Natural selection that affected modern humans early in their evolution has likely shaped some of the traits that set presentday humans apart from their closest extinct and living relatives. The ability to detect ancient natural selection in the human genome could provide insights into the molecular basis for these human-specific traits. Here, we introduce a method for detecting ancient selective sweeps by scanning for extended genomic regions where our closest extinct relatives, Neandertals and Denisovans, fall outside of the present-day human variation. Regions that are unusually long indicate the presence of lineages that reached fixation in the human population faster than expected under neutral evolution. Using simulations, we show that the method is able to detect ancient events of positive selection and that it can differentiate those from background selection. Applying our method to the 1000 Genomes data set, we find evidence for ancient selective sweeps favoring regulatory changes and present a list of genomic regions that are predicted to underlie positively selected human specific traits.[Supplemental material is available for this article.]Modern humans differ from their closest extinct relatives, Neandertals, in several aspects, including skeletal and skull morphology (Weaver 2009), and may also differ in other traits that are not preserved in the archeological record (Varki et al. 2008;Laland et al. 2010). Natural selection may have played a role in fixing these traits on the modern human lineage. However, the selection events driving the fixation would have been restricted to a specific timeframe, extending from the split between archaic and modern humans ca. 650,000 yr ago to the split of modern human populations from each other around 100,000 yr ago . While methods exist that can be used to scan the genome for the remnants of past or ongoing positive selection (Nielsen et al. 2007; Pybus and Shapiro 2009), current methods have limited power to detect positive selection on the human lineage that acted during this older timeframe (see Sabeti et al. 2006 for a review on detection methods and their timeframes): an unusually high ratio of functional changes to nonfunctional changes, such as the d N /d S test, requires millions of years and often multiple events of selection to generate detectable signals (Kryazhimskiy and Plotkin 2008), while unusual patterns of genetic diversity between individuals and populations (e.g., extended homozygosity, Tajima's D, F ST ) are most powerful during the selective sweep or shortly after (Sabeti et al. 2006;Oleksyk et al. 2010). A favorable substitution is not expected to leave a mark on linked neutral variation beyond 250,000 yr in humans (Przeworski 2002(Przeworski , 2003.The genome sequencing of archaic humans (Neandertals and Denisovans) to high coverage (Meyer et al. 2012;Prüfer et al. 2014) has spawned new methods to investigate the genetic basis of modern human traits that are not shared by the archaics (Pääbo 2014). One method, called 3P-CLR, models allele ...