Milk yields and milk component percentages of dairy goats are collected through Dairy Herd Improvement (DHI) with nearly the same procedures that are used for dairy cattle. This production information, which is forwarded by six dairy records processing centers to the Animal Improvement Programs Laboratory (AIPL) of USDA's Agricultural Research Service, is the basis of an annual calculation of estimates of genetic merit for yield traits. The American Dairy Goat Association (ADGA) conducts a type appraisal program in which does receive linear scores for their physical characteristics and a final score. This appraisal information also is the basis of an annual calculation of estimates of genetic merit. The evaluations, which are calculated by AIPL, are used in selection programs to increase milk and component yields and to improve animal conformation. The evaluations are available on the AIPL web site (http://aipl.arsusda.gov) and also are included in ADGA publications. During 1999, 11,273 does were enrolled in DHI test plans that have records that can be used in genetic evaluations; 3784 does participated in linear appraisal programs, which was lower than the average of 4285 does for the last 5 years. Annual genetic progress for does that were born during 1996 as a percentage of average breed yield was lowest for Toggenburgs (!0.1 for milk; 0.0 for fat and protein) and highest for Saanens (0.9 for milk and protein; 1.0 for fat). Corresponding trend for type traits across breeds was 0.67 for stature, 0.37 for rump angle, 0.34 for teat placement, 0.22 for suspensory ligament, 0.20 for strength, 0.16 for teat diameter 0.12 for rump width and fore udder width,, 0.09 for rear legs, 0.06 for dairyness, 0.05 for final score, 0.02 for fore udder attachment, and 0.01 for udder depth. To achieve higher rates of improvement, more extensive application of artificial insemination (AI) is needed so that greater use can be made of bucks of superior genetic merit for desired traits. However, the economics of dairy goat milk production would probably have to change for such an increase in AI usage to occur.