Children's acquisition of tool use abilities is an important part of development but is not yet well understood. This study compares two modes of tool-use learning, observation and individual haptic experience. Two- and 3-year-olds had haptic experience with tools, observed tool use by others, had both haptic and observational experience, or no tool exposure. Their tool choice and use were evaluated across six problem-solving tasks that varied in degree of difficulty. Children learned about tools better by observation than by individual learning through manual exploration. Performance also varied by task difficulty, with more complex tasks proving more difficult. Findings are discussed from cognitive and evolutionary perspectives. (Contains 2 figures.)