Attachment anxiety has been consistently linked with increased vulnerability to depression, and hyperactivating emotion regulation strategies (e.g., rumination) have been shown to mediate this relationship. Investigations of mediators of the attachment avoidance to depression relationship have yielded inconsistent findings, and the nature of this relationship remains to be clarified. There is evidence to suggest that the constructs of thought suppression and self-compassion are associated with attachment avoidance and also with depressive symptomology. In order to further clarify the nature of this relationship, the present study tested a serial mediation model, whereby it was hypothesised that thought suppression and self-compassion were serial mediators of the relationship between attachment avoidance and depression. One hundred and forty-eight participants completed an online composite questionnaire consisting of the Experiences in Close Relationships-Revised Questionnaire, the White Bear Suppression Inventory, the Self-Compassion Scale, and the Depression Anxiety and Stress Scale-21. Initial results supported the hypothesised serial mediation model (Model A); that is, higher attachment avoidance predicted higher thought suppression, higher thought suppression predicted lower levels of self-compassion and lower levels of self-compassion predicted higher depression. However, this model was no longer significant following the inclusion of attachment anxiety as a covariate within the post-hoc analysis. A second, post-hoc serial mediation model was tested (Model B), with the only difference being that attachment anxiety replaced attachment avoidance as the independent variable. This model was significant, with and without the inclusion of attachment avoidance as a covariate. The study provides evidence for the central role of thought suppression and self-compassion as mechanisms underlying the relationship between insecure attachment and depression, and indicates that these factors operate in opposing directions. The findings are discussed in terms of explicating some of the processes through which insecure attachment confers vulnerability to depression. The implications of the observed degree of shared variance between the two attachment dimensions suggests these constructs may be more appropriately considered overlapping, rather than orthogonal.