Suction-feeding in fish is a ubiquitous form of prey capture whose outcome depends both on the movements of the predator and the prey, and on the dynamics of the surrounding fluid, which exerts forces on the two organisms. The inherent complexity of suction-feeding has challenged previous efforts to understand how the feeding strikes are modified when species evolve to feed on different prey types. Here, we utilize the concept of dynamic similarity, commonly applied to understanding the mechanisms of swimming, flying, walking, and aquatic feeding. We characterize the hydrodynamic regimes pertaining to 1) the forward movement of the fish (ram), and 2) the suction flows for feeding strikes of 71 species of acanthomorph fish. A discriminant function analysis revealed that feeding strikes of zooplanktivores, generalists, and piscivores could be distinguished based on their hydrodynamic regimes. Furthermore, a phylogenetic comparative analysis revealed that there are distinctive hydrodynamic adaptive peaks associated with zooplanktivores, generalists, and piscivores. The scaling of dynamic similarity across species, body sizes, and feeding guilds in fish indicates that elementary hydrodynamic principles govern the trophic evolution of suction-feeding in fish.

Data on fish feeding kinematics were obtained from previously published studies. These encompassed feeding strikes of 15 species of Centrarchidae (Holzman et al. 2012), 27 species of Serranidae (Oufiero et al. 2012), and 21 species of Cichlidae (Hulsey et al. 2010). Original movies from the cichlid study (Hulsey et al. 2010) were reanalyzed
to extract all the kinematic variables relevant to this study (ESM 1. Data collection). These data were supplemented with unpublished data on two species of Antennariidae, two species of Centrarchidae, four species of Cichlidae, four species of Pomacentridae, and six species of Serranidae.

All data were collected in a similar way, with fish being filmed from a lateral view at 500 fps (Hulsey et al., 2010; Holzman et al., 2012; Oufiero et al., 2012, using NAC Memrecam Ci Tokyo Japan), or 1000 fps (Oufiero et al., 2012, using Fastec HiSpec 1 San Diego CA USA), while striking an evasive prey (small guppies Poecilia reticulata in Hulsey et al., 2010; ghost shrimp in Holzman et al., 2012; and zebrafish Danio rerio and mosquitofish Gambusia sp. in Oufiero et al., 2012). After each trial, a ruler was filmed to enable scaling. For each recorded strike, landmarks on the fish were tracked in space and time using the DLTdv package for MATLAB (Hedrick, 2008) or a similar digitizing software. Specifically, gape size was measured as the distance between the anterior tips of the premaxilla and the lower jaw. Because gape expansion is typically asymptotic at onset and end of a strike, the mouth opening phase was demarcated by the point where gape is 20% of maximum gape and the point where gape is 95% of maximum gape (Oufiero et al., 2012). The time to peak gape (TTPG) was measured as the duration of the mouth opening phase. Ram speed was measured as the slope of the linear regression of the forward cumulative displacement of the body during the mouth opening phase as a function of time, measured by tracking a landmark on the body.

We used published literature to determine feeding guild, and prioritized sources that reported primary data (stomach content analysis). We classified fish as herbivores, detritivores, molluskivores, generalists, zooplanktivores, or piscivores. When primary data were available, species were classified according to the dominant food type, recognizing that many fish are opportunistic and may consume any food item that is readily available, irrespective of feeding specializations, in which case we also acknowledged a feeding guild designation given by the authors, if backed up by data (for example Lepomis microlophus is a molluskivore specialist, but stomach content depends on both season and fish body size (Desselle et al., 1978)). Fish that were primarily recorded as feeding on zooplankton, copepods and calanoids were classified as zooplanktivores and fish that primarily fed on other fish were classified as piscivores. Fish that were recorded as feeding on a variety of prey, such as gastropods, decapod crustaceans, mollusks, and plants were classified as generalists. Fish that were recorded as specializing on mollusks, plants, or detritus were classified as molluskivores, herbivores and detritivores, respectively.

Original data sometimes contained multiple recordings from each fish, or used several individual fish from each species. To avoid bias, we calculated the average value for each individual and the the average value for the species. The total number of fish and the average number of strikes from each fish are listed in columns No. fish and No. strikes, respectively.