Detecting conspecific brood parasitism using egg morphology in black brant Branta bernicla nigricans

Numerous methods have been proposed to indirectly detect conspecific brood parasitism (CBP) in birds. Egg morphology has been suggested as a predictor of parasitism, assuming that variation in egg size is greater among females than within females. Here we use microsatellite data to assess the use of egg morphology to detect CBP in a sample of black brant Branta bernicla nigricans nests. We attempted to repeat a previously demonstrated technique using cluster analysis and maximum Euclidean distance (MED) to detect parasitized nests within black brant. Additionally we attempted a new technique based on a discriminant function analysis of egg morphology in an attempt to detect brood parasitic eggs. When detecting parasitized nests using egg morphology, the cluster analysis revealed that the MED between the two most dissimilar eggs in each nest was significantly greater for parasitized nests than for non‐parasitized nests (1.62±0.06 and 1.43±0.08, respectively). The extent of overlap in sizes of eggs between parasitized and non‐parasitized nests, however, was such that we were unable to effectively identify parasitized nests. In most cases for each parasitized nest correctly identified, 3 non‐parasitized nests were incorrectly identified as parasitic. When we attempted to detect parasitic eggs we found that parasitic eggs were more different from the expected egg volume than host eggs: mean absolute residual volume of parasitic eggs=2.59±5.79 cm3 while that for host eggs=1.82±2.14 cm3. Overall, we found that the discriminant function analysis was moderately effective in determining whether eggs belonged to the host female using a resubstitution technique (error rate=9.71%) or a jackknife technique (error rate=6.12%). Additionally, we found a higher but moderate error rate when using an independent data set to validate the function (error rate=14.07%). In both cases, however, parasitic eggs accounted for most of the error and were not correctly classified 75%, 70% and 100% of the time respectively. We suggest when developing a predictive function for detecting conspecific brood parasitism based on egg morphology that an appropriate technique be used to validate the function, particularly those techniques that utilize unambiguous identifiers such as molecular and protein fingerprinting techniques.