use with children and adolescents (Co hen, Dibble, Grawe, & Pollin, 1973). The results of the first stage of this study are published elsewhere (Eley & Stevenson, 1999).

Two groups were selected from the stage 1 data for the second stage. The proband pairs were those in which at least one of the twins scored at least one standard deviation above the mean on either the CDI (a score of 17) or the STAIC-Trait (a score of 44). The score of 17 used to select the probands on the CDI is in line with the suggested clinical cutoff (Fundudis et al., 1991). The STAIC-State was not used to classify probands because it specifically measures transient anxious feelings. The control pairs were a random selection from those pairs in which both twins scored below the cutoff on the CDI and both scales from the STAIC. Sixty-one (61) proband pairs and 29 control pairs were visited. The mean CDI scores for the proband pairs, in which at least one twin scored above the cutoff for at least one of the measures, was 13.3 (SD = 8.10), whereas for the control pairs this was only 5.59 (SD = 4.38, t = 5.86, df = 86.44, p [less than] .001). For the STAIC-Trait, the proband mean was 40.63 (SD = 7.25), wher eas the control mean was 33.21 (SD = 5.49, t = 5.37, df = 71.15, p [less than] .001). Thus, in both measures the proband group (which also includes many lower-scoring co-twins of identified probands) scored at least one standard deviation above the controls. Furthermore, the mean CDI score for those children selected as probands on the CDI was 21.46 (SD = 4.90), more than three standard deviations greater than the mean score for the remainder of the sample of 7.4 (SD = 4.55, t = 18.26, df = 180, p [less than] .001). The mean STAIC-Trait score for those children selected as probands on that measure was 46.80 (SD = 2.74), as compared to the mean for remainder of 34.69 (SD = 5.53, t = 14.67, df = 180, p [less than] .001). Thus, both in terms of pairs in which one child is a proband and in terms of individual children scoring as probands, the scores were significantly higher than the corresponding control scores.

Although the selection of pairs for stage 2 had to be made using the CDI and STAIC-Trait scores, these measures contain a lot of similar items that artificially inflate their correlation (r .67). For this reason, once all the first stage data had been collected, in order to maximize the chances of identifying environmental influences specific to each symptom type, factor scores of purer measures of depression and anxiety were created using factor analysis. All 27 items from the CDI and all 20 items from the STAIC-Trait were entered into the factor analysis, using all 1058 individuals from the first stage of the study. An oblique rotation with a delta of zero was used, which allowed the factors to correlate. The first stage of the factor analysis produced 10 factors with an eigenvalue of 1.0 or more, and on which 2 or more variables loaded. These were self-blame, self-esteem, lonely, insecure, sad/crying, physiological arousal, worry about school, general worry, sleep problems, and decision making. The factor scores were calculated and reentered into a further factor analysis. The results of the second-order factor analysis are given in Table I. As can be seen, self-blame, self-esteem, lonely, insecure, and sad/crying all loaded by .4 or more onto one factor subsequently named depression. Physiological arousal, worry about school, general worry, sleep problems, and decision making loaded by .4 or more onto the second factor, which was therefore named anxiety. Four factors loaded at .3 or more onto both second-order factors. The sad/crying factor also loaded onto the anxiety factor, physiological arousal also loaded negatively onto the depression factor, and general worry and decision making also loaded on the depression factor. These cross-loadings reflect the clinical manifestation of these symptoms so the second-order factor scores were used rather than simple sums of each of the two pairs of five first-order factors. The depression and anxiety second-order factors correlated by .27. Further details of the cons truction of these scales are published elsewhere (Eley & Stevenson, 1999). Although the families were selected according to cutoffs on the original measures (CDI and STAIC-Trait), these purer-factor scores were used to reclassify the children for the analyses of the impact of life events and chronic experiences. The scores were not age adjusted, as neither the depression nor the anxiety dimensions had high correlations with age (.03 and — .09, respectively). For the analyses, cutoffs of one standard deviation above the mean for the entire unselected population on the depression and anxiety dimensions were used to identify probands and controls.

A semistructured interview, the Psychosocial Assessment of Childhood Experiences (Glen, Simpson, Drinnan, & Sandberg, 1993; Sandberg et al., 1993), was conducted to ascertain information relating to life events and chronic stressors during the preceding 12 months. This interview was based on the principles of the Life Events and Difficulties Schedule (LEDS; Brown & Harris, 1978b) and was constructed specifically to obtain life events about children from interviews with both the mother and the child. Different interviewers blind to the proband status of the children interviewed each of them. One of the interviewers then interviewed the mother. Events and chronic stressors were rated by whoever conducted the interview, and then checked with the other interviewer. Following discussion of all information from both the mother and child interview about each child, a comprehensive set of events and chronic stressors was produced for each child. In this way, the data collected from each child was combined with that from the mother about that child in order to provide the fullest possible representation of information. The interviews were conducted by the first author and a graduate student, both of whom undertook extensive training with Dr. Seija Sandberg, who designed the interview (Glen et al., 1993; Sandberg et al., 1993).

