Action video games have received intensive study because they improve players’ visual attention and perception abilities more than other types of video games. However, experimenters’ ability to attribute visual functioning improvements to particular action video game components has been hampered due to the use of non-action video games as controls. Here, we employ a newly developed, tightly controlled experimental paradigm in which two groups of participants trained in two versions of the same custom video game world. One version contained both friends and foes and the other contained foes only. A third group of participants served as a no-training control. People trained for two hours in friend vs. foe discrimination showed several modest improvements in their visual skills, including a significant increase in attentional filtering and a marginal decrease in flanker interference. This experimental design establishes that friend vs. foe discrimination per se led to the observed visual improvements.