Examining the Shared and Unique Features of Self-Concept Content and Structure in Borderline Personality Disorder and Depression

While we feel that dimensional designs are particularly well suited to answering questions of specificity, we are not intending to make any strong claims here as whether dimensional designs are superior to diagnostic ones in general. This issues remains contentious in the field. In our view, categorical and dimensional models and design can happily co-exist alongside one another (see Arntz 1999). Each approach brings with it particular theoretical and clinical pros and cons (e.g. see Paris et al. 2009) and researchers and clinicians should choose the framework that most suits their particular purpose.

List of words used in the PDST available from the corresponding author.

To examine convergent validity of the continuous BPD measure, we repeated this analysis with BPD diagnostic status as a between-groups factor rather than PAI-BOR score as a continuous covariate. An identical pattern of results emerged. This was also true for the interconnectedness and clustering analyses. See Online Resource 1.

To validate our clustering measure as a sensitive and unique measure of self-concept disturbance in BPD, we conducted some additional exploratory analyses to see if clustering was most clearly related to the identity disturbance factor of the PAI-BOR. As expected, greater identity disturbance was associated with greater clustering, r = .42, p < .001, and this held when controlling for depression severity, rp = .37, p = .001. However, in both zero-order and partial correlation analyses (controlling for depression) the other factors of the PAI-BOR were also related to clustering: affect instability, r = .34, p < .01, rp = .28, p = .02; negative relationships, r = .39, p = .001, rp = .34, p < .01; and self-harm, r = .26, p = .02, rp = .19, p = .11. Next, we examined if these associations for each PAI-BOR factor held when also controlling for the other PAI-BOR factors (and depression severity). There remained a trend significant relationship for identity disturbance, rp = .22, p = .07, but the affect instability, rp = .6, p = .64, negative relationships, rp = .15, p = .21, and self-harm, rp = .01, p = .96, were no longer significant. These pattern of findings show that the identity disturbance factor of the PAI-BOR is most clearly related to clustering on the PDST as expected, validating this index as a useful additional outcome measure for future PDST studies. Effectively, these analyses move beyond a latent variable approach (where a single “BPD” dimension is driving results) to an overlapping network systems approach (where particular symptom clusters within the BPD construct are driving results; see Cramer et al. 2010).