Spatial training has been only modestly effective at improving the performance of adolescents and adults on the water-level task. Based on previous findings with the task, a self-discovery training procedure was developed that involved having participants proceed from easier to more difficult problems along a dimension of increasingly greater competing perceptual cues. The training was effective in (a) eliminating the gender differences on the drawing task, and (b) significantly improving females' knowledge of the physical (invariance) principle, although not to the level of males. Training effects did not transfer to a related spatial task. The water-level task (Piaget & Inhelder, 1956) requires participants to predict how the surface of water would look in a tilted container. Typically, research participants are asked to draw the water line in a twodimensional depiction of a tilted bottle or glass, in which case a horizontal line is the correct response. Although a surprisingly large proportion of both men and women have difficulty with this task, gender differences on it are nevertheless robust. Among adolescents and adults, many studies have reported that males are more likely than females to indicate, both graphically and verbally, that water remains horizontal regardless of the orientation of its container (Kalichman,