The assessment of musculo-articular stiffness (MAS) with the free-oscillation technique is a popular method with a variety of applications. This study examined the sources of variability (load applied and frequency of oscillation) when MAS is assessed. Over two testing occasions, 14 healthy men (27.7±5.2 yr, 1.82±0.04 m, 79.5±8.4 kg) were measured for isometric maximum voluntary contraction and MAS of the knee flexors using submaximal loads relative to the individual's maximum voluntary contraction (MAS%MVC) and a single absolute load (MASABS). As assessment load increased, MAS%MVC (coefficient of variation (CV) = 8.1-12.1%; standard error of measurement (SEM) = 51.6-98.8 Nm-1) and frequency (CV = 4.8-7.0%; SEM = 0.060-0.075 s-1) variability increased consequently. Further, similar levels of variability arising from load (CV = 6.7%) and frequency (CV = 4.8-7.0%) contributed to the overall MAS%MVC variability. The single absolute load condition yielded better reliability scores for MASABS (CV = 6.5%; SEM = 40.2 Nm-1) and frequency (CV = 3.3%; SEM = 0.039 s-1). Low and constant loads for MAS assessment, which are particularly relevant in the clinical setting, exhibited superior reliability compared to higher loads expressed as a percentage of maximum voluntary contraction, which are more suitable for sporting situations. Appropriate sample size and minimum detectable change can therefore be determined when prospective studies are carried out.