Results
Preliminary analyses
Compared to participants from intact families, those from divorced families reported lower family income (t = −3.56, p < .01), but did not differ on any of the other demographic variables (see Table 1). Further, none of the demographic variables interacted with divorce status to predict cortisol values. Divorced and intact families significantly differed in reported amounts of family conflict, but did not differ in current depression and anxiety.
Table 1.
Sample characteristics.
Total (n = 94) Intact (n = 51) Divorce (n = 43) Comparison (p)
Gender (N, % in each group) p = .99
Male 37 (39.4%) 20 (39.2%) 17 (39.5%)
Female 57 (60.6%) 31 (60.8%) 26 (60.5%)
Age (M, SD) 19.9 (2.2) 19.9 (2.3) 19.9 (2.1) p = .93
BMI (M, SD) 23.4 (3.9) 23.5 (3.7) 23.2 (4.1) p = .88
Hormonal contraceptives p = .18
Yes (N, % of women in each group) 31 (54.4%) 15 (48.4%) 16 (61.5%)
No (N, % of women in each group) 11 (19.3%) 8 (25.8%) 3 (11.5%)
Ethnicity (N, % in each group) p = .49
White, non-Hispanic 73 (77.7%) 39 (76.5%) 34 (79.1%)
African-American 2 (2.1%) 1 (2.0%) 1 (2.3%)
Hispanic 13 (13.8%) 8 (15.7%) 5 (11.6%)
Other 6 (6.4%) 3 (5.9%) 3 (7.0%)
Family conflict (range 0–9) 3.8 (2.5) 2.9 (2.1) 5.0 (2.5) p < .001
Depressive symptoms (range 0–37) 9.5 (7.8) 9.0 (7.2) 10.3 (8.3) p = .39
State anxiety (range 20–59) 37.0 (9.5) 37.1 (10.2) 37.3 (8.9) p = .94
Time of participation (range 12:55–16:00) 14:07 (1:03) 14:13 (1:04) 14:00 (1:01) p = .29
Family income (N, % in each group) p = .001
0–US$ 29,999 5 (5.3%) 0 (0%) 5 (11.6%)
US$ 30,000–US$ 44,999 11 (11.7%) 3 (5.9%) 8 (18.6%)
US$ 45,000–US$ 59,999 10 (19.6%) 4 (7.8%) 6 (14.0%)
US$ 60,000–US$ 79,999 10 (19.6%) 7 (13.7%) 3 (7.0%)
US$ 80,000–$99,999 18 (19.1%) 10 (19.6%) 8 (18.6%)
$100,000 and above 37 (39.3%) 26 (50.1%) 11 (25.6%)
Table options
Primary analyses
Parental divorce and cortisol
The study's primary hypothesis was supported in that participants from the divorce group showed lower levels of cortisol across the task than those from intact families, β = −.044, F(1,94) = 4.8, p = .031, Cohen's d = .45 (see Fig. 1). The groups did not significantly differ in the magnitude of reactivity to the task as measured by a univariate score of the difference between time 1 and time 3, p = .29, or as measured by a univariate score of the difference between time 2 and time 3, p = .31. Univariate analyses predicting cortisol at each sampling time point separately found that the groups significantly differed in cortisol pretask (p = .01), and immediately after the task (p = .01), and showed a trend towards significance at 15 min (p = .06), 30 min (p = .07), and 45 min (p = .10) after the task. The family group main effect remained significant after multilevel models that additionally controlled for gender, age, BMI, use of hormonal contraceptives, average caffeine intake, smoking status, and time of last meal, β = −.039; F(1,87) = 4.3; p = .042; Cohen's d = .44. Within the divorce group, multilevel models found no significant effect of age at the time of divorce on cortisol levels (p = .90) or reactivity (p = .21).
Cortisol levels before and after a challenging speech task in young adults from ...
Figure 1.
Cortisol levels before and after a challenging speech task in young adults from divorced and intact married families (M, SE).
Figure options
Current distress
Current symptoms of depression and anxiety were not associated with cortisol levels as analyzed with multilevel modeling (depression p = .19; anxiety p = .39), reactivity to the task as measured by a change score from time 1 to time 3 (depression p = .96; anxiety p = .79), or reactivity to the task as measured by a change score from time 2 to time 3 (depression p = .20; anxiety p = .25). The main effect of parental divorce in multilevel analyses remained significant after controlling for depression and anxiety, β = −.042, F(1,92) = 4.1, p = .045, Cohen's d = .42.
Family conflict
Reports of family conflict did not predict cortisol levels as analyzed with multilevel modeling (p = .38), reactivity measured as a change score from time 1 to time 3 (p = .82), or reactivity measured as a change score from time 2 to time 3 (p = .78). The main effect of childhood parental divorce remained significant after controlling for family conflict, β = −.045, F(1,93) = 4.0, p = .050, Cohen's d = .41.
Family income
Lower family income was significantly associated with lower cortisol across the task as analyzed using multilevel modeling, β = .033, F(1,91) = 7.8, p = .006. When income was included in the same model, childhood parental divorce was no longer a significant predictor of cortisol, β = −.035, F(1,90) = 2.6, p = .11, Cohen's d = .34. However, family income remained significant, β = .027, F(1,90) = 4.4, p = .038. Mediation of the effects of parental divorce by family income was evaluated. Path a, relating parental divorce to family income was significant, a = 1.22, sa = .334, p < .001. Path b, relating income to cortisol adjusted for parental divorce was also significant, b = .0266, sb = .0126, p = .038. The mediated effect was statistically significant at p < .05, (ab) = .032, 95% CI, .003–.072. Family income did not predict reactivity measured as a change score from time 1 to time 3 (p = .48), or reactivity measured as a change score from time 2 to time 3 (p = .45)
