Testing Experimental Results in the Field: Comment on Ficetola and De Bernardi (2005)

Abstract

Ficetola & De Bernardi (2005) test the effects of reproductive interference between Rana latastei and Rana dalmatina on the former’s reproductive success under natural conditions. Their study examines predictions by Hettyey & Pearman (2003), based on natural history observations and experimental results, that reproductive failure may occur because of heterospecific matings when species relative abundance is highly skewed. Ficetola & De Bernardi (2005) measure R. latastei reproductive success, represented by mean embryo viability across clutches raised in the laboratory, using three populations that are syntopic with R. dalmatina and two other populations, where this congener is absent. Ficetola and De Bernardi find no decrease in reproductive success of R. latastei populations in syntopy. They also find no relationship between the relative abundance of heterospecifics and R. latastei reproductive success. Ficetola & De Bernardi (2005) conclude that heterospecific individuals do not interfere with R. latastei reproduction under natural conditions and they emphasize validating experimental results using natural populations. We agree with Ficetola & De Bernardi (2005) that tests of experimental results under natural conditions are needed. Determining an optimal experimental setup is difficult because different setups may yield different results (e.g. Michalak & Rafinski 1999). Further, influential factors in experimental manipulations may be unimportant under natural conditions (e.g. Skelly & Kiesecker 2001). In this sense, the aim of Ficetola & De Bernardi (2005) to test the validity of experimental results using natural populations is admirable. Nonetheless, correlative tests of experimental studies must be designed with statistical rigor and a thorough understanding of the original studies in order to provide valid tests of predictions. Ficetola & De Bernardi (2005) is flawed in this respect. The design in Ficetola & De Bernardi (2005) creates two biases toward accepting the null hypothesis that reproductive interference by R. dalmatina does not affect R. latastei in nature. Firstly, Ficetola & De Bernardi (2005) ignore the result that the relationship between species relative abundance and effectiveness of sexually isolating mechanisms is strongly non-linear (Hettyey & Pearman 2003). Abundance of R. latastei and R. dalmatina in the syntopic study populations of Ficetola and De Bernardi is estimated to be 1:1, 1:2, and 1:5. Hettyey & Pearman (2003) observe no significant decrease in fertilization success in trials, where R. latastei males are not outnumbered by R. dalmatina males by at least 1:5, probably because of the mitigating effect of mate recognition mechanisms. The results and model of Hettyey & Pearman (2003) predict significant decrease in average reproductive success in nature only when one species occurs at very low relative abundance. Decreased reproductive success where R. latastei’s relative abundance is 1:1 and 1:2 is not expected, thus grouping all syntopic populations together biases Ficetola and De Bernardi’s test. Further, absence of a bimodal distribution of embryo viability and no decrease in embryo viability in the one population where R. latastei was outnumbered by 1:5 would suggest that conspecific–heterospecific ratios be <1:5 in some natural populations to disrupt sexual isolation. However, one cannot draw such a conclusion from the single population in Ficetola & De Bernardi (2005) in which the phenomenon was expected. Secondly, Ficetola & De Bernardi (2005) suffers from small sample size, which biases the study Ethology