Rating Life Events and Chronic Experiences

Two types of influence were assessed during the interview: discrete life events, and chronic stressors lasting 4 weeks or more. Severe negative life events (those that still had a moderate or high negative impact 2 weeks after the event) were first rated for independence from the behavior of the child. A 4-point scale rated the role of the child in the event (see Paykel, 1983), and this was collapsed into two ratings: probably or definitely independent of the child, and probably or definitely behavior related. Only events that were probably or definitely independent of the child, and therefore unlikely to have been caused by the child’s symptoms, were included in this analysis. The severe independent life events were then rated on dimensions of loss and threat. A pictorial representation of the whole rating procedure is given in Fig. 1.

The loss dimension was made up of two variables: loss of an attachment figure and loss of a valued idea. Loss of a valued idea was applicable to events that resulted in a change in the status quo, including losses of an ideal, hope, or cherished notion. For example, the failure of a school entrance exam, and thus the loss of a future possibility, would be rated as loss of an idea. The threat dimension was made up of four variables: risk of loss of attachment figure, trauma as a witness, physical jeopardy, and psychological challenge. Psychological challenge referred to events in which the child had to overcome some hurdle; for example, taking an exam would be rated on that variable. The threat dimension is comparable to the notion of danger described previously (Finlay-Jones & Brown, 1981). Risk of loss of an attachment figure (threat) and loss of an attachment figure (loss) differed purely in terms of whether the loss had occurred before the interview (e.g., death of a parent), or whether at interview it se emed likely to occur in the future (e.g., terminal illness of a parent). The loss and threat ratings were continuous, but a score of 2 or more on either of these dimensions was equivalent to the event being classified as a severe negative event. For each child, the number of severe independent events that scored 2 or more on loss and the number of severe independent events that scored 2 or more on threat was calculated and used for the analysis.

The chronic stressors were categorized according to the area of life they impinged on. The stressors considered here are those that were rated as moderate or severely negative. Three categories of chronic stressors were considered: friendship problems, family relationship problems, or stressors related to schoolwork. This latter category included quite a wide range of situations, from academic difficulties such as dyslexia to the experience of taking high-school exams. Stressors were included in the analysis regardless of whether they were independent of the behavior of the child, because very few chronic stressors are probably or definitely independent of the behavior of the child, but are likely to be the result of ongoing interactions between the child and his or her environment.

Reliability of Ratings

The reliability of these ratings was tested in two stages. In the first stage, 16 families were interviewed, which resulted in data from 32 children and 16 mothers about 2 children each. Interrater reliability between the two raters was calculated for independence of both life events and chronic stressors, loss and threat ratings for life events, and negative impact ratings for the chronic stressors. For the independence ratings, kappas were calculated and these ranged from 0.72 (p [less than] .001) to 1.00 (p [less than] .001). The loss, threat, and negative impact ratings were continuous, so intraclass correlations were calculated. These ranged from .63 to .84 for the number of severe loss and threat events reported by either the mother or the child. The interrater reliability for the number of serious negative chronic experiences ranged from .78 to .83. The second stage considered the reliability of the ratings made by pooling both the child- and mother-reported data (i.e., that which was to be used in th e analyses). These ratings were agreed on during discussion between the two raters. For the purposes of the reliability study, 10 such sets of ratings were then compared with ratings made by an external expert rater. The kappas for the independence ratings ranged from 0.86 (p [less than] .001) to 1.00 (p [less than] .001). The reliability of the number of severe loss and threat events was .62 and .59, respectively. These lower figures were due to a mean difference rather than a difference in ranking of events. The raters of the present study underrated events compared to the outside rater, an effect that would, if anything, tend to lower association between the life event measures and symptomatology. Finally, the interrater reliability for the pooled information about the number of negative chronic stressors was .78.

The reliability of the CDI and STAIC-Trait was calculated in the first stage sample using alpha, and was estimated at .86 for both measures.

