The present study examines different lacteal matrices (varying in fat content) in order to develop a protocol for extraction and quantification of fat-soluble vitamins. The samples (whole milk, both liquid and powdered; milk with reduced fat content; chocolate milk, and soy-modified milk) were evaluated using various protocols. Organic phase extraction alone succeeded in isolating no more than 20 to 40% of esterified and non-esterified forms of vitamins A and E. A combination of overnight saponification isolated more than 100% of the vitamin content indicated by manufacturers' labels. All vitamin compounds were separated by reversed-phase HPLC and detected at 265 nm, or alternatively, at the maximum λ of each individual form. The best mobile phase consisted of methanol: water 99: 1 (v/v) used in all the samples. The following experimental factors were optimized: saponification-KOH concentration, aqueous and alcoholic concentrations, duration and pyrogalol concentration; extraction of individual non-polar solvents, and solvent combinations (hexane, diethylether, and light petroleum). In order to confirm the effectiveness of these methods, matrices free of fats and vitamins (zero matrices) were spiked with known quantities of all-trans-retinol, vitamins D2, D3, and α-tocopherol. Other matrices were also spiked. Recoveries were followed by saponification with KOH methanol (1.9 N) at 30, 21, 42, and 28%. These improved to 104, 42, 85, and 96%, respectively, using KOH methanol (3.8 N). Similar conditions used diethylether: light petroleum 20: 80 (v/v) as an extraction solvent. The time required for exhaustive saponification was 3 h, since the pyrogallol content was increased from 0.1 to 1% w/v. The new extraction protocols show advantages relative to established methods for vitamins A, D, and E. This protocol could also be used for other food matrices.