Based on quantitative electromyography, a muscle can be categorized as normal or affected by a neuromuscular disorder. The objective of this work was to compare the utility of probabilistic to conventional means and outlier methods of categorization of myopathic and normal muscles. Various sets of motor unit potential (MUP) features detected in biceps brachii muscles of control subjects and patients with facioscapulohumeral muscular dystrophy were used to categorize them as normal or myopathic based on conventional means and outlier categorization (CMC) as well as a new probabilistic muscle categorization (PMC). The sensitivity, specificity, and accuracy provided by each categorization method were compared. The categorizations made using PMC were significantly more accurate (by at least 10%) compared with CMC (P < 10(-10)) for muscles evaluated in this study. Area, duration, and thickness were highly discriminative MUP features.