Statistical Analysis

The data were analyzed in two ways. First, mean scores for loss, threat, and number of friendship, family relationship, and schoolwork stressors were compared using independent sample t-tests for the probands and controls on the purer measures of anxiety and depression. There were 60 probands and 120 controls for the depression measure, and for the anxiety measure there were 69 probands and 111 controls. Due to the remaining overlap on these two scales, some individuals (28) scored above the cutoff on both measures, but unfortunately this was too small a number to be analyzed separately. As the sample consisted of twin pairs the samples were not truly independent, but the results were very similar when only one member of each pair was used. The second stage of the analysis used discordant pairs only. Probands were selected for either anxiety or depression using the purer measures, and only those pairs in which the co-twin was a nonproband for that measure were included. Paired t-tests were used to compare the probands and co-twins. There were 29 pairs discordant for depression and 37 pairs discordant for anxiety. These two groups of discordant pairs overlapped in that being classified as discordant for depression did not preclude the pair from also being classified as discordant for anxiety. There were no age differences between the anxious probands, the depressed probands, and the controls. A diagrammatic represenation of the sampling procedure and demographic details for each group is given in Fig. 2.

Differences revealed in the independent sample t-tests could be due to either shared or nonshared influences, as the analysis was simply probands versus controls regardless of the status of the probands’ co-twins. In contrast, only nonshared influences are identified in the paired t-test analysis, as these were discordant pairs. In this way, it was possible to identify which of the influences were shared and which were nonshared. It is important to note that these shared and nonshared influences could include shared and nonshared genes as the sample included DZ pairs, who share only half their genes. Although we had no method by which to disentangle genes of shared effect and environment of shared effect, we were able to distinguish between nonshared environment and genetic sources of difference by analyzing discordant MZ twin pairs alone. Any associations found in this group could only be due to environmental influences differentially influencing two members of an MZ pair. It should be noted that the number of these pairs was very small. There were 15 MZ pairs discordant on the depression measure and 18 discordant on the anxiety measure.

The effect size of the difference between the proband and control means or between the proband and co-twin group means was calculated by subtracting the mean of the control or co-twin group from the mean of the proband or proband group and dividing this score by the standard deviation of the control or co-twin group. One-tailed significance values were used because all hypotheses were unidirectional.


Loss, Threat, Depression, and Anxiety

Table II presents the independent t-tests comparing mean values of loss and threat scores in depressed versus nondepressed children and in anxious versus nonanxious children. These results offer some support for the first hypothesis. Anxious children scored significantly higher than the nonanxious children for threat but not for loss. The depressed children also scored significantly higher on loss than the nondepressed children, but not on threat. However, it can be seen that the effect size for threat in the comparison of depressed and nondepressed children is as large as those that reach significance, but due to variance differences between the groups, the degrees of freedom require adjustment and the comparison does not reach statistical significance.

The results from the paired t-tests presented in Table III reveal only one statistically significant difference between probands and their co-twins. The anxious probands score significantly higher on threat events than their nonanxious co-twins. This suggests that the factors resulting in the association between threat events and anxiety are individual specific. In contrast, those resulting in the association between loss events and depression are either shared by members of a twin pair, and thus not seen in an analysis of discordant pairs, or are of too small effect size to be revealed in the analysis of the smaller number of discordant pairs. The nonsignificant trend for association between depression and number of threat events appears to be individual specific, as it is identified to the same extent by both types of analysis.

Analysis of the MZ pairs alone produced very similar results, with only one effect reaching significance: probands within MZ pairs discordant for anxiety obtained a significantly higher mean number of threat events (1.11 vs. 0.61, effect size = 0.59, p [less than] .005).

Chronic Stressors, Depression, and Anxiety

Table IV gives the results from the independent t-tests of the mean number of friendship problems, schoolwork stresses, and family relationship problems for the probands and controls, selected by depression and then by anxiety. It is clear from this table that all three types of chronic stressors are significantly associated with depression but not with anxiety. The effect sizes are particularly large for both friendship and family relationship problems, revealing that depressed children are scoring almost one standard deviation higher on these two variables than nondepressed controls.

The paired t-tests given in Table V reveal that friendship problems are significantly greater in depressed than nondepressed twins from discordant pairs. The effect size is such that the depressed probands are experiencing almost twice as many friendship problems as their nondepressed co-twins. This suggests that the effects of these problems are nonshared in that they tend to make members of a twin pair different from one another. There is also a nonsignificant trend for the depressed probands to experience more family relationship problems, implying perhaps both some shared and some nonshared influence. However, this variable is highly correlated within twin pairs (r = .84), and as such provides a far more powerful test than that for other less correlated experiences such as schoolwork stressors (r = .38). Indeed, although the effect size for schoolwork stressors was very similar to that for family relationship problems (0.180 vs. 0.177), the mean difference between depressed probands and their co-twins fo r school stresses was not significant. This suggests that we only have power with this variable to detect the shared influence of schoolwork stresses resulting in similarity within the pair, but with a larger sample size, perhaps nonshared influences would be identified.

