The Statistical Downscaling Model (SDSM) is used to investigate future projections of daily minimum and maximum temperature extremes for 45 stations and rainfall extremes for 39 stations across the Caribbean and neighbouring regions. Models show good skill in reproducing the monthly climatology of the mean daily temperatures and the frequencies of warm days, warm nights, cool days and cool nights between 1961 and 2001. Models for rainfall exhibit lower skill but generally capture the monthly climatology of mean daily rainfall and the spatial distribution of the mean annual maximum number of consecutive dry days (CDD) and mean annual count of days with daily rainfall above 10 mm (R10). Future projections suggest an increase (decrease) in warm (cool) days and nights by 2071–2099 under the A2 and B2 scenarios relative to 1961–1990. An increase in CDD is suggested for most stations except some eastern Caribbean stations and Bahamas. Decreases in RX1 (monthly maximum 1-day precipitation), R10 and R95p (annual total rainfall above the 95th percentile) are also suggested for some northern Caribbean locations and Belize under the A2 scenario, compared to a mixture of increases and decreases for the eastern Caribbean. Atmospheric predictors used in SDSM correlate well with known oceanic and atmospheric drivers of Caribbean climate, e.g. the Atlantic Multidecadal Oscillation (AMO) on a seasonal timescale. Atlantic sea surface temperatures and the Caribbean low level jet appear to have significant influence on Caribbean temperature and rainfall extremes.