We present ShadowGuides, a system for in-situ learning of multi-touch and whole-hand gestures on interactive surfaces. ShadowGuides provides on-demand assistance to the user by combining visualizations of the user’s current hand posture as interpreted by the system (feedback) and available postures and completion paths necessary to finish the gesture (feedforward). Our experiment compared participants learning gestures with ShadowGuides to those learning with video-based instruction. We found that participants learning with ShadowGuides remembered more gestures and expressed significantly higher preference for the help system.