In mammals, X-chromosome inactivation (XCI) ensures equal expression of X-linked genes in XX and XY individuals by transcriptionally silencing one X-chromosome in female cells. In this review, we discuss an imprinted form of X-inactivation in which the paternal X (Xp) is preferentially silenced. Believed to be the ancestral mechanism of dosage compensation in mammals, imprinted X-inactivation can still be observed in modern-day marsupials and in the extraembryonic tissues of some eutherians such as the mouse. Recent experiments have addressed the nature of the gametic imprint and focused on the regulatory interaction between the noncoding RNA gene, Xist, and its antisense partner, Tsix. Our review of the literature has inspired an unconventional view of imprinted XCI in mice. First, the evidence strongly argues that imprinted XCI is inabsolute, so that a stochastic number of extraembryonic cells escape imprinting. Second, contrary to conventional thinking, we would like to consider the possibility that the paternal X might actually be transmitted to the zygote as a pre-inactivated chromosome. In this model, the gamete initiates and establishes imprinted XCI, while the zygote maintains the pre-established pattern of gametic inactivation. Finally, we hypothesize that the inabsolute nature of imprinting is caused by imperfect zygotic maintenance. We propose that the mouse represents a transitional stage in the evolution of random XCI from an absolutely imprinted mechanism.