Objective. There is a suggestion in the literature that more variable affect increases suicidal ideation through the repeated re-activation of latent suicidal cognitions. The hypothesis that affective variability would be a better predictor of suicidal ideation and related behaviour than affect level was tested in individuals at ultra-high risk of developing psychosis. This study also examined the prediction that affective variability is a suicide-specific mechanism and would not predict levels of attenuated psychotic phenomena. Method. Twenty-seven ultra-high risk individuals were required to complete ambulant ratings of their affect when prompted by an electronic wristwatch for six days (the experience sampling method). In the debriefing session, participants were assessed with a semi-structured interview (the Comprehensive Assessment of At-Risk Mental State), which assessed the severity and frequency of suicidality and psychosis-related phenomena. Results. The variability of negative and positive affect was predictive of the frequency of suicidal thoughts and behaviour. More variable negative, but not positive affect, was also associated with more severe suicidal ideation and related behaviour. Affect variability was not significantly related to the severity of attenuated psychotic phenomena. Conclusion. Affective variability appears to be a specific risk factor for suicidal ideation in individuals at ultra-high risk of developing psychosis. Early intervention should focus on providing individuals with skills for regulating their own affect. © 2011 The British Psychological Society.