Repeating these analyses with the discordant MZ pairs only produced a very similar pattern of results. Probands from MZ pairs discordant for depression scores had significantly higher mean scores on both friendship problems (0.33 vs. 0.07, effect size = 1.03, p [less than] .05) and family relationship problems (0.73 vs. 0.53, effect size = 0.27, p [less than] .05).


Limitations of the Study

The current study used a two-stage strategy in order to ascertain information across the full range of scores for anxiety and depression symptoms, but as such was not focused on identifying a large number of highly depressed anxious and highly anxious individuals. For this reason, the sample size for these analyses from the second stage of the study is rather small. This meant that for the independent sample t-tests all individuals had to be used, where using just one child from each pair would have been preferable and would have lead to truly independent samples. Furthermore, the analyses of discordant pairs is much more straightforward to interpret where only identical twins have been used, as in this case all differences must be due to nonshared environment. By including DZ twins as well, we were unable to separate nonshared environmental influences from genetic differences within the pairs that would result in within-pair differences in outcome.

In addition to sample-size related issues, clinical ratings of depressive and anxiety disorders were not made on these children. Furthermore, as the cutoff used was only one standard deviation above the mean, it is unlikely that many of the children would have met diagnostic criteria.

Finally, it should be noted that the sample here, as often found with volunteer twin registers, is biased toward the upper end of the SES distribution. As such, these results may not apply to those at the very low end of the SES distribution, as they are only marginally represented in the dataset. As such, the implications of these findings are limited to a nonclinical middle-class sample, and it is clear that replication with a large representative sample of clinically diagnosed pairs of twins would greatly add to the findings of this study.

Loss, Threat, Depression, and Anxiety

In the current study, the specificity of loss and threat events to depression and anxiety was investigated. There was some evidence for the relationship between loss events and depression, strong evidence for the association between threat events and anxiety, and a suggestion of a relationship between threat and depression. These results offer some support for the hypothesis that event types can be identified that are principally associated with one type of symptomatology–that is, symptom-specific events.

By comparing results from the independent sample t-tests and the paired t-tests, it was possible to identify whether the life event types being investigated were acting as shared or nonshared influences. Here we refer to shared events as being those whose influence on outcome is shared for the twin pair, whereas nonshared influences are child specific. This is distinct from the symptom specificity discussed above, in which the influence is specific to one outcome measure rather than one individual. From these analyses the effects of threat events on anxious (and depressive) symptomatology were shown to be individual specific, making members of a pair different from one another for proband status. This was found not to be due to genetic differences between the pair, as the result was maintained when only MZ pairs were analyzed. In contrast, the effects of loss events on depression scores appeared to be shared within the pair. This result is slightly surprising in the light of the behavioral genetic literature that indicates only a small influence of shared environment on depression symptoms in children and adolescents, although there is some evidence for this being greater at the extremes (Eley, 1997a; Eley et al., 1998; Eley & Stevenson, 1999; Rende et al., 1993; Thapar & McGuffin, 1994). However, even if the influence of the shared environment is relatively small, this means only that the likelihood of identifying such influences is reduced, it does not mean that such influences are not detectable. Furthermore, as loss events in this study tended to be the loss of a grandparent or family friend or some other shared loss, it was not surprising that this variable had the effect of making depression scores similar within a pair.

Chronic Stressors, Depression, and Anxiety

As well as considering the influence of acute stressful life events on internalizing symptoms, this study investigated the role of ongoing stressors lasting 4 weeks or more. Schoolwork stresses were identified as being associated with depression when considering the whole sample, but not when the discordant pairs were analyzed. This suggests that such stressors were shared influences, making members of a twin pair similar, although the effect size in the independent samples t-test was small, and it may have been too small to be identified within the analysis of the smaller number of discordant pairs. Friendship and family relationship problems were also found to be associated with depression but not anxiety. In contrast to the result for schoolwork stresses, these findings were replicated in the analysis of the discordant pairs, and again this association was found not to be due to genetic influences as it remained present in the analysis of discordant MZ pairs alone. This suggests that these stressors are i ndividual specific in their effects, and are nongenetic. As such, they can be described as nonshared environmental influences. This is unsurprising for the friendship measure, as one would expect that to be individual specific. In contrast, the family relationship measure might be thought to be likely to be a shared environmental influence. This is an example of how something that happens within the context of the family may only have an effect on outcome for one individual, and is as such a nonshared influence. Furthermore, as a paired t-test has increased power the more highly correlated the two groups or measures are (Cohen, 1992), we had very high power here to detect nonshared aspects of family relationship influences on outcome. However, it should be stressed that the power of the paired t-test for family relationships was very high due to the high within-pair correlation for this measure. We therefore had higher power to detect any nonshared influence of this variable relative to the others. Overall, t he finding that there is an association between friendship stressors and depression is particularly interesting as it parallels findings from Finlay-Jones (1989) that depressed but not anxious subjects were significantly more likely to report lack of an intimate confidant than normal controls. These results provide considerable support for the hypothesis that there are environmental influences associated specifically with anxiety or depression.

