G-protein-coupled receptors (GPCRs) play a crucial role in mediating e!ects of extracellular messengers in a wide variety of biological systems, comprising the largest gene superfamily at least in mammals. Mammalian GPCRs are broadly classi"ed into three families based on pharmacological properties and sequence similarities. These sequence similarities are largely con"ned to the seven transmembrane domains, and much less in the extracellular and intracellular loops and terminals (LTs). Together with the fact that the LTs vary considerably in length and sequence, the LT length of GPCRs has not been studied systematically. Here we have applied a statistical analysis to the length of the LTs of a wide variety of mammalian GPCRs in order to examine the existence of any trends in molecular architecture among a known mammalian GPCR population. Tree diagrams constructed by cluster analyses, using eight length factors in a given GPCR, revealed possible length relations among GPCRs and de"ned at least three groups. Most samples in Group J ( joined) and Group M (minor) had an exceptionally long N-terminal and I3 loop, respectively; and other samples were considered as Group O (other/original). This length-based classi"cation largely coincided with the conventional sequence-and pharmacology-based classi"cation, suggesting that the LT length contains some biological information when analysed at the population level. Principle component analyses suggested the existence of inherent length di!erences between loops and terminals as well as between extracellular and intracellular LTs. Wilcoxon rank transformation tests unveiled statistically signi"cant di!erences between Group O and Group J, not only in the N-terminal and I3 loop, but also in the E3 loop. Correlation analyses identi"ed an E1-I2 length-correlation in Group O and Group J and an N-E3 length-correlation in Group J. Taken together, these results suggest a possible functional importance of LT length in the GPCR superfamily.