This paper assesses the relationship between a regional index of rainfall (SWER) over Devon and Cornwall, South-west England, and concurrent gridded (5° × 5°) sea-surface temperature anomalies (SSTAs) for the North Atlantic–European domain (10–70°N, 80°W–20°E) over the period 1950–97. Monthly and seasonal SSTA : SWER correlation fields are derived, and stepwise regression models are then constructed to specify SWER through SSTA variations. In particular, this paper emphasizes the importance of assessing all correlation matrices for field significance, an analysis stage that is often ignored by many researchers. The most temporally reproducible signal is found between depressed SSTs to the west of the British Isles and above-average rainfall over Devon and Cornwall, although this eastern Atlantic signal is by no means constant over time in terms of its location and spatial extent. Another distinctive feature of the anomaly correlation maps is for locally significant positive SSTA : SWER associations to emerge at high latitudes (60–70°N) in summer and early autumn, e.g. Davis Strait, Norwegian Sea. Field significance testing reveals that seasonal SSTA : SWER correlation fields are statistically more robust than their monthly counterparts. This is demonstrated by the fact that the seasonal fields (excluding summer) all achieve field significance at the 0.1 level, whereas only five of the 12 monthly fields (January, May, July, October and December) do so, suggesting that some of the locally significant SSTA : SWER correlations presented in this paper could have been obtained by merely correlating SWER with a random number series. Copyright © 2002 Royal Meteorological Society.