Summary.1. The ten recognized subspecies ofTamandua tetradactyla(Linn.) are defined mainly by skull proportions. The differences between these subspecies have been analysed with the aid of statistical methods based on the allometry formula, in order to discover how far neglect of the fact that many proportions change with size has led to false conclusions.2. In order to provide a background for the statistical analysis, all available information on the characteristics and distribution of each race is first summarized, and controversial points discussed. The main points are as follows:–(a) Tetradactylaandlongicaudataare not clearly defined. The latter seems to be distinct only in fur pattern, lacking the usual striping.(b) Instabilisandquichuashow parallel characteristics with races of other mammals from the same localities, suggesting their peculiarities are due to ecological factors.(c) Onlylongicaudata, quichua,and possiblytetradactylaare distinguishable by their coat characters.(d)Tail‐lengths appear to fall into two size groups,chapadensisandinstabilishaving shorter tails than the rest. Tail‐length ininstabilisis uncorrelated with head and body‐length, suggesting that its short tail is an independent character, not due to its small body‐size.3. Allen's classification ofinstabilis, chapadensis, mexicana,andchiriquensisinto two groups, based on relative nasal length, is found to be erroneous.Instabilisappears to be a small edition ofmexicanaandchiriquensisin the skull characters studied, whilechapadensisdiffers from the others in having a greater extension of the frontals at the expense of the nasals.4. Study of other skull characters suggests that the status of most of the races is dubious. Skulls from nearly all regions show marked variability in proportions, and there is no clear tendency for proportions to change with geographical position.5. Comparison of ranges and means suggests that onlychapadensisandmexicanahave any claim to be considered distinct subspecies on the basis of skull proportions.Chiriquensis, punensis, tambensis,and specimens from the Middle Amazons cannot with certainty be distinguished from the British Guiana form,tetradactyla. Sellatais probably identical withmexicana.The latter's sole peculiarity seems to be its relatively narrow rostrum.6. It is clear that proportional differences between two groups of animals cannot be accepted as evidence of taxonomic distinctness, when size differences occur, without further analysis. The tests described have been found satisfactory for the purpose of analysing proportional differences.7. The danger is emphasized of basing taxonomic classifications on small samples. Before subspecies are defined it is necessary to collect large samples from different regions and to study the fluctuational variability. Among mammals this is usually large for all quantitative characters.8. Adequate data for the study of rare animals such asTamanduawill only accumulate if museum taxonomists regularly publish comprehensive measurements of the specimens in their care. It is important to make available the individual measurements, since otherwise statistical analysis is severely restricted.9. All original measurements will be deposited in the library of the British Museum (Natural History