One purpose of classification systems is to group objects with similar properties so that meaningful generalizations can be made concerning their properties, behavior and potential use. It has been difficult to measure the success of such groupings. Similarly it has been difficult to determine how well soil classification systems group properties different from the criteria upon which the systems are based. Eighty soil samples (0–15 cm depth) of the Andept suborder were collected from the Island of Hawaii. Soil chemical properties (pH, exchangeable cations, extractable P, P sorption, and silicon) were measured. The 80 soils were grouped according to Soil Taxonomy, the proposed Andisol order of Soil Taxonomy, the FAO Legend for the Soil Map of the World, the French Classification System, and the 1938 Soil Classification System. Each system's effectiveness in grouping was compared by using regression techniques based on the statistical model of the classification scheme. Each system had a specific category which most effectively grouped soils with similar properties. The most effective category (that which decreased the pooled variance most) was the great group level of Soil Taxonomy, the suborder level of the proposed Andisol order of Soil Taxonomy, the Groupe level of the French System, and the suborder level of the 1938 System. In considering the FAO legend, only soil units were examined. These effectively separated variation in most soil properties in this group of soils. As indicated by the greatest precision (smallest pooled variance) Soil Taxonomy was effective in grouping all soil chemical properties except Na, whereas grouping 80 soils into 3 orders of the 1938 System did not increase homogeneity within groups. The Groupe category of the French System which formed the 2 taxa of saturé and desaturé was effective for most properties, especially soil pH. In our case the two Andisol suborders were identical within the Groupe level of the French System ( saturé and desaturé), whereas the four taxa of the Andept great group category corresponded with four great groups of the proposed Andisol order. Lower categories of Soil Taxonomy, the 1938 System, or the French System did little to increase precision, suggesting that differentiae of the lower highly defined categories studied here were correlated with those of more general higher categories.