2 (SAS Institute Inc., Cary, NC, USA). Differences in ant choice for natural inflorescence scent or control, and deviations of ant choice from a neutral preference between wicks with synthetic compounds and control were assessed by fitting generalized linear models (procedure GENMOD of SAS) with the binomial error distribution and logit link function. Differences in the number of ants attending to flower scent stimuli and control treatment, and differences in the number of ant visits between synthetic compounds and control,

were compared using procedure GENMOD with the Poisson distribution and log as the link function. A scale check details parameter, estimated by the square root of the deviance of the model divided by its degrees of freedom, was used to correct for overdispersion in the model. Tukey post hoc tests were used to determine which treatments differed significantly. Regardless of population and daytime, compounds emitted by Cytinus flowers consisted of aromatics (eight compounds) and irregular terpenes (three compounds) ( Table 1). Inflorescence axes did

not emit these volatiles. Within inflorescences, emissions from female and selleck inhibitor male flowers conformed to the same scent profile (PERMANOVA: Pseudo-F1,31 = 0.58, P = 0.62), hence further analyses focused exclusively on the inflorescence level. Depending on the inflorescence sampled, (E)-cinnamaldehyde, (E)-cinnamyl alcohol, 4-oxoisophorone, or 4-oxoisophorone epoxide were the most Carnitine dehydrogenase abundant scent compounds ( Table 1). Only rarely (1 of the 18 sampled inflorescences) did benzaldehyde dominate the scent profile. Many samples contained considerable amounts of (E)-cinnamaldehyde along with high amounts of one or two of the other compounds ( Table 1, Fig. 2). The PERMANOVA analysis suggests that semiquantitative variation in scent within populations could be considered

comparable to variation among populations (Pseudo-F3,17 = 1.56, P = 0.14). One would be tempted to suggest that these results point to scent homogeneity across Cytinus races and populations. However, because of the small sample size, these inferences should be interpreted with caution. Results from measurements with ant antennae were very noisy, probably because of strongly chitinized antennae resulting in high electrical resistance (see Material and Methods). However, three runs resulted in responses to compounds clearly differentiated from the noise and demonstrated that ants can perceive the main compounds occurring in Cytinus floral scent ( Fig. 3). Two antennae from two different individuals of A. senilis responded to (E)-cinnamaldehyde, (E)-cinnamyl alcohol and 4-oxoisophorone ( Fig. 3), and one antenna of P. pallidula responded to (E)-cinnamyl alcohol. Six different ant species (A. senilis, C. auberti, C. scutellaris, P. pallidula, P. pygmaea, and T. semilaeve) were recorded in the experimental trials, accounting for 154 visits.