Forty-two first-year psychology students took part in an exploratory study of conceptions of emotions using Kelly's (1955) Repertory Grid technique. Each participant generated eight different personal constructs for comparing and contrasting eight different emotional states, then rated these emotions on the basis of their own constructs. Analysis of the group data revealed five significant principal components. The first four of these components related to evaluation (positive vs. negative), direction (inner-directed vs. outer-directed), duration (long-term vs. short-term), and causation (internal vs. external cause). The fifth component was not so readily interpretable. The internal complexity of the components suggested that participants' implicit theories of emotions are relatively well articulated and do not depend on simple unitary dimensions of comparison. Potential advantages of the repertory grid technique in this kind of research are considered.