A selection experiment on maximum duration of fertility of Brown Tsaiya ducks after artificial insemination (AI) with pooled Muscovy semen has been conducted since 1992. The Brown Tsaiya ducks were divided into two lines: a control line (T) with no selection and a selected line (S). The traits measured were the number of eggs set that were laid from Days 2 to 15 after one AI (NES), the number of fertile eggs at candling (NEF), the total number of dead embryos (NED), the maximum duration of fertility (MD), and the number of hatched mule ducklings (NEH). The selected trait was NEF. Six generations with a total of 2,127 females were measured. The variance components were estimated for each line in a multiple-trait animal model, using the restricted maximum likelihood (REML) methodology, which yields estimates free of bias caused by selection and inbreeding. Estimates of the heritability and genetic correlation from the two lines were very similar. Heritabilities in the S and T lines, respectively, were 0.14 and 0.10 for NES, 0.30 and 0.26 for NEF, 0.06 and 0.09 for NED, 0.28 and 0.21 for MD, and 0.18 and 0.19 for NEH. High and favorable genetic correlations existed between NEF and MD (0.96 and 0.92), between NEF and NEH (0.86 and 0.91), and between MD and NEH (0.90 and 0.82). The results suggested that selection for NEF could improve the maximum duration of fertility, but it could be useful to check the estimates of genetic parameters in a meat-type female duck.