Do Cleaner Fish (Labroides dimidiatus) Have General Cognitive Ability? A Reanalysis of Individual Differences Data and Consideration of Phylogenetic Context

Aellen et al. (2022) recently suggested on the basis of principal component analysis (PCA) that there is no general cognitive ability (GCA) factor in various cognitive ability measures of wild-caught cleaner fish (Labroides dimidiatus), making this species an oddity-given the apparent ubiquity of this dimension in many animal taxa. They report the presence of three approximately co-equal factors instead, with the first exhibiting a mixture of positive and negative loadings. Reanalysis of their data employing unit-weighted estimation yielded a GCA factor with all positive loadings accounting for 29.9% of the variance. Adding a fourth ability (feeding against preference) yielded a positive manifold accounting for 19.3% of the variance. As this technique for factor estimation typically yields latent variables exhibiting higher generalisability than those obtained via differentially weighted techniques (such as PCA), it is suggestive of what might be found were sample specificity effects to be reduced via more extensive sampling of individuals. Consistent with this possibility, it is found that the proportion of variance associated with unit-weighted estimated GCA in these data is not significantly different from the proportion of variance associated with this factor in a meta-analysis of 12 other animal taxa. Adaptationist theories of GCA make explicit predictions concerning where in the phylogenetic landscape this factor might be expected to be strongly or weakly present, or even absent altogether. These are discussed in detail.

Automated summary

Based on principal component analysis (PCA), Aellen et al. (2022) argue against the existence of a general cognitive ability (GCA) factor in wild-caught cleaner fish. Instead, they find three equally important factors, with the first showing a mix of positive and negative loadings. Reanalyzing the data using unit-weighted estimation reveals a GCA factor with all positive loadings, accounting for 29.9% of the variance. These findings suggest that more extensive sampling could reduce sample specificity effects and lead to the discovery of a GCA factor. Additionally, the proportion of variance associated with the GCA factor in the wild-caught cleaner fish is similar to that found in a meta-analysis of 12 other animal taxa, supporting adaptationist theories of GCA.