For over 30 y sleep restriction therapy (SRT) has been used to treat insomnia but we know very little about how this therapy exerts its effects. When SRT was first described, it was hypothesised to treat insomnia by addressing four key factors: strengthening homeostatic sleep pressure, inhibiting perpetuating practices (excessive time in bed), attenuating hyperarousal and tightening regulatory control of sleep by the endogenous circadian pacemaker. We conducted a systematic literature review in search of evidence for these putative mechanisms-of-action. A total of 15 randomised and non-randomised studies investigating SRT met inclusion criteria. For each study, we extracted all variables associated with the proposed mechanisms and assessed study quality using a structured appraisal tool. The extracted variables were: time in bed (TIB), napping, variability in sleep, markers of circadian rhythmicity, measurements of sleep pressure/sleepiness, and assessments of arousal. Overall study quality was poor as indicated by a mean quality score of 17 (out of a possible range of 0–31). No study indicated, or indeed was designed to test, whether changes in the proposed mechanisms act as mediators of treatment outcomes. Of all reviewed studies, most reported a reduction in TIB (10/10) and/or revealed a decrease in sleep onset latency (10/14), indexing increased sleep pressure. However, such changes were most often reported at the end of treatment, reflecting an outcome and not a mechanism of SRT per se. Evidence for reduction in arousal (4/4) and night-to-night sleep variability (2/2) was found in only a small number of uncontrolled studies while there was no evidence for change in circadian phase or periodicity (0/1). Our review suggests that SRT targets some of the hypothesised processes but specifically-designed mechanistic evaluations are needed. We introduce a new testable model of SRT mechanism-of-action (Triple-R) and set out a research agenda aimed at stimulating prospective investigations.