Cause, Effect, or Shared Vulnerability?

Having identified these associations, the next stage is to attempt to understand what causes them. One possibility is that these life events and chronic stressors are specifically involved in the etiology of depressive and anxious symptoms. A second is that the presence of such symptoms results in the individual being predisposed to experience such events. A third is that there may be shared vulnerability to experience both symptoms and events. Finally, there may be a reporting bias in which depressed or anxious individuals overreport events.

A longitudinal study of 1103 adult women provides some evidence for life events being involved in the etiology of symptoms (Fergusson & Horwood, 1984). The women completed a self-report measure of depression and a reduced version of the Social Readjustment Rating Scale (Holmes & Rahe, 1967) that asked about life events in the previous 12 months. These measures were collected at two time points, 2 years apart. Structural equation modeling of the data revealed that there was a significant path leading from life events at time 1 to depressive symptoms at time 2. However, there was also a significant path from depressive symptoms at time 1 to life events at time 2, although this was of smaller magnitude. Thus the relationship between life events and depression in this study was at least in part due to the fact that life events were predictive of later depression, giving some support for the first hypothesis above. However, the path from depressive symptoms to later life events also requires explanation, and lead s to the examination of the next hypothesis.

It seems likely that in some circumstances symptoms would increase the subject’s chances of experiencing a certain event. For example, a shy child may avoid social functions and thus reduce contact with peers and increase isolation, which would here be coded as a chronic experience to do with friendships. As such, the association between the symptoms and the environment may be as a result of a person–environment correlation, possibly as a result of gene–environment correlation (g-e correlation: Plomin, DeFries, & Loehlin, 1977; Plomin et al., 1977; Scarr & McCartney, 1983). Several studies have demonstrated genetic influences on depressive symptoms in children and adolescents (Eley, 1997a; Eley & Stevenson, 1999; O’Connor, McGuire, Reiss, Hetherington, & Plomin, 1998a; O’Connor, Neiderhiser, Reiss, Hetherington, & Plomin, 1998b; Rende et al., 1993; Thapar & McGuffin, 1994), and furthermore, there is a small body of evidence suggesting that life events are heritable (Kendler, Neale, Kessler, Heath, & Eaves, 1993; Thapar & McGuffin, 1996). The association between the two may therefore be as a result of g-e correlations, in which the person experiencing the symptoms influences their environment as a result of genetic vulnerabilities. As such, a g-e correlation occurs in which genetic influences on the individual result in particular environmental situations.

As described, we did not analyze events unless they were independent of the behavior of the child, which greatly reduces the likelihood of g-e interactions, but does leave the possibility that the parents or other family members were involved in the event occurring. As such, the association between symptoms and environmental influences may be as a result of direct and indirect influences, both of which originate in the parents. This leads to the final hypothesis–that there are shared vulnerability factors resulting in both symptoms and life events or experiences.

There is some evidence that the association between measures of parenting and depression is in part due to shared genetic influences (Pike, McGuire, Hetherington, Reiss, & Plomin, 1996). As such, it was possible that the associations found in this study might have been due to genetic influences on both the depression and anxiety measures and on the environment measures. We were able to test for such by analyzing just discordant MZ pairs for whom genetic differences could not be resulting in their different scores. In this small subsample we found exactly the same pattern of results as when we considered the whole group of discordant pairs. In summary, we are therefore able to say that the nonshared associations identified in the discordant pairs analyses (threat with anxiety, family relationship, and friendships with depression) were not due to genetic influences.

Finally, depressive or anxious symptoms may result in an overreporting of events or an overreporting of the impact of events. In order to combat this possibility, we used only those events that were reported by both the mother and the child, unless there appeared to be a good reason why only one respondent would have reported them, such as events of a very personal nature. Also, to minimize the influence of the respondent’s own emotional state on ratings, we used only our own ratings in the analyses.

