Near infrared reflectance spectroscopy (NIRS) was used to study the reflectance properties of intact and minced lamb muscles in two presentations to the instrument to predict their chemical composition. A fetal of 306 muscles were examined from 51 lambs, consisting of the following muscles: longissimus dorsi, supraspinatus, infraspinatus, semimembranosus, semitendinosus and rectus femoris. Modified partial least squares (MPLS) regression models of chemical variables yielded R-2 and standard error of cross-validation (SECV) of 0.76 (SECV: 10.4), 0.83 (SECV: 5.5) and 0.73 (SECV: 4.7) for moisture, crude protein and intramuscular fat in the minced samples expressed as g/kg on a fresh-weight basis, respectively. Calibrations for intact samples had lower R2 and higher standard error of cross validation (SECV) compared with the minced samples.