Species Comparisons:
Qualitatively, as hypothesized, in P. barbatus with limited dimorphism, females (workers and gynes) did not group by shape to the exclusion of workers, but rather males were intermediate in shape between workers and gynes (MDS2 in Fig. 2C). Contrary to the hypothesis, though, in the dimorphic P. badius , workers most differed from sexuals in shape and each workers and sexuals had distinct size-shape axes (Fig. 2D). Interestingly, the workers in P. barbatus do have a significant size by shape relationship (F1,35 = 44.52, P < 0.0001, R2 = 0.55) while shape is not predicted by size in males and gynes of P. barbatus . In P. badius , there is not a significant size-shape relationship in any caste, though visually this appears to possibly be a sample size artifact. That said, size predicts shape for each workers and sexuals, and these slopes are parallel (total model: F3,78 = 183, P << 0.0001, Size x shape: P << 0.0001, interaction: P = 0.14). Also very interestingly, the size-shape slope for P. barbatus workers is not different from that ofP. badius workers, potentially suggesting conservation of the growth mechanisms governing shape change with size in this genus.
Predictable traits helped separate castes in each species; note, below, trait size as referenced is trait size accounting for body size. The distance from the origin in Fig. 2C-D corresponds to the strength of correlation between each MDS axis and the measured characters. In both species, trunk/mesosomal height helped differentiate sexual castes due to the thoracic enlargement that accommodates wing muscle. Leg length (for all legs and leg segments) and head size tended to help separate workers from sexuals in P. barbatus , but helped separate minor workers/males from major workers/gynes in P. badius . Gaster size tended to help separate gynes in both species. These qualitative comparisons of castes by the measurements correspond with casual observation and common sense.
Quantitatively, as expected, the first multivariate scaling axis is highly correlated with ant length in both species, thus approximating size (P. barbatus : F1,77 = 757, P << 0.0001, R2 = 0.91, P. badius : F1,80 = 1269, P << 0.0001, R2 = 0.94)(Fig. 2A-B). In both P. barbatus andP. badius , all castes were distinct in size-shape space (PERMANOVA: P. barbatus : F2,76 = 14.366, P < 0.001, P. badius : F3,78 = 89.881, P < 0.001; all castes different from each other at P < 0.005 in both species)(Fig. 2C-D) as well as in shape/MDS2 (ANOVA:P. barbatus : F2,76 = 51.57, P << 0.0001, P. badius : F3,78 = 197, P << 0.0001; all castes different from each other at P < 0.005 in both species).