No disease-modifying treatments for dementia are available. Sleep disturbance is strongly associated with cognitive impairment. Non-pharmacological treatments targeting sleep may offer an alternative therapeutic approach. We searched PubMed, CINAHL, EMBASE and the Cochrane library for non-pharmacological treatments for sleep disturbance in mild cognitive impairment (MCI) and dementia, published in English from October 1965 to 2018, including all designs, excluding studies of drug therapies. In all, 53 papers representing 48 studies were included. Participant age ranged from 67.3 to 89.4 years. Most studies (79%) had small samples (<50 participants, range 1-173) and were conducted in long-term/residential care (62%). The majority (85%) recruited participants with moderate-severe dementia; mean MMSE scores ranged from 0 to 28.3/30. Four studies examined MCI. Light therapy delivered over 1-10 weeks was the most studied stand-alone intervention (n¿=¿27), and the majority (81.5%) of these studies found improvements on objective or subjective sleep measures, though the evidence was inconclusive with significant clinical and methodological heterogeneity. Seven multi-modal intervention studies were identified, all incorporating light exposure, and six of these reported improved sleep. Other interventions included electrotherapy stimulation (n¿=¿4), physical exercises/activities (n¿=¿4), acupressure/acupuncture (n¿=¿3) and mindfulness/cognitive behavioural therapy (n¿=¿3). Those examining MCI utilised different mono-modal approaches. A meta-analysis of data from randomised controlled trials showed a statistically significant (mean difference = 3.44, 95% CI: 0.89-5.99, I2=0%; p¿=¿0.008) improvement in sleep efficiency between interventions and controls, favouring the pooled interventions (bright light, multi-domain and other therapies). No other significant differences in sleep or non-sleep outcomes were found. While evidence is available for non-pharmacological sleep interventions, particularly multi-domain approaches, studies were diverse and had small samples. More research examining multi-modal interventions, community-dwellers and those with MCI is required.