A study is presented into the effect of familiar and abstract task content on self-directed inductive learning. Two groups of learners performed a self-directed inductive learning task in which they conducted experiments to discover the model describing the relations between five independent variables and a dependent variable. One group performed a task of familiar content, the other performed an isomorphic abstract task. During learning, indicators of the inductive learning process were gathered by means of think-aloud protocols and asking standardized questions to the learners. As expected, learning outcome was higher in the group performing the concrete task. It appeared that the number of hypotheses, research plans and inferences stated were predictive for successful learning outcome, especially in the abstract task. The content effect was explained in terms of a narrowing down of the number of possible hypotheses to be considered in the concrete task, compared to the abstract task.