Statistics and reproducibility

Isotopic datasets were evaluated to determine if the data were normally distributed and if the variance was equal between groups using Levene’s test. A non-parametric Kruska–Wallace test followed by a Dunn’s post hoc test with a Bonferroni correction for pairwise comparisons was used when the requirements for ANOVA were not satisfied; otherwise, analysis of variance (ANOVA; one-tailed) was used to identify statistically significant differences in isotopic values between groups for modern African fauna. Where ANOVA indicated statistical significance, pairwise comparisons were made using a Tukey–Kramer HSD post hoc test to determine which groups differed from one another in their isotope values. Statistically significant relationships between paired isotopic values from the same individuals were determined using Pearson’s correlation (two-tailed), Kendall Rank correlation, or Spearman Rank correlation (two-tailed) where appropriate. Statistical analyses were performed using Paleontological Statistics Version 4.09 (PAST4) and JMP®, Version 16 statistical software using an alpha level for the significance of 0.05. Detailed results of pairwise comparisons can be found in Tables S2–S4. Sample sizes and the number of replicates are stated in the Results section as well as in Supplementary Data 1–4.