Background: Fatigue considerably affects rehabilitation and ergonomics. Many approaches to this complex phenomenon, ranging from physiological to psychological, have been used to obtain meaningful fatigue measurements. However, none of the methods in the literature measure fatigue directly. It is therefore of considerable interest to determine which indirect methods best represent the state. Method: Fatiguing contraction was measured at maximum voluntary contraction (MVC) and 40% MVC in the biceps brachii, quadriceps and erector spinae muscles (three separate experiments). The most-reported objective measurements (contraction force, median electromyographic frequency, heart rate, muscle bed oxygenation and muscle blood volume) and subjective measurements (visual analog score, body part discomfort rating and perceived exertion rate) were simultaneously recorded. The data from the three experiments underwent separate statistical analysis. Descriptive statistics, linear mixed effects (to examine force fatigue predictability) and trend analysis (with between and within-subject correlations) were calculated. Results: Univariate ANOVA on all objective variables showed that gender was a significant factor (p<0.001). All subjective and objective variables were significant (p<0.05-0.001) in predicting force fatigue. However, the percentage variability explained remained small. By combining variables, the variability explained increase to between 60.2% and 71.9%. Correlations between variables were small but significant (p<0.05-0.001). Among single variables, median electromyographic frequency was a slightly better predictor of MVC fatigue (p<0.001) and visual analog score for sub-MVC (p<0.001). Conclusion: It is desirable not to measure or predict fatigue based on one variable alone. Combining variables improves measurement and prediction. Single-variable indexing of localized muscle fatigue is problematic.