How to improve clinical practice and, in particular, that of physical therapy? Currently, several strategies are used which all fit the label 'scientification'. These scientific strategies have to make physical therapy's clinical practice more homogeneous. Sometimes this homogenization is thought to be necessary for other strategies of innovation including effectiveness research. But it has also been suggested that more homogeneity in the clinic is already itself an improvement. In this article we comment on these strategies. More specifically, we direct our attention at attempts in physical therapy to establish a uniform, generally adopted scientific language. We argue that these attempts fail to appreciate the specificity of therapeutic work. The effectiveness research that follows is therefore liable to take irrelevant variables into account. We illustrate this argument with examples taken from analyses of diverging therapies for stroke patients.