The tree of life is highly asymmetrical in its clade wise species richness and this has often been attributed to variation in diversification rates either across time or lineages. Variations across lineages are usually associated with traits that increase lineage diversification. Certain traits can also hinder diversification by increasing extinction and such traits are called evolutionary dead-ends. Ecological specialization has usually been considered as an evolutionary dead-end. However, recent analyses of specializations along single axes have provided mixed support for this model. Here, we test if fossoriality, a trait that forces specialization at multiple axes, acts as an evolutionary dead-end in squamates (lizards and snakes) using recently developed phylogenetic comparative methods. We show that fossoriality is an evolutionary dead-end in snakes but not in lizards. Fossorial snakes exhibit reduced speciation and increased extinction compared to non-fossorial snakes. Our analysis also indicates that transition rates from fossoriality to non-fossoriality in snakes are significantly lower than transition rates from non-fossoriality to fossoriality. Overall our results suggest that broad scale ecological interactions that lead to specialization at multiple axes limit diversification.

Appendix S1_Table1_Snake_microhabitat_dataMicrohabitat data described as fossorial and non-fossorial along with references for 1256 species of snakesAppendix_S1_Table_SuppInfo_Snake_microhabitat_data_22Dec.xlsxAppendix_S2_Table2_Lizard_microhabitat_dataMicrohabitat data described as fossorial and non-fossorial compiled for 822 species of lizards from the GARD database (http://www.gardinitiative.org/data.html).Appendix_S2_Table_SuppInfo_Lizard_microhabitat_data.xlsxAppendix_S3_SupportingInfoSupporting information containing results of model selection and parameter estimation from the BiSSE, HiSSE and STRAPP analyses carried out on the snake and lizard microhabitat data.Appendix_S4_supplementary methodsThe file describes the methods used to evaluate the effect of tip ratio bias on HISSE along with the results of the analyses.