A developmental investigation of the ability to control variables was conducted. One or two logically similar physics problems were individually administered to 120 subjects age 12 to 14. The problems involved either familiar or unfamiliar variables. Each problem was presented in three different informational formats. Success on the problems in the various formats ranged between 5% and 95%. Both familiarity of the variables and format of the question influenced success. A qualitative change in ability to control variables between 12 and 16 was observed. Results are discussed in relation to Piagetian theory and to various information processing conceptualizations. (Author)