Morphometric analyses were based on 15 continuous
morphometric variables (listed in table I, with abbreviations
in squared brackets) of 287 examined specimens, with the
addition of 236 specimens examined by Boeseman (1968:
tables 1-20), leading to a total of 523 specimens representing
43 samples of populations for 15 nominal species. Specimens
smaller than 30 mm were excluded from the analyses
to minimize the bias introduced by allometric growth, and
because morphometric data are highly correlated between
them, missing data were estimated using the least squares
method with the standard length (SL) used as explanatory
variable. To estimate the absence of significant differences
between measurements provided by Boeseman and ours, a
Wilcoxon matched pairs signed rank test was performed on
all common measurements recorded from the holotype of
H. gymnorhynchus. In order to highlight the morphological
structure of the species and populations under study, and to
construct the model used for the assignment of the Linnean
types to an extent species, the data were submitted to multivariate
analyses