The otoliths of teleost fishes exhibit a great deal of inter- and intra-species shape variation. The ecomorphology of the saccular otolith is often studied by comparing its shape across species and populations inhabiting a range of environments. However, formal tests are often lacking to examine how closely variation in otolith shape follows the genetic drift of a neutral trait. Here, we examine patterns of saccular otolith shape variation in four species of African cichlid fishes, each sampled from three field sites. All four species showed the greatest level of otolith shape variation along two principal component axes, one pertaining to otolith height and another to the prominence of an anterior notch. Fish collected from the same site possessed similarities in saccular otolith shape relative to fish from other sites, and these ‘site-difference’ signatures were consistent across species and observable in both sexes. Sex-differences in saccular otolith shape differed in magnitude from site to site. Population differences in saccular otolith shape did not covary with neutral genetic differentiation between those populations. Otolith height, in particular, displayed large site similarities across species, weak correlation with neutral genetic variation, and strong sex differences, collectively suggesting that otolith shape represents a selectively non-neutral trait.

Saccular otoliths and fin clips were sampled from both males and females (spanning a wide body size range) of four Lake Tanganyikan cichlid species (Neolamprologus caudopunctatus, N. pulcher, N. savoryi, Variabilichromis moorii) from three distinct field sites at the southern tip of the lake. Two-dimensional otolith shape was computed using elliptical Fourier shape analyses followed by principal component analysis. Neutral genetic population differentiation was estimated by sequencing, and quantifying variability in, the mitochondrial control region. Differences in otolith shape between populations were then compared to the degrees of neutral differentiation between those same population pairs (see Methods).

There are no missing values in these data sets (although any otoliths that were chipped or vateritic were not used in this study). Column names in the .csv files are self-explanatory (see Methods).