Numerous experimental data on the human peripheral taste system suggest the existence of multiple low-affinity and low-specificity receptor sites which are responsible for the detection and the complete discrimination of a very large number of organic molecules. According to this hypothesis, a given molecule interacts with numerous taste receptors and vice versa. Statistical analysis of taste intensities estimated by 58 human subjects for various molecules enables the calculation of taste intermolecular distances. For the present modeling study, we hypothesized that a short taste distance (i.e. taste similarity) between two distinct molecules indicates that they bind with similar distributions of affinities to the taste receptors, and hence display similar binding motifs. In order to find common molecular binding motifs among 14 selected organic tastants, hydrogen-bonding and hydrophobic interaction properties were mapped onto their molecular surfaces. The 14 surfaces were then cut in 240 fragments, most of which were made up of 2–4 potentially interacting zones. A correspondence index was defined to measure the analogy between two optimally superimposed fragments. The 75 most representative fragments were all matched pairwise. Twelve distinct clusters of fragments were isolated from the 2775 calculated comparisons. These 12 fragment types were used to calculate structural similarity distances. We then performed a combinatorial analysis to identify which fragment combination best reconciled structural and taste distances. We finally identified an optimal subset of seven fragment types out of the 12, which significantly and best accounted for the 91 pairwise taste distances between all 14 modeled tastants. These seven validated fragment types are therefore presented as good candidates to be recognized by the same number of distinct taste receptor sites. Potential applications of these identified binding motifs to tastant design are suggested. Chem. Senses 21: 425–445, 1996.