Dissociation is strongly associated with trauma, but explanatory accounts that build trauma into the construct itself risk confusing cause with mechanism. Fragmentation is clinically meaningful in some severe presentations, yet the metaphor is often read as a static division into separate parts, whereas many dissociative experiences fluctuate across time and context. We propose the Dynamic Self-Instability (DSI) model. Its central claim is that dissociative phenomena arise when four normally coordinated aspects of self-experience, integration, accessibility, self-attribution and temporal continuity, become unreliably coordinated (weakly or unstably coupled). For each domain we give a narrower working definition, candidate indicators that do not reproduce dissociative symptom items, and an example of selective change, and we outline designs in which coordination is measured before dissociative episodes rather than inferred from them. Trauma remains the best-established and often the most important cause, but other perturbations, including sleep and arousal dysregulation, interoceptive disturbance, substances, neurological insult and psychotic self-disturbance, may reach the same immediate mechanism. Structural fragmentation is treated as one relatively stable pattern within a broader dynamic system , and we specify tests that would discriminate DSI from structural, temporal-depth and transtheoretical accounts . The dissociation–psychosis interface provides a demanding test case: dissociation-first and psychosis-first sequences can produce the same cross-sectional overlap, and we specify what would be measured, how often and over what interval to distinguish them. These are hypothesized temporal patterns, not diagnoses, and temporal precedence alone would not establish causation. We separate findings that would refute the core model from findings that would require abandoning only one extension or would indicate limited practical value.