Actions can be represented as a hierarchy of goals and means, whereas the granularity of the top level depends on the action's complexity. So far, most evidence comes from imitation studies. We explored this issue by evaluating the detection of means and end changes in a recognition paradigm across ages and cultures. German and Chinese preschoolers (n = 53) and adults (n = 58) studied videos of tool-use actions differing in complexity and familiarity. In the test phase, participants judged identical, means-changed and end-changed actions as same or different. Changes of the actions' end were generally identified more accurately independent of age and culture. Means' changes, but not ends' changes, were detected more frequently for complex t...