Landmarks could be recorded from a total of 504 out of 631 photographed specimens (60.1% of all collected specimens): 442 shells of C. galea, 55 of C. caribaea and seven of C. curacaoensis sp. nov. (Table 3). Principal component analysis on landmark data of all three species revealed six axes, explaining 71.0% of all observed variance, with an ICC > 0.80 that could therefore be considered repeatable. While overlap in shell shape between species existed, all three species were separated on the first and third PC axis, which explained 31.2% and 10.4% of all variances in shell shape (Fig. 8). On the first PC axis, Coralliophila galea shells separated from both C. caribaea (p < 0.0001) and C. curacaoensis sp. nov. (p < 0.0001). On the third PC axis, shells from C. curacaoensis sp. nov. were also separated from C. caribaea shells (p = 0.016). In total, species identity accounted for 19.5% of all observed variance in shell shape.
Intraspecific variation in shell shape of both C. galea and C. caribaea was high (Figs. 9-10). For C. galea, principal component analysis again revealed five repeatable axes (ICC > 0.80, explaining 62.3% of variance in shell shape). For C. caribaea, four repeatable axes were found (ICC > 0.80, explaining 72.2% of variation) by the principal component analysis. Compared to C. galea, the first two PC axes of C. caribaea explained more of the intraspecific variance in shell shape. Most of the intraspecific variation on the repeatable PC axes of both species was related the shape and relative size of the shell spire. Despite high intraspecific variation, no distinct ecotypes could be distinguished in either species, as all specimens clustered together into one cloud without gaps.
For both C. galea and C. caribaea, all factors and interactions had a significant influence on shell shape (Procrustes ANOVA model; Table 4). Though all tested factors contributed to shell shape, the explained variance in shell shape by any factor was low (R2 < 0.20 for all factors and interactions) and residual variance was high (R2 = 0.777 for C. galea and R2 = 0.429 for C. caribaea).
Host-associated differences in shell shape accounted for some of the intraspecific variation in shell shape of Coralliophila spp. In C. galea, differences in host species explained 6.9% of variance in shell shape (Table 4). Among snails from different hosts species, eleven pairwise differences in shell shape were found (Table S4 in Online Supplementary Material 4), that, even though statistically significant, were subtle, and strong overlap in shell shape existed among snails from different host species (Figs. 9-10).
Shell shape of C. caribaea also differed among host species. Firstly, snails originating from hosts of the order Alcyonacea and snails from Scleractinia differed in shell shape, which accounted for 9.3% of the observed variance in shell shape (Table 4). At the host genus level, two pairwise differences were significant (Table S5).
Depth had a small, though significant, effect on shell shape in both C. galea and C. caribaea (Table 4). On top of an overall effect of depth, and a small host-specific effect of depth was observed in both C. galea and C. caribaea.
Since C. curacaoensis sp. nov. only occurred on M. auretenra, host-related differences could not be assessed. Shell length did not contribute to intraspecific variations in shell shape within this species (F = 2.1; p = 0.070).