In this article results are reviewed from different experimental approaches to determine the size of the power stroke generated by myosin molecules during their ATPase cycle. While data from fiber studies and protein crystallography predict a stroke size of about 10 nm for skeletal muscle myosins, single molecule studies imply a stroke size for these myosins of only about 5 nm. Single molecule studies also showed the stroke size to be proportional to the length of the light chain binding domain, acting like a lever arm. At the same lever arm length, however, the stroke size of smooth muscle myosin II is found about twice as large and a stroke size of about 14 nm was reported for class-I myosins. It was proposed that such different stroke sizes for molecules with same lever arm length result from different extend of converter domain rotation. Only for class-I myosins, however, an about 30 degrees larger rotation of the converter was found so far by protein crystallography. This, however, is far too small to account for the almost 3-fold larger stroke size reported from single molecule studies. In this contribution we discuss some factors that might account for the apparent discrepancies between single molecule studies on the one hand and protein crystallography as well as some fiber studies on the other hand. In addition, we present some modeling to illustrate that the power stroke very likely is underestimated to a large extent in current single molecule approaches. We further show that differences in the stroke size for various classes of myosins reported from single molecule studies might be related to small differences in the probability to execute the power stroke kinetics. We demonstrate that such small changes in power stroke kinetics can seriously affect the extent to which the 'true' power stroke is underestimated by present single molecule approaches.