In conclusion, therefore, we were able to identify specific associations between symptoms, life events, and chronic stressors. As only independent life events were used in the analysis of loss and threat, these factors are likely to be genuine components in the etiology of depression and anxiety. Furthermore, the child-specific nature of the association between threat events and anxiety was not due to shared genetic influences on the two. The associations between chronic stressors and depression revealed that schoolwork stresses have a shared effect in that they influence both members of a twin pair. This may be due to the fact that this group of stresses included taking national exams over a period of several weeks, which would have been experienced by both members of a twin pair. In contrast, the family relationship and friendship problem measures were associated with depression in a child-specific way, and this was shown not to be as a result of genetic influences. In conclusion, there are clear indicatio ns for associations between environmental measures and symptoms, but further research is needed to clarify whether this association is due to stressful life events and chronic stressors resulting in increased levels of symptoms in children and adolescents.


The authors acknowledge with thanks the substantial contribution made to this study by Dr. Bettina Hohnen, who interviewed one child in every twin pair. Thanks are also due to Dr. Seija Sandberg for training the interviewers in the use of the PACE and rating of the events, and Clive Hillary for spending time discussing the interview procedure during the pilot stage of the project. Finally, we would like to express our gratitude to the families that took part in this study and made it possible.

This project was funded as a Ph.D. grant by the Child Health Research Appeal Trust, Institute of Child Heath, University of London, UK.

(1.) Behavioural Sciences Unit, Institute of Child Health, University of London.

(2.) Centre for Psychological Development, Department of Psychology, University of Southampton.

(3.) A11 correspondence concerning this article should he addressed to Dr. Thalia C. Eley, who is now at the Social, Genetic and Developmental Psychiatry Research Centre, 111 Denmark Hill, Institute of Psychiatry, Kings College, London SE5 8AF, UK. Correspondence via the internet can he sent to t.eley@iop.kcl.ac.uk.


Armsden, G. C., McCauley, E., Greenburg, M. T., Burke, P. M., & Mitchell, J. R. (1990). Parent and peer attachment in early adolescent depression. Journal of Abnormal Child Psychology, 18, 683-697.

Asarnow, J. R., Goldstein, M. J., Tompson, M., & Guthrie, D. (1993). One-year Outcomes of depressive disorders in child psychiatric in-patients: Evaluation of the prognostic power of the brief measure of expressed emotion. Journal of Child Psychology and Psychiatry, 34, 129-137.

Brown, G. W., & Harris, T. O. (1978a). Social origins of depression: A study of psychiatric disorder in women. London: Tavistock.

Brown, G. W., & Harris, T. O. (1978b). The Bedford College Life-Events and Difficulty Schedule: Directory of contextual threat ratings of events. Unpublished manuscript, Bedford College, University of London.

Carlson, G. A., & Cantwell, D. P. (1983). Case studies in prepubertal childhood depression. In D. P. Cantwell & G. A. Carlson (Eds.), Affective disorders in childhood and adolescence (pp. 39-59). Spectrum Publications Inc.

Cohen, D. J., Dibble, E., Grawe, J. M., & Pollin, W. (1973). Separating identical from fraternal twins. Archives of General Psychiatry, 29, 465-469.

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-159.

Costello, A., & Angold, A. (1988). Scales to assess child and adolescent depression: Checklists, screens, and nets. Journal of the American Academy of Child and Adolescent Psychiatry, 27, 726-737.

Eley, T. C. (1997a). Depressive symptoms in children and adolescents: Etiological links between normality and abnormality: A research note. Journal of Child Psychology and Psychiatry, 38, 861-866,

Eley, T. C. (1997b). General genes: A new theme in developmental psychopathology. Current Directions in Psychological Science, 6, 90-95.

Eley, T. C. (1999). Behavioral genetics as a tool for developmental psychology: Anxiety and depression in children and adolescents. Clinical Child and Family Psychology Review, 2, 21-36.

Eley, T. C., Deater-Deckard, K., Fombonne, E., Fulker, D. W., & Plomin, R. (1998). An adoption study of depressive symptoms in middle childhood. Journal of Child Psychology and Psychiatry, 39, 337-345.

Eley, T. C., & Stevenson, J. (1999). Using genetic analyses to clarify the distinction between depressive and anxious symptoms in children and adolescents. Journal of Abnormal Child Psychology, 27, 105-114.

Fergusson, D. M,, & Horwood, L. J. (1984). Life events and depression in women: A structural equation model. Psychological Medicine, 14, 881-889.

