Within a symptom framework (e. g. Bentall, 1990), the present research investigated the self-structure of individuals who were actively experiencing paranoid and persecutory delusions. The structural qualities of self-schemata were determined in two experimental groups (15 acutely unwell paranoid patients and 29 non-clinical controls) using a multiple free-sorting task. Participants were asked to endorse pre-selected self-attributes and position their chosen elements into social roles or identities. Linville's unitary index was calculated using Attneave's H algorithm. Separate positive and negative self-complexity indices were also computed following Woolfolk's model, as were levels of differentiation and integration (Rafaeli-Mor et al., 1999). Paranoid individuals displayed reduced unitary and positive self-complexity. Their responses on the negative index mirrored those of controls. The clinical group exhibited less differentiation within endorsed attributes, no overlap of self-descriptive elements and reduced identified social roles. Psychological well-being was inversely related to negative self-complexity and directly associated with positive self-structure. Against predictions, greater self-complexity did not buffer the effects of stress life-events on psychological functioning, with unitary complexity exacerbating levels of anxiety and low self-esteem. A strong inverse trend was observed between negative self-structure and length of present admission. Schizotypy was consistently associated with increased negative self-complexity, indicating a possible vulnerability marker for high-risk populations. Clinical implications include the use of interpersonal therapeutic processes to develop self-reflection skills, the importance of early interventions to prevent the fragmentation and simplification of the self and its internal working models of being, and the potential predictive use of structure to indicate schizotypy. Limitations of the research include a small sample size and under power, lack of appropriate control groups, poor definitional criteria of self-structure and the omission of multi-dimensional phenomenological measures of delusional pathology.