3.1. Variation in Sexual Size Dimorphism (SSD)
Overall, the average body size differed significantly across the four study populations (Kruskal-Wallis; chi-squared=64.713, df=3,P <0.0001) (Figure 3). Males were significantly larger than females in both Singapore populations (Kruskal-Wallis: CCNR: chi-squared=18.451, df=1, P <0.0001, Pulau Ubin: chi-squared=38.501, df=1, P <0.0001) and Langkawi (Kruskal-Wallis; chi-squared=7.931, df=1, P <0.01), indicating a clear male-biased SSD. Although males in Central Peninsula MY were also larger than females but this difference was not statistically significant (Kruskal-Wallis; chi-squared=1.818, df=1,P >0.05). In addition, the SDI was more pronounced in Singapore populations (CCNR=-1.09, Pulau Ubin=-1.12; Central Peninsular MY=-1.03 and Langkawi=-1.03), even though the average body size of males in Malaysia populations are bigger than Singapore population.
The average body size of females differed significantly among all populations (Kruskal-Wallis; chi-squared=69.037, df=3,P <0.0001). Females from Malaysia populations were significantly larger than females from Singapore populations (post-hoc Dunn’s test; Central Peninsular MY versus CCNR and Pulau Ubin:P <0.0001, P <0.0001; Langkawi versus CCNR and Pulau Ubin: P <0.0001,P <0.0001).  The average body size of males also differed significantly among all populations (Kruskal-Wallis; chi-squared=16.145, df=3, P <0.01). Using post-hoc Dunn’s test, males from Langkawi were significantly larger than males from Singapore populations (CCNR:  P <0.01; Pulau Ubin:P <0.01) but did not differ significantly with Central Peninsular MY. Body size also did not differ significantly between the Singapore populations and Central Peninsular MY (post-hoc Dunn’s test; Central Peninsular MY versus CCNR and Pulau Ubin:P >0.05) (Figure 3).

3.2 Variation in male reproductive traits as a function of body size

Using a log-transformed data and the breakpoint model, a hyperallometric relationship (allometric coefficient, β>1, Figure 4a, Table 1) was found between horn length and body size (pronotum width) for all four populations of O . babirussa (see Appendix 3 for the best fit model for each population). The adjusted R2 values for equation one of the breakpoint model were high for all four populations, signalling a strong positive correlation. In addition, 95% confidence intervals (CI) for equation one of all populations excluded zero, ruling out the likelihood of a zero slope, indicating a significant relationship between horn length and body size. On the other hand, 95% CI for equation two includes zero for all populations except Pulau Ubin, signalling a non-significant relationship, hence only the allometric coefficient of equation one (β1) was taken into consideration. Together, these results suggest that body size is a significant factor in explaining horn length variation, where larger males have disproportionately longer horns. Interestingly, there is an overlap in CI values for all populations which suggests that there were no significant population-level differences in allometric relationships.
Next, using log-transformed data, a positive and hypoallometric relationship (allometric coefficient β<1, Figure 4b, Table 1) was found between testes weight and body size (pronotum width) for CCNR, Pulau Ubin and Langkawi, and a hyperallometric relationship was found for Central Peninsular MY. The adjusted R2 values for the linear model of all populations were low, indicating a weak correlation. In addition, the p-values of the slopes of the four linear models (P>0.05, Table 1) and 95% CI (includes zero) suggests that body size is not a significant factor in explaining testes weight variation.
Following that, using log-transformed data, a positive and hypoallometric relationship (allometric coefficient β<1, Figure 4c, Table 1) was found between sperm length and body size (pronotum width) for CCNR, Pulau Ubin and Langkawi, and a negative and hypoallometric relationship was found for Central Peninsular MY. The adjusted R2 values for the linear model of all populations were low, signalling a weak correlation. In addition, the p-values of the slopes of the four linear models (P>0.05, Table 1) and 95% CI (includes zero) suggests that body size is not a significant factor in explaining sperm length variation.