Assessment of attachment in psychosis: A psychometric cause for concern

Attachment has recently been proposed as a key developmental construct in psychosis, in particular with respect to interpersonal functioning and social cognition. The current study examined the latent structure of the self-report Psychosis Attachment Measure (PAM) and its relationship to lower-level perceptual and higher-order inferential social cognitive processes. The PAM was administered to 138 psychiatrically stable outpatients with schizophrenia alongside a battery of symptom, social cognitive, and functional measures. PAM responses were analyzed using latent variable measurement models, which did not yield evidence of the coherent two-dimensional structure predicted by previous literature. A unidimensional subscale comprising 6 of the 16 original PAM items possessed the strongest psychometric properties. This subscale was generally uncorrelated with social cognitive measures and showed weak correlations with some symptoms measures and with community functioning. These results suggest that either the PAM may not measure attachment in psychosis or it may measure only attachment anxiety but demonstrate little construct validity in this population. These results tell a cautionary tale regarding making theoretical inferences on the basis of measures without coherent latent structure. Attachment measures with stronger psychometric properties will help clarify putative relationships between attachment and social cognitive processes in psychosis.