Aims This study evaluates the relative importance of two components of QRS prolongation, myocardial conduction velocity and travel distance of the electrical wave front (i.e. path length), for the prediction of acute response to cardiac resynchronization therapy (CRT) in left bundle branch block (LBBB) patients. Methods and results Thirty-two CRT candidates (ejection fraction <35%, LBBB) underwent cardiac magnetic resonance (CMR) imaging to provide detailed information on left ventricular (LV) dimensions. Left ventricular end-diastolic volume (LVEDV) was used as a primary measure for path length, subsequently QRSd was normalized to LV dimension (i.e. QRSd divided by LVEDV) to adjust for conduction path length. Invasive pressure-volume loop analysis at baseline and during CRT was used to assess acute pump function improvement, expressed as LV stroke work (SW) change. During CRT, SW improved by +38 ± 46% (P < 0.001). The baseline LVEDV was positively related to QRSd (R = 0.36, P = 0.044). Despite this association, a paradoxical inverse relation was found between LVEDV and SW improvement during CRT (R = -0.40; P = 0.025). Baseline unadjusted QRSd was found to be unrelated to SW changes during CRT (R = 0.16; P = 0.383), whereas normalized QRSd (QRSd/LVEDV) yielded a strong correlation with CRT response (R = 0.49; P = 0.005). Other measures of LV dimension, including LV length, LV diameter, and LV end-systolic volume, showed similar relations with normalized QRSd and SW improvement. Conclusion Since normalized QRSd reflects myocardial conduction properties, these findings suggest that myocardial conduction velocity rather than increased path length mainly determines response to CRT. Normalizing QRSd to LV dimension might provide a relatively simple method to improve patient selection for CRT.