Self-Discrepancies Among Individuals with a History of Depression: The Role of Feared Self-Guides

Miller and Chapman (2001) raise concerns about use of ANCOVA to control group differences when groups were not based on random assignment. In particular, Miller and Chapman make the case that it is inappropriate to covary out conceptually substantive variance from two groups who by definition differ on the covariate. To do so, leaves variance in the dependent variable that reflects a phenomenon that either does not exist in nature, is no longer conceptually meaningful or that is conceptually distinct from the phenomenon of interest. The issue then is whether or not current depressive symptoms reflect “essential variance” in the history of depression variable that cannot be removed without substantively altering the construct. In other words, are elevated current depressive symptoms a fundamental component of having once met lifetime criteria for major depression? Is the history of depression variable no longer substantively meaningful after removing variance related to current symptoms?

If current symptoms are an essential feature of past depression, it would be inappropriate to treat them as a covariate, just as it would be inappropriate to covary out variance related to current symptoms from an analysis comparing individuals who meet criteria for current MDD versus those who do not. However, given that depression typically involves a fluctuating course that includes well periods with minimal or no symptoms, we believe that our use of an ANCOVA to handle group differences in current depressive symptoms is appropriate. Indeed this is the same logic used in prospective studies investigating predictors of change in depression over time. Such studies typically treat baseline depressive symptoms as a covariate because these are strongly predictive of future symptoms (the dependent variable) and are usually correlated with the variables that are being tested as predictors of change in depression.

Vergara and Roberts (2011) examined temperament and goal setting among individuals with and without a past history of depression in this same sample. Results indicated that previously depressed individuals reported greater numbers of avoidance goals and higher scores on the behavioral activation system (BAS). In order to provide a stringent test of the current hypotheses, we included number of avoidance goals and BAS as covariates. Results were similar to those reported above. In particular, previously depressed individuals reported higher scores on actual-feared self-guide congruence, F (1,79) = 14.06, P < .01, η ² = 15, and higher expectations that feared self-guides would be attained in the future, F (1,79) = 8.90, P < .01, η ² = .10, compared to never depressed individuals.

When feared self-guides that met DSM-IV MDD symptom criteria were not removed from analyses, previously depressed individuals reported higher scores on actual-feared self-guide congruence, F (1,81) = 10.53, P < .01, η ² = 12, and higher expectations that feared self-guides would be attained in the future, F (1,81) = 8.63, P < .01, η ² = 10, compared to never depressed individuals. Similar null results to those reported in the text were found when examining group differences in importance of feared self-guides, F (1,80) = .03, P = .85, η ² = 0.