This study aimed to compare the swimming kinematics and hydrodynamics within and among aquatic and semi-aquatic/terrestrial frogs. High-speed video was used to obtain kinematics of the leg joints and feet as animals swam freely across their natural range of speeds. Blade element analysis was then used to model the hydrodynamic thrust as a function of foot kinematics. Two purely aquatic frogs, Xenopus laevis and Hymenochirus boettgeri, were compared with two semi-aquatic/terrestrial frogs, Rana pipiens and Bufo americanus. The four species performed similarly. Among swimming strokes, peak stroke velocity ranged from 3.3+/-1.1 to 20.9+/-2.5, from 6.8+/-2.1 to 28.6+/-3.7 and from 4.9+/-0.5 to 20.9+/-4.1 body lengths per second (BL s(-1)) in X. laevis, H. boettgeri and R. pipiens, respectively (means +/- s.d.; N=4 frogs for each). B. americanus swam much more slowly at 3.1+/-0.3 to 7.0+/-2.0 BL s(-1) (N=3 frogs). Time-varying joint kinematics patterns were superficially similar among species. Because foot kinematics result from the cumulative motion of joints proximal to the feet, small differences in time-varying joint kinematics among species resulted in species-specific foot kinematics (therefore hydrodynamics) patterns. To obtain a simple measure of the hydrodynamically useful motion of the foot, this study uses 'effective foot velocity' (EFV): a measure of the component of foot velocity along the axis of swimming. Resolving EFV into translational and rotational components allows predictions of species-specific propulsion strategies. Additionally, a novel kinematic analysis is presented here that enables the partitioning of translational and rotational foot velocity into velocity components contributed by extension at each individual limb joint. Data from the kinematics analysis show that R. pipiens and B. americanus translated their feet faster than their body moved forward, resulting in positive net translational EFV. Conversely, translational EFV was slower than the body velocity in H. boettgeri and X. laevis, resulting in negative net translational EFV. Consequently, the translational component of thrust (caused mostly by hip, knee and ankle extension) was twofold higher than rotational thrust in Rana pipiens. Likewise, rotational components of thrust were nearly twofold higher than translational components in H. boettgeri. X. laevis, however, was the most skewed species observed, generating nearly 100% of total thrust by foot rotation generated by hip, ankle and tmt extension. Thus, this study presents a simple kinematics analysis that is predictive of hydrodynamic differences among species. Such differences in kinematics reveal a continuum of different propulsive strategies ranging from mostly rotation-powered (X. laevis) to mostly translation-powered (R. pipiens) swimming.