It has been proposed that insecure attachment can have adverse effects on the course of psychosis once symptoms have emerged. There is longitudinal evidence that increased insecure attachment is associated with increased severity of psychotic symptoms. The present study examined whether in the flow of daily life attachment insecurity fluctuates, whether elevated stress precedes the occurrence of attachment insecurity, and whether elevated attachment insecurity precedes the occurrence of paranoia. Twenty clinical participants with a psychosis-spectrum diagnosis and twenty controls were studied over six consecutive days using the experience sampling method (ESM). The findings revealed that fluctuations in attachment insecurity were significantly higher in the clinical group, that elevated stress predicted a subsequent increase in attachment insecurity, and that elevated attachment insecurity predicted a subsequent increase in paranoia; this effect was not observed in auditory hallucinations once co-occurring symptoms were controlled for. Finally, although previous ESM studies have shown that low self-esteem precedes the occurrence of paranoia, attachment insecurity continued to predict paranoia even when self-esteem was controlled for. The findings suggest that attachment security may be associated with a lower risk of paranoia, and that psychological interventions should address attachment beliefs and work towards establishing a sense of attachment security.