Adult females of An. homunculus and An. cruzii are morphologically similar; consequently, distinction between these species is generally problematic when using only the classical morphological characteristics described in identification keys . The correct identification of species is essential for recognition of the vectors involved in the transmission of malaria and for helping researchers to develop control strategies . Geometric morphometrics analyses revealed that An. homunculus, An. cruzii and An. bellator can be distinguished based on wing characteristics.
The results of the discriminant analysis showed that An. bellator is well separated from both An. cruzii and An. homunculus in the morphospace of the canonical variables. This finding is consistent with previous results indicating that An. bellator can be easily distinguished by the adult external morphology (hindtarsomere 5). However, one would expect higher phenotypic similarity between An. bellator and An. homunculus if the wing shape is directly associated with the close phylogenetic relationships between the species . An. cruzii and An. homunculus, which are occasionally misidentified , were also discriminated despite partial overlapping in the morphospace of canonical variables.
The wing shape divergence among these three species was not as significant as that for species of the genera Culex and Aedes[16, 24], which may be a result of recent diversification of the subgenera  or due to evolutionary constraints. The close evolutionary relationships among Kerteszia representatives might be reflected in the wing shape because of the heritability of this structure, as proposed for other insects . Anopheles bellator and An. homunculus can coexist in bromeliads and compete for resources , and this close association could impose constraints or favour canalisation to the observed phenotype.
Considering that the results obtained either with or without allometry were similar, we conclude that size variation did not interfere with species delimitation. Anopheles homunculus had the largest size among species in this study; however, we cannot ascertain that this comparison holds in nature because size is commonly subject to plasticity [15, 24, 26]. At least for An. homunculus and An. cruzii, it is plausible to consider that the size disparity may be associated with genetic determinism because both species were collected in the larval stage and reared to adults in the laboratory under similar environmental conditions and food resource availability.
Specimen collections were not simultaneous (years 2009 and 2011), what could lead us to believe that interspecific wing shape divergency may be partly a result of asynchronic sampling and microevolutionary changes. This idea is unlikely because the three-year interval is much shorter than the divergence time among the Anopheles species involved . It has recently been reported that wing shape variation in Aedes albopictus can occur within four years  however, such variation is slight in comparison to macroevolutionary changes. Additionally, phenetic distance between Ae. aegypti and Ae. albopictus based on wing shape remained equivalent over several years .
In addition to helping taxonomists identify species, geometric morphometrics maybe used by health professionals to identify species in the future. This technique will not end up with other methods of identification, such as those based on the costal wing spots, but will complement them. Whereas present work was essentially based on wing landmarks, papers from Wilkerson and Peyton (1990)  and Motoki et al. (2009)  successfully used wing spot relative sizes to identify Anopheles species. Although it is not an easy task to simultaneously analyse landmarks and dark spots in Anopheles wings, we hope in a future to combine landmark and spot-based morphometrics, as suggested by Dujardin (personal communication: Jean Pierre Dujardin).
The mean ratio of dimensions x and y is 2/3 in An. bellator, whereas this ratio is nearly 1/1 in An. cruzii and An. homunculus. Remarkably, the length of segment 13–14 (distance x) in some individuals of An. bellator is so short that the segment is almost nonexistent. As far as we know, this vein pattern has not been observed in other culicids. Although the occasionally vestigial segment does not directly contribute to the diagnosis proposed here, it may be worth an investigation. As an example, the absence of a wing vein in Drosophila melanogaster was characterised as an informative mutation named crossveinless. Additionally, the mean ratio z/w of An. homunculus was lower than those of the other two species. Accordingly, as shown in Figure 5, the wings of Anopheles homunculus are narrower anteroposteriorly.
Apart from the morphological characteristics used in this study, molecular taxonomic markers have been developed for Kerteszia species , facilitating species identification and delimitation. The employment of geometric morphometric methods in taxonomic studies is promising and should be performed in conjunction with other methods to facilitate the correct identification of anopheline species.