The emphasis given to experimental problem-solving skills in science curriculum innovation has not been matched by the development of comparable assessment tools. Multiple-choice tests were constructed for seven skills using learning hierarchies based on expert-novice differences. The instruments were refined in three phases of field testing. The reliabilities of the tests are sufficient for making judgments of group performance, but are insufficient in a single administration for individual assessment. Evidence of the validity of the tests is presented and their worth is discussed within the framework of a theory of instruction.