Traps. Traps were 150ml cylindrical plastic vials (diameter 45 mm length 102 mm) containing a roll of (380 x 95mm) standard corrugated cardboard (corrugation height 3mm, width 0.79mm). Vials were sprayed silver to prevent solar heating. We placed traps at the corners of a 1m quadrat; quadrats were located haphazardly on each of the islands. Traps were placed open-end-down, and attached to a cane. The open end of the vial was in contact with the ground and vegetation was pulled over the vial.
Biomass of nesting birds. The species surveyed were gannet Morus bassanus, lesser black backed Larus fuscus and herring gull L. argentatus, puffin Fratercula arctica, sandwich tern Sterna sandvicensis, common tern Sterna hirundo, arctic tern Sterna paradisaea, roseate tern Sterna dougalliiand eider Somateria mollissima. With the exception of eiders all the birds have altricial young and therefore produce terrestrial organic matter during chick rearing. Cliff-nesting species were not included in the data set as storms wash organic matter from the cliffs each winter. The average number of survey years was 10.6 ± 1.6 per species, per island, and the average number of annual surveys of any species was 65.1 ± 11.9 per island. Lindisfarne and Crammond are connected to mainland by causeways and are large enough to support terrestrial predators. These islands have fewer ground nesting birds than expected for their area; Crammond is not surveyed for this reason. Bass Rock is a gannet colony, consequently the huge numbers (39,000) of these heavy birds means the island has a much greater biomass of nesting birds than expected for its area. Excluding Crammond, Lindisfarne and Bass Rock from the analysis, a strong isometric relationship exists between log island area and the log biomass of ground nesting birds (F(1,17)= 48.52, P < 0.001, r2 = 0.74, predicted log biomass = -1.214 + log area x 1.01) suggesting that by-and-large the biomass estimates do correlate with stable habitat features. In order to calculate the biomass of birds relative to island area, we used the latter regression equation to obtain residuals from what is the principle across-island relationship.
Sex Ratio. Female earwig density was a decreasing function of male density across islands: the reduced major axis slope of log number of females on log number of males was 0.742 (CI = 0.63-0.88), significantly less than unity (F1,21 = 13.82, P = 0.001). To distinguish between the effects of density and increasing male bias in the sex ratio with density, we calculated the deviance in female density from a 1:1 ratio. Deviance in sex ratio was partialled from its correlation with population density (r22 = 0.83 P< 0.001) using a bivariate principal components analysis: PC 1 is the variance in the major axis i.e density and PC 2 is the variance in the, perpendicular, minor axis i.e. deviations in sex ratio independent of density.