Abstract A multiple regression analysis was performed upon selected environmental variables for a series of islands in the British Isles, to establish their effects upon the size of the butterfly fauna, measured as the number of species regularly breeding, S B . So that the data be normally distributed, the regression analyses were performed upon log 10 transformed data only, with the data for outliers, mainland Britain and Ireland, the two largest islands, excluded. Most highly correlated with the number of butterfly species breeding upon an island is the number breeding within a 25 km radius of the nearest point of the mainland, r 2 = 0.5941, followed by the correlations with the latitude of the mid-point of the island, r 2 = 0.5541, the number of plant species comprising the island flora, r 2 = 0.5225, and the distance separating the island from the mainland, r 2 = 0.4514. A partial correlation analysis confirms the importance of the parameters distance separating the island from the mainland, D 1, and the size of the faunal source S F , and rejects the importance of the size of the Hora and the latitude of the island. This is further confirmed by the results of a step-wise regression analysis, the two variables D 1 and S F accounting for 66% of the variation of the butterfly fauna. If an alternative measure of isolation, D 2, which allows for the geographical clumping of islands, is combined with the variable S F , then 69% of the variation of the butterfly fauna is accounted for.