6 probleas dn comparcius populations sainpled differently. the comparison of variaat ereguencics Tu addition, from different laboratories, and expos ially sverige Evaienciss when the mix of systems studied miy Vary widely, is to be approached with caution, For the purposes ot this ‘oreliminary comparison, we will content ourselves with reference to two series which originate in this Jaboratory and involve essentially the same proteins and tachniques as the original material herein reported, plus a third composite series for Caucasians. We will again use an unweighted average, based on as many of the systems listed in Table 2 as ara covered in the references cited, By this convention, the variant trequency for South American Indians, based on systems-1, 2, 8-10, Li-16, 18 and 20-24 of Table 2, is 1.7/1000 (2, and unpublished). lower frequency than in our last sublicationon this subject (2) The is duc toa the exclusion of the Yanomama~-2 albumin variant from this calculation because its frequancy in the total sample now exceeds 2 percent (11). The frequency in West European Caucasians, based on systems 1, 2, 8-16, 18-24 and the data of Fleischer and Mohr (27); Moullec et al (28); Sick et al (29) 3 Uajatzadeh and walter (30); Fine (31); Rex-Kiss and Fesus (32) 5 and Harris, Hopkinson, and Robson (33) is 2.7/1000, and for Japanese adults, based on-systems 1, 2, 4, 8-10, 12, 13, 15, 16, 18, and 20-24, 1.9/1000 (34). As noted, there are obvious difficulties in statistical comparisons between these sertes. Our own Microuesian data, and we suspect to some extent that of others, contains closely related individuals, as does our data on Amerindians, MM By contrast, the data on Caucastaus and Japanese 509433 hat at