Animal performance can be determined by subjective observations or objective measurements. Numerical data are only then superior to results of subjective observations when they are the result of measurements carried out to test a well-defined hypothesis or to give the answer to a clear, precisely formulated question. In the analysis of kinematics a careful evaluation of the set-up of the measurement equipment and the resulting accuracy in the data is required. Measurements in three dimensions (3D) are theoretically better than those in 2D. Practically, however, collection, analysis, interpretation and presentation of 3D data are so much more complicated that frequently 2D analysis appears to be more useful. The minimal size of markers necessary to obtain a certain accuracy in kinematic data is usually too big for practical use. Smaller markers impair accuracy. Reduction of measurement noise is obligatory when time derivatives are to be calculated. Skin movement artefacts cannot be removed by data smoothing. Forces occurring between the digits and the ground can be determined using a force plate or an instrumented shoe. A force plate is accurate, but repeated trials are necessary. Using a force shoe each ground contact results in useful data. However, the shoe itself may affect locomotion. Surface strains on long bones can be recorded relatively easily. Determination of loading forces from surface strains is complicated but can be carried out using multiple strain gauges and a post-mortem calibration test. Strain in tendons is difficult to measure due to problems in defining a ‘zero’ or reference length. Tendon forces can either be calculated from signals of parallel, compliant transducers or estimated by application of tendon buckles or wedge-shaped implantable transducers. In both cases a calibration procedure is required to convert transducer signals to tendon for