Finlay-Jones, R. (1989). Anxiety. In G. W. Brown & T. O. Harris (Eds.), Life events and illness (pp. 95-112). London: Unwin Hyman.

Finlay-Jones, R., & Brown, G. W. (1981). Types of stressful life events and the onset of anxiety and depressive disorders. Psychological Medicine, 11, 803-815.

Fundudis, T., Berney, T. P., Kolvin, I., Famuyiwa, O. O., Barrett, M. L., Bhate, S. R., & Tyrer, S. P. (1991). Reliability and validity of two self-rating scales in the assessment of childhood depression. British Journal of Psychiatry, 159, 36-40.

Glen, S., Simpson, A., Drinnan, D., & Sandberg, S. (1993). Testing the reliability of a new measure of life events and experiences in childhood: The Psychosocial Assessment of Childhood Experiences (PACE). European Child and Adolescent Psychiatry, 2, 98-110.

Goodyer, I. M., & Altham, P.M. E. (1991). Lifetime exit events and recent social and family adversities in anxious and depressed school-age children and adolescents–I. Journal of Affective Disorders, 21, 219-228.

Goodyer, I. M., & Cooper, P. J. (1993). A community study of depression in adolescent girls II: The clinical features of identified disorder. British Journal of Psychiatry, 163, 374-380.

Goodyer, I. M., Wright, C., & Altham, P. M. E. (1989). Recent friendships in anxious and depressed school age children. Psychological Medicine, 19, 165-174.

Holmes, T. H., & Rahe, R. H. (1967). The Social Readjustment Rating Scale. Journal of Psychosomatic Research, 11, 213-218.

Kendler, K. S., Heath, A. C., Martin, N. G., & Eaves, L. J. (1987). Symptoms of anxiety and depression: Same genes, different environments? Archives of General Psychiatry, 44, 451-457.

Kendler, K. S., & Karkowski-Shuman, L. (1997). Stressful life events and genetic liability to major depression: Genetic control of exposure to the environment. Psychological Medicine, 27, 539-547.

Kendler, K. S., Neale, M. C., Kessler, R. C., Heath, A. C., & Eaves, L. J. (1992). Major depression and generalized anxiety disorder. Same genes (partly) different environments? Archives of General Psychiatry, 49, 716-722.

Kendler, K. S., Neale, M. C., Kessler, R. C., Heath, A. C., & Eaves, L. J. (1993). A twin study of recent life events and difficulties. Archives of General Psychiatry, 50, 789-796.

Kovacs, M. (1981). Rating scales to assess depression in school age children. Acta Paedopsychiatrica, 46, 305-315.

Kovacs, M. (1985). The Children’s Depression Inventory (CDI). Psychopharmacology Bulletin, 21, 995-1124.

Kranzler, E. M., Shaffer, D., Wasserman, G., & Davies, M. (1989). Early childhood bereavement. Journal of the American Academy of Child and Adolescent Psychiatry, 29, 513-520.

Loss, N., Beck, S. J., & Wallace, A. (1995). Distressed and nondistressed third- and sixth-grade children’s self-reports of life events and impact and concordance with mothers. Journal of Abnormal Child Psychology, 23, 397-409.

O’Connor, T. G., McGuire, S., Reiss, D., Hetherington, E. M., & Plomin, R. (1998a). Co-occurrence of depressive symptoms and antisocial behavior in adolescence: A common genetic liability. Journal of Abnormal Psychology, 107, 27-37.

O’Connor, T. G., Neiderhiser, J. M., Reiss, D., Hetherington, E. M., & Plomin, R. (1998b). Genetic contributions to continuity, change, and co-occurrence of antisocial and depressive symptoms in adolescence. Journal of Child Psychology and Psychiatry, 39, 323-336.

Parker, J. G., & Asher, S. R. (1993). Friendship and friendship quality in middle childhood: Links with peer group acceptance and feelings of loneliness and social dissatisfaction. Developmental Psychology, 29, 611-621.

Paykel, E. S. (1983). Methodological aspects of life events research. Journal of Psychosomatic Research, 27, 341-352.

Paykel, E. S., Myers, J. K., Dienelt, M. N., Klerman, G. L., Lindenthal, J. J., & Pepper, M. P. (1969). Life events and depression. Archives of General Psychiatry, 21, 753-760.

Peterson, A. C., Compas, B. E., Brooks-Gunn, J., Stemmler, M., & Grant, K. E. (1993). Depression in adolescence. American Psychologist, 48, 155-168.

