intra-tester reliability is fundamental in determining the quality of data collected in research. few controlled studies have reported the reliability of strength tests and, in spite of most published studies reporting it to be good (0.79 to 0.99), differences between test and retest are observed to be statistically significant. thus, for research purposes, it is suggested that values should be taken from a second test, at least, so that changes in strength may be attributed to treatment effect and not simply to adaptation to the test protocol. the relationships between maximum strength tests and submaximal tests or anthropometric variables have been investigated in order to predict maximal strength without submitting subjects to a maximal load test, so as to avoid the risk of injury. maximal load, or a percentage of it, is commonly used to better prescribe training. prediction of one repetition maximum (1rm) from submaximal tests seems to be good (in general, correlation coefficients > 0.90), although studies have mostly failed to cross-validate prediction equations. thus, care should be taken especially in relation to specificity of the population, of the exercise, and performance technique when developing and applying these equations. anthropometric variables have not proven to be good predictors of 1rm. the number of repetitions for a given % of 1rm is different for different exercises, so is the load for a given number of repetitions maximum (nrm) when performed at different velocities. exercise prescription based, indifferently, on number of repetitions or %1rm should be carefully considered.