To refer to this article use this url: http://ctoz.nl/vol85/nr03/a03

#####
*Statistics*

The density of muricids and its variation over depth was analysed through both Mann-Whitney-Wilcoxon tests and linear regression. The distribution of age groups over depth and substrates were compared through Chi-squared tests. Distribution over reefs in general for *D*.* rugosa* was investigated with linear regressions and generalized linear models. Median snail group sizes on coral colonies and among age groups were compared through Kruskall-Wallis tests and investigated further through Mann-Whitney-Wilcoxon tests. Statistical analyses were performed in RStudio (R Development Core Team 2010).

Prey preferences were assessed for *D*.* rugosa* through resource selection functions (Manly et al., 1993). This was done by estimating the selection ratio (ωi) for all potential prey genera and growth forms and the associated Bonferroni corrected 95 and 99% confidence interval with the formulas:

where *o*
_{i} is the proportion of occupied colonies of coral genus* i *among all occupied colonies of all genera, *a*
_{i} is the proportion of available colonies of coral genus *i* among all available colonies of all genera, *Z*
_{a/2k}
is the critical value of a standard normal distribution upper tail area of ∝*/2k*, *k* is the total number of coral genera in the analysis and *u*
^{+} is the total number of coral colonies of all genera that are occupied. The confidence interval suggested that a prey was attacked less than expected from its relative availability when below 1 and more than expected from its relative availability when above 1. When the confidence interval encompassed 1, the prey species was not considered attacked significantly less or more than could be expected from its relative abundance. In addition to the resource selection of all *D*.* rugosa*, selection of juveniles and adults was calculated as well. The functions were applied according to Schoepf et al. (2002), although coral genera were taken into account, not species. All present scleractinian genera were included, as they were all considered potential prey species. This approach widens the resulting confidence intervals, making it less likely to find significant deviations from expected colony occupations. When availability of a coral genus was considered too low to be representative (< 7 colonies), such a genus was omitted from the analysis. Colonies of < 10 cm in diameter were also omitted from the analysis.

Since the analysis does not take *D*.* rugosa* group size into account, the mean group size per coral genus was compared with the results afterwards. Furthermore, when noting presence/absence (use/non-use) it can be difficult to demonstrate if the host coral is also used as prey (Boyce et al., 2002). This seems less problematic for *Drupella* snails, which feed and rest on the same colony, directly linking presence to use and justifying the use of this method.