Our session in Dey Hall yielded a very nice diagram.
The income effect is greater for people far away from the pivot point than it is for somebody near the pivot point. If most people are in the latter group, then substitution effects dominate and demand curves do slope downward.