Submarine landslides are one of the most important processes for moving sediment across our planet. Landslides that are fast enough to disintegrate can generate potentially hazardous tsunamis, and produce long run-out turbidity currents that break strategically important cable networks. It is therefore important to understand their frequency and triggers. This thesis aims to do so using extensive datasets (N>100) suitable for statistical analysis. The influence of temporally non-random variables on landslide and turbidity current frequency is assessed statistically. In light of predicted future global and sea level rises this is a timely study. Analysis of large volume turbidites (>0.1 km3) reveals two distinct frequency distribution forms for submarine landslide recurrence. A Poisson (time-independent) form is observed in three basins which may indicate similar controls on landslide frequency and triggers occur in disparate areas. A log-normal (time-dependent) distribution is seen in the Iberia Abyssal Plain over much longer timescales (20 Myr). Physiographic and palaeoclimatic effects are thought to explain the two different distribution forms. The influence of sea level is either shown to be statistically insignificant (Poisson form) or has a significantly delayed (~1.2 Myr) influence (log-normal form) on landslide and turbidite recurrence. Two sequences that cross the Initial Eocene Thermal Maximum (IETM) hyperthermal, show a reduction in turbidity current and landslide activity, rather than the increase hypothesised by many studies. Therefore, predicted future sea level rise and global warming may not necessarily result in significantly increased submarine landslide or turbidity current frequency on human timescales. Finally, a unique direct monitoring dataset from the Squamish Prodelta, British Columbia provides new insights into the links between rivers and offshore deltas on very short (