Abstract The Technicon AutoAnalyzer dye binding system for protein determination in milk was compared with the Kjeldahl and Udy methods. Individual cow and mixed herd milk samples were analyzed, including such variables as the effect of aging, chemical preservation, laboratory manipulation of protein content, and experimental changes in instrumental parameters. The Udy and Kjeldahl methods showed better agreement than the Technicon procedure in all treatments. Aging resulted in slight decreases in apparent protein content by both dye binding methods. Variation due to treatment with formaldehyde and mercuric chloride was considerable for the Technicon system (+ 0.72 and +0.20% protein, respectively). Individual cow milks varied more than herd milks and Guernsey more than Holstein milk. Milk samples standardized for protein content showed a high correlation with the Kjeldahl results. Replication by the Technicon system was superior to both the Udy and Kjeldahl methods. Performance of the Technicon system improved with increased dialyzing time, decreased dye concentration, and decreased sampling rate. Instrument performance was also influenced by dialysis membrance replacement.