Paradoxical Intention (PI) is a cognitive treatment approach for sleep-onset insomnia. It is thought to operate by eliminating voluntary sleep effort, thereby ameliorating sleep performance anxiety, an aroused state incompatible with sleep. However, this remains untested. Moreover, few PI studies have employed objective sleep measures. The present study therefore examined the effect of PI on sleep effort, sleep anxiety and both objective and subjective sleep. Following a seven-night baseline, 34 sleep-onset insomniacs were randomly allocated to 14 nights of PI, or to a control (no PI) condition. Consistent with the performance anxiety model, participants allocated to PI, relative to controls, showed a significant reduction in sleep effort, and sleep performance anxiety. Sleep-onset latency (SOL) differences between PI participants and controls using an objective sleep measure were not observed, although an underlying trend for significantly lowered subjective SOL amongst PI participants was demonstrated. This may relate to actigraphic insensitivity, or more probably confirms recent suggestions that insomniacs readily overestimate sleep deficit, due to excessive anxiety about sleep. Together, results help determine putative mechanisms underlying PI, have important implications for the clinical application of PI, and emphasize the need for further PI research within an experimental cognitive framework.