Due the universal occurrence of n-alkanes, this type of hydrocarbon is assumed not to be relevant in ant communication, and indeed experimental data proved that ants do not respond to n-alkanes (see reviews by Martin and Drijfhout, 2009 and van Wilgenburg et al., 2011). It is therefore unlikely that paraffin influenced the outcome of the behavioural assays. Observers were situated 1.5 m FK228 chemical structure from focal trial, and ant behaviour and number of visits were observed for 1-min periods in 910 censuses. Censuses began at 9 AM and continued

up to 4 PM during three days accounting for a total of 910 min of field observations. In the course of the experiment, we additionally recorded the presence of all ant species that were active in the area occupied by the Cytinus population, irrespective of their activity or their attraction to Cytinus plants. Regardless of population, inflorescence and flower sex, the amount of scent trapped was quite variable (overall 0.2–31.4 ng on a per hour and flower basis). We therefore focused our analysis on relative (percentage of the total peak area) rather than absolute amounts of scent components. Semiquantitative similarities in floral scent patterns among samples were calculated with the Bray–Curtis similarity index in the statistical

software PRIMER 6.1.11 (Clarke and Gorley, 2006). To test for scent differences between female and male flowers, we calculated a PERMANOVA (10,000 permutations, in PRIMER 6.1.11) based on the Bray–Curtis similarity matrix. PERMANOVA is a technique else for testing the simultaneous 17-AAG order response of one or more variables to one or more factors in an ANOVA experimental design on the basis of a (dis)similarity (distance) matrix with permutation methods (Anderson et al., 2008). The analysis employed a two-way crossed design with sex as the fixed factor and inflorescence as the random factor. This

analysis revealed that female and male flowers of a specific inflorescence emitted the same scent (see Results). We therefore calculated the mean relative amount of scent for each inflorescence, computed semiquantitative similarities (Bray–Curtis similarity index) in scent patterns among inflorescences, and used these data for all further analyses. Nonmetric multidimensional scaling (NMDS) was performed (based on the Bray–Curtis similarity index) to depict variation in floral scent among the inflorescences (Clarke and Gorley, 2006). Nocturnal and diurnal samples occupied similar locations in a 2-dimensional odour space, and similarity within nocturnal and diurnal samples was not higher than similarity between nocturnal and diurnal samples (PERMANOVA: Pseudo-F1,17 = 0. 65, P = 0.62). A PERMANOVA analysis to test differences in scent among populations (10,000 permutations; fixed factor:population) was then applied to pooled diurnal and nocturnal data.