The paper presents a statistical test for ancient admixture between closely related populations, called the D statistic. This test exploits the asymmetry in the frequencies of two nonconcordant gene trees in a three-population tree. The D statistic is sensitive to asymmetry under various demographic scenarios and can detect archaic admixture even when no archaic sample is available. The test is insensitive to some demographic assumptions, such as ancestral population sizes, and requires only the assumption that ancestral populations were randomly mating. The D statistic can be used to estimate the proportion of archaic ancestry in present-day populations and to detect recent admixture. The paper also investigates the effect of sequencing error on the false-positive rate of the test and shows how to estimate the proportion of archaic ancestry. It also explores a model of subdivision in ancestral populations that can result in D statistics indicating recent admixture. The paper concludes that the D statistic is robust to sequencing error and ascertainment bias, and that it can be used to detect ancient admixture between closely related populations. The test is applied to Neandertal data and shows that Neandertals share more genetic variants with non-Africans than with Africans.The paper presents a statistical test for ancient admixture between closely related populations, called the D statistic. This test exploits the asymmetry in the frequencies of two nonconcordant gene trees in a three-population tree. The D statistic is sensitive to asymmetry under various demographic scenarios and can detect archaic admixture even when no archaic sample is available. The test is insensitive to some demographic assumptions, such as ancestral population sizes, and requires only the assumption that ancestral populations were randomly mating. The D statistic can be used to estimate the proportion of archaic ancestry in present-day populations and to detect recent admixture. The paper also investigates the effect of sequencing error on the false-positive rate of the test and shows how to estimate the proportion of archaic ancestry. It also explores a model of subdivision in ancestral populations that can result in D statistics indicating recent admixture. The paper concludes that the D statistic is robust to sequencing error and ascertainment bias, and that it can be used to detect ancient admixture between closely related populations. The test is applied to Neandertal data and shows that Neandertals share more genetic variants with non-Africans than with Africans.