3.3 | Correlated evolution of morphology and behavior
Of the morphological features that we tested for correlation with behavior (see Table 1), we identified significant patterns of correlation in eleven (Table 2, Fig. 3). Analysis using Pairwise Comparisons (PC) and Maximum Likelihood (ML) were largely corroborative, with strongest hypotheses of correlation returning the strongest significance values in the PC analysis, and the highest delta-AICc values for the independent model in the ML analysis (indicating the poor performance of this model relative to models of correlation). Because only cases that returned significant correlation in the PC analysis were reanalyzed using ML, in the ML analysis the independent (uncorrelated) model always performed the worst, returning the highest delta-AICc values. Delta-AICc values were usually low for all dependent models, signifying little difference in model-fit between different dynamics of correlation.
Patterns of correlation between the spinnerets and behavior were as expected: elongate posterior lateral spinnerets, widely-spaced spinnerets, and pseudo-segmented apical segments of the posterior lateral spinnerets were strongly positively correlated with the construction of a web at the retreat entrance and to a lesser extent with opportunist retreat construction (almost all web-builders are opportunists), and were negatively correlated with burrowing and entrance modification. In contrast, short apical segments of the posterior lateral spinnerets were most strongly correlated with structural modification to the burrow entrance, and negatively correlated with opportunism and web-building (opportunists virtually never structurally modify their retreat entrance). The rastellum returned strong positive correlation with both burrowing and entrance-modification, however the second pairwise comparison analysis (PC2) revealed many cases of ‘neutral change’ with respect to the rastellum and both of these behaviors. The serrula returned the opposite pattern to the rastellum, being positively correlated with web-building and opportunism, and negatively correlated with burrowing and entrance modification, with negative correlation with burrowing returning the strongest correlation. ‘Digging spines’ did not return a significant correlation with burrowing, but instead with entrance modification (positive). The presence of scopulae showed no correlation with any of the behaviors tested. All modifications to the posterior legs – enlargement relative to the anterior legs, dorsal bias in macrosetation, and presence of a thorn patch on prodorsal patella III – showed a strong positive correlation with both burrowing and burrow-entrance modification. Finally, the presence of a common eye tubercle was positively correlated with opportunistic burrowing (although there are many cases of neutral change, see PC2), and a compact rectangular eye group was negatively correlated with burrow-entrance modification, however the two specific modifications to the eye group that were tested for correlation (widening of the eye group and anteriorly-positioned anterior lateral eyes) did not return significant correlation, despite each only occurring in taxa with modified burrow entrances.