The distribution of NBHF clicks on the odontocete tree is identical to that of non-whistling species after the divergence of the Physeteroidea (Physeteridae & Kogiidae) from other odontocetes (Fig. 7). The lack of congruence before this split is explained easily; Kogia evolved NBHF clicks independently, but not whistle loss because they never whistled in the first place.
Several authors have suggested that the peak frequency of echolocation clicks is negatively correlated with body length in odontocetes (Watkins, 1980; Thomas et al., 1988; see review Tyack & Clark, 2000). We reanalysed the data, adding recently published data from various sources based on reliable recordings (Table 1, Fig. 8). Some of the scatter in the figure is likely due to the uncertainty of this measurement (Madsen et al., 2004b). In spite of this, the higher peak frequency, lower peak frequency and centroid frequency of echolocation clicks show negative correlations with body length in odontocete species except NBHF species (anova; F1,13 = 6.05, P < 0.05; F1,13 = 4.99, P < 0.05; F1,13 = 22.9, P < 0.0001, respectively). If we averaged the data from species within the same family to minimize effects of phylogeny, the higher peak frequency and centroid frequency of clicks still show negative correlations with body length (anova; F1,3 = 60.3, P < 0.05; F1,2 = 15.4, P < 0.05, respectively), but the lower peak frequency shows only weak negative correlations with body length (anova; F1,2 = 14.3, P = 0.06). However, in NBHF species, the peak frequency of clicks exhibits no correlation with body length (anova; F1,6 = 0.32, P = 0.59 all species included; F1,2 = 0.47, P = 0.56 for family averages). The surprisingly high frequency clicks reported for L. albirostris (Rasmussen & Miller, 2004) are interesting but other delphinids may render equally high bandwidths when recorded with the same wide band gear used by Rasmussen & Miller (2004). For the species with a body length < 3 m (except L. albirostris), higher peak frequency, lower peak frequency and centroid frequency of clicks do not show any correlations with body length (anova; F1,8 = 1.51, P = 0.25; F1,8 = 0.03, P = 0.86; F1,8 = 0.01, P = 0.92, respectively). The peak frequency of the clicks by NBHF species differed significantly from the lower peak frequency and centroid frequency of the clicks by non-NBHF species, but did not differ from the higher peak frequency of the clicks by non-NBHF species (anova; F3,34 = 51.2, P < 0.001; Tukey–Kramer's HSD post hoc test, α =0.05). This implies strongly that selection acted against the lower peak frequency in NBHF species.