Effects of Childhood Abuse on Overgeneral Autobiographical Memory in Current Major Depressive Disorder

We also conducted some alternative analyses to ensure the robustness of this regression. We choose to enter childhood physical and sexual abuse on Step 1 of the regression to show test their effects uncomplicated by the other covariates, but it should be stressed that the final regression model presented in Table 2 shows the unique effects of childhood physical and sexual abuse above and beyond the other covariates. We verified this by rerunning the regression with childhood sexual and physical abuse on the last step. Doing the regression in this way leaves the results absolutely unchanged; the ΔR ² for childhood sexual and physical abuse entered on the last step was .08 in this model, indicating that these two dichotomous variables accounted for 8 % of the variance beyond the other five covariates, p = .046. For the final model, the values for B, SE (B), and β statistics were all exactly as reported in Table 2, as would be expected.

We also ran the regression shown in Table 2 also including age at depression onset as an additional covariate in Step 3 of the regression along with the number of depressive episodes, as well as gender as an additional covariate in the final step of the model. We did not present this as the primary regression because the ratio of cases to parameters would drop to less than 10, and two cases would need to be excluded due to missing data on age of onset (N = 75). The results for childhood physical abuse were nevertheless unchanged in this model—It was still a significant predictor of fewer specific memories in the final model including all of the other covariates. Thus, the final interpretation of the data would remain absolutely unchanged.

Finally, we examined bivariate associations between the number of specific memories abuse variables. The correlation was only significant for childhood physical abuse, r = −.27, p = .019; for childhood sexual abuse, the correlation was r = .06, p = .62, ns.

Our trauma interview contained some individual items that could speak to these questions in a future, larger sample. Unfortunately our sample sizes for participants with a history of abuse were too small to conduct analyses within only these individuals. Future studies along these lines would need to be large enough for sufficient variability of abuse characteristics to determine which characteristics are reliably related to overgeneral memory.