Data compiled from the source papers for this dataset were generated using the following methods.

Microbial abundance (HMA vs LMA status) is enumerated using a combination of scanning and transmission electron microscopy (Weisz et al., 2007, 2008; Gloeckner et al., 2014); chemical defense allocation patterns and palatability are determined by incorporating chemical extracts from sponges into artificial food cubes that are then fed to generalist fish or fish assemblages on coral reefs (Loh and Pawlik, 2014); nutritional quality and physical characteristics of sponge tissue are determined by combustion of dried and weighed tissue in a muffle furnace at 450 C for 12 hours (ash); measuring the force required to tear tissue (tensile strength); NaOH-solube protein analysis using bovine serum albumen as a standard (protein); TCA-soluble carbohydrate content using glycogen as a standard (carbohydrate); a gravimetric lipid assay; and combustion in a bomb calorimeter for caloric energy content (Chanas and Pawlik, 1995). Photosymbiont abundance (chlorophyll a concentration) is determined by extracting sponge tissue in acetone and using a spectrophotometer to estimate the concentration of this pigment (Erwin and Thacker, 2007; Freeman et al., 2020); microbial community diversity (via the Inverse Simpson's index) is determined via 16S metabarcoding of the microbial symbiont community (Easson and Thacker, 2014; Thomas et al., 2016; Gantt et al., 2019; Freeman et al., 2020); NO_x production was assessed using ex situ experimental incubations or collections of water before and after passing through the sponge (Southwell et al., 2008); pumping rate is assessed based on the advancement of a dye front in the excurrent water plume or an acoustic Doppler velocimeter (Weisz et al., 2007; 2008; Pawlik et al., 2018); the proportion of carbon obtained from dissolved organic matter, detritus, and living particulate organic matter (DOC, DET, LPOC) is assessed by collection of incurrent (before entering the sponge) and excurrent (after entering the sponge) water, followed by analysis of water samples using a combination of flow cytometry to enumerate cells and high temperature catalytic oxidation to measure DOC (McMurray et al., 2018; Pawlik et al., 2018); and tissue density is measured as dry mass of tissue divided by volume (measured as water displacement) (Weisz et al., 2007; 2008). Lastly, phylogenetic signal analysis was carried out as in Freeman et al., 2020 using Pagel’s K statistic (Pagel, 1999; Münkemüller et al., 2012). This statistic measures phylogenetic dependence of trait data, with K values close to zero indicating phylogenetic independence and a value of one suggesting that traits are distributed as would be expected under Brownian Motion.