Educational research has thus far failed to greatly improve education because it is based on a model of reality that emphasizes unilateral control in both research and practice. Efforts to gain unilateral control presume that researchers or practitioners know what is significant from the outset, that they are essentially outside the situation they are observing or acting on, and that knowledge is to be used to implement a preconceived plan of action. Scientific research based on the unilateral control model seeks only to develop descriptive theories; when faced with the task of proposing alternatives to what it describes, such research fails because it does not reflect on its own assumptions. In contrast to the unilateral control model, which offers only a fundamentally anti-educational "reflective" science "about" action, a new model is needed on how to conduct an "action science." Such a science would consider the researcher to be an interactive participant, rather than a detached observer, in the situation under consideration and would welcome the possibility of change through dialog between the actor-researcher and others. Research, according to this new model of collaborative inquiry in the social sciences, would be regarded as an "experiment-in-practice," an opportunity for studying the self as well as others that would require a new politics and a new ethics. (JBM)