The mathematical prediction of muscle and joint force requires a quantitative knowledge of muscle origins and insertions. A model is presented based upon marking the origins and insertions in three cadavers (six limbs). Right-to-left biological variations and/or making errors are sometimes significant, but they rarely result in moment arm calculation variations of greater than 20 percent and usually the variations are less than 10 percent. The muscle origin and insertion differences between small and large cadavers is great, as would be expected, and the use of single specimen or average data will result in large errors in muscle force predictions. A scaling scheme is presented which substantially reduces those errors. The inherent limitations of developing a straight line muscle model include: 1) right-to-left biological variations and/or marking errors; 2) difficulties in establishing "effective" origins or insertions when the locations of the actual origin or insertion do not accurately reflect muscle function; and 3) intersubject variability which cannot be accounted for by simple scaling schemes.