A variable proportion of cardiac resynchronization therapy (CRT) patients do not benefit from treatment (termed `non-responders'). The problem of non-response to CRT might become increasingly important, because it is anticipated that larger groups of heart failure patients are indicated to the therapy. This article will discuss the definition of response to CRT, the parameters related to response to CRT, and finally whether response to CRT might be predicted. The effort to improve patient selection in order to maximize human and financial resource utilization has fallen short so far. It is, however, conceivable that rather than the identification of a `universally' applicable cut-off value, risk strata-in which inclusion of method for determination of left ventricular volumes, etiology, QRS duration and morphology, etc-might better serve the goal of defining non-responders. Potentially simple clinical scores might help in refining outcome and by doing so, allow to more precisely measure response to CRT in daily practice in the individual patient at the time of CRT implantation. Although sophisticated cardiac imaging modalities have been intensively utilized for improving patient outcome, it seems that many mechanical dyssynchrony measures suffer from technical limitations and from difficult interpretation of the complex signals, which lack reproducibility outside highly specialized laboratories.