This study aimed to examine within-day and between-days intratester reliability of mechanomyography (MMG) in assessing muscle fatigue. An accelerometer was used to detect the MMG signal from rectus femoris. Thirty one healthy subjects (15 males) with no prior knee problems initially performed three maximum voluntary contractions (MVCs) using an ISOCOM dynamometer. After 10 min rest, subjects performed a fatiguing protocol in which they performed three isometric knee extensions at 75% MVC for 40 s. The fatiguing protocol was repeated on two other days, two to four days apart for between-days reliability. MMG activity was determined by overall root mean squared amplitude (RMS), mean power frequency (MPF) and median frequency (MF) during a 40 s contraction. RMS, MPF and MF linear regression slopes were also analysed. Intraclass Correlation Coefficients (ICC); ICC1,1 and ICC1,2 were used to assess within-day reliability and between-days reliability respectively. Standard error of measurement (SEM) and smallest detectable difference (SDD) described the within-subjects variability. MMG fatigue measures using linear regression slopes showed low reliability and large between-days error (ICC1,2 = 0.43-0.46; SDD = 306.0-324.8% for MPF and MF slopes respectively). Overall MPF and MF, on the other hand, were reliable with high ICCs and lower SDDs compared to linear slopes (ICC1,2 = 0.79-0.83; SDD = 21.9-22.8% for MPF and MF respectively). ICC1,2 for overall MMG RMS and linear RMS slopes were 0.81 and 0.66 respectively; however, the SDDs were high (56.4% and 268.8% respectively). The poor between-days reliability found in this study suggests caution in using MMG RMS, MPF and MF and their corresponding slopes in assessing muscle fatigue. © 2008 Elsevier Ltd. All rights reserved.