It has been reported that preimplantation human embryos secrete HLA-G, and the levels may be predictive of their ability to implant. However, it is not known which of the membrane-bound (HLA-G 1–4) and soluble (HLA-G 5–6) alternatively spliced forms are present, nor the developmental stage at which they appear. Therefore, we have investigated HLA-G mRNA isoform expression on single embryos at the two-, four-, six-, and eight-cell, morula, and blastocyst stages. The percentage of embryos expressing each HLA-G isoform mRNA increased with developmental stage, but contrary to expectation, HLA-G5 mRNA was not detected in single two- to eight-cell embryos and was only expressed by 20% of morulae and blastocysts. Similarly, soluble HLA-G6 mRNA was not detected until the blastocyst stage and then in only one-third of embryos. In contrast, labeling with MEM G/9 Ab (specific for HLA-G1 and -G5) was observed in 15 of 20 two- to eight-cell embryos and 5 of 5 blastocysts. This disparity between mRNA and protein may be due to HLA-G protein remaining from maternal oocyte stores produced before embryonic genome activation and brings into question the measurement of soluble HLA-G for clinical evaluation of embryo quality. Although HLA-G is expressed in the preimplantation embryo, later it is primarily expressed in the invasive trophoblast of the placenta rather than the fetus. Therefore, we have investigated whether down-regulation of HLA-G first occurs in the inner cell mass (precursor fetal cells) of the blastocyst and, in support of this concept, have shown the absence HLA-G1 and -G5 protein and mRNA.