Pike, A., McGuire, S., Hetherington, E. M., Reiss, D., & Plomin, R. (1996). Family environment and adolescent depressive symptoms and antisocial behavior: A multivariate genetic analysis. Developmental Psychology, 32, 590-603.

Pike, A., & Plomin, R. (1996). Importance of nonshared environmental factors for childhood and adolescent psychopathology. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 560-570.

Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from each other? Behavioral and Brain Sciences, 10, 1-16.

Plomin, R., DeFries, J. C., & Loeblin, J. C. (1977). Genotype–environment interaction and correlation in the analysis of human behaviour. Psychological Bulletin, 85, 309-322.

Puig-Antich, J., Kaufman, J., Ryan, N. D., Williamson, D. E., Dahl, R. E., Lukens, E., Todak, G., Ambrosini, P., Rabinovich, H., & Nelson, B. (1993). The psychosocial functioning and family environment of depressed adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 32, 244-253.

Reinherz, H. Z., Stewart-Berghauer, G., Pakiz, B., Frost, A. K., Moeykens, B. A., & Holmes, W. M. (1989). The relationship of early risk and current mediators to depressive symptomatology in adolescence. Journal of the American Academy of Child and Adolescent Psychiatry, 28, 942-947.

Rende, R. D., Plomin, R., Reiss, D., & Hetherington, E. M. (1993). Genetic and environmental influences on depressive symptomatology in adolescence: Individual differences and extreme scores. Journal of Child Psychology and Psychiatry, 34, 1387-1398.

Rutter, M., & Brown, G. W. (1966). The reliability and validity of measures of family life and relationships in families containing a psychiatric patient. Social Psychiatry, 1, 38-53.

Sandberg, S., Rutter, M., Giles, S., Owen, A., Champion, L., Nicholls, J., Prior, V., McGuinnesss, D., & Drinnan, D. (1993). Assessment of psychosocial experiences in childhood: Methodological issues and some illustrative findings. Journal of Child Psychology and Psychiatry, 34, 879-897.

Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of genotype-[greater than]environmental effects. Child Development, 54, 424-435.

Schwartz, C. E., Doter, D. J., Beardslee, W. R., Lavori, P. W., & Keller, M. B. (1990). Maternal expressed emotion and parental affective disorder: Risk for childhood depressive disorder, substance abuse, or conduct disorder. Journal of Psychiatric Research, 24, 231-250.

Spielberger, C. (1973). Preliminary test manual for the State-Trait Anxiety inventory for Children. Palo Alto: Consulting Psychologists Press, Inc.

Stevenson, J., Batten, N., & Cherner, M. (1992). Fears and fearfulness in children and adolescents: A genetic analysis of twin data. Journal of Child Psychology and Psychiatry, 33, 977-985.

Tejerina-Allen, M., Wagner, B. M., & Cohen, P. (1994). A comparison of across-family and within-family parenting predictors of adolescent psychopathology and suicidal ideation. In E. M. Hetherington, D. Reiss, & R. Plomin (Eds.), Separate social worlds of siblings: The impact of non-shared environment on development (pp. 143-157). New Jersey: Lawrence Erlbaum Assoc. Inc.

Thapar, A., & McGuffin, P. (1994). A twin study of depressive symptoms in childhood. British Journal of Psychiatry 165, 259-265.

Thapar, A., & McGuffin, P. (1996). Genetic influences on life events in childhood. Psychological Medicine, 26, 813-820.

Thapar, A., & McGuffin, P. (1997). Anxiety and depressive symptoms in childhood-a genetic study of comorbidity. Journal of Child Psychology and Psychiatry, 38, 651-656.

Tisher, M., Tonge, B. J., & Horne, D. J. (1994). Childhood depression, stressors and parental depression. Australian and New Zealand Journal of Psychiatry, 28, 635-641.

Topolski, T. D., Hewitt, J. K., Eaves, L. J., Silberg, J. L., Meyer, J. M., Rutter, M., Pickles, A., & Simonoff, E. (1997). Genetic and environmental influences on child reports of manifest anxiety and symptoms of separation anxiety and overanxious disorders: A community-based twin study. Behavior Genetics, 27, 15-28.

Weller, R. A., Weller, E. B., Fristad, M. A., & Bowes, J. M. (1991), Depression in recently bereaved prepubertal children. American Journal of Psychiatry, 148, 1536-1540.

Williams, S., & McGee, R. (1991). Adolescent’s self-perceptions of their strengths. Journal of Youth and Adolescence, 20, 325-337.

COPYRIGHT 2000 Plenum Publishing Corporation

COPYRIGHT 2000 Gale Group