Prediction of body composition was assessed in thirty-eight female and thirty-seven male random source dogs using in vivo methods of deuterium oxide dilution, subjective and objective morphometry, bioelectrical impedance and ultrasound, either separately or in various combinations. Carcass composition determined by chemical analyses of carcass homogenates served as criterion measures of body composition. Dogs were selected based on gender, body weight and body condition score. Body weight ranged from 7.3 to 34.5 kilograms (kg), subdivided into 4.5 kg increments with 6 female and 6 male dogs per increment. Body condition was categorized as thin, optimum or obese using a defined criteria, body condition scoring system (subjective morphometry) with 12 female and 12 male dogs per body condition category. Selection criteria produced 18 body weight condition groups with 2 female and 2 male dogs per group. One additional male and 2 female dogs were included for economic and ethical reasons. Equations to predict carcass composition from in vivo measurements were derived using standard regression techniques. Influence diagnostics, residual analysis and data splitting were used to validate predictive equations. Predictions from deuterium oxide dilution produced the most precise estimates of body composition. Average standard errors of estimation (SEE) from deuterium equations were 1.3, 1.8, 1.0, and 0.4 percent for percentages of body moisture, fat, protein and ash, respectively, and 0.39, 0.57,0.21 and 0.08 kg for absolute quantities of moisture, fat, protein and ash, respectively. Morphometry produced the most imprecise, but economical, estimates. Average SEE from morphometry equations for proportions were 3.0, 4.0, 1.3, and 0.4 percent, and 0.9, 0.9, 0.3 and 0.07 kg for absolute quantities of moisture, fat, protein and ash, respectively. Subjective morphometry could estimate body fat with an average SEE of 3.4 percent and correctly categorized 75 percent of the dogs. Bioelectrical impedance and ultrasound produced predictions with average SEE intermediate to deuterium and morphometry. Bioelectrical impedance was equivalent to deuterium dilution on the basis of cost per unit improvement in SEE, but ultrasound was not cost effective. / Ph. D.