During normal muscle shortening, the myosin heads must undergo many cycles of interaction with the actin filaments sliding past them. It is important to determine what range of configurations is found under these circumstances, and, in terms of the tilting lever arm model, what range of orientations the lever arms undergo. We have studied this using the X-ray interference technique described in the previous article, focusing mainly on the changes in the first order meridional reflection (M3) as compared to isometric. The change in ratio of the heights of the interference peaks indicates how far the mean lever arm angle has moved towards the end of the working stroke; the total intensity change depends on the angle change, on the number of heads now attached at any one time, and on the dispersion of lever arm angles. The latter provides a measure of the distance over which myosin heads remain attached to actin as they go through their working strokes. Surprisingly, the mean position of the attached heads moves only about 1 nm inwards (towards the center of the A-band) at low velocity shortening (around 0.9 T0): their dispersion changes very little. This shows that they must be detaching very early in the working stroke. However, at loads around 0.5 T0, the mean lever arm angle is about half way towards the end of the working stroke, and the dispersion of lever arm angles (with a uniform dispersion) is such as to distribute the heads throughout the whole of the working stroke. At higher velocities of shortening (at 0.3 T0), the mean position shifts further towards the end of the stroke, and the dispersion increases further. The details of the measurements, together with other data on muscle indicate that the force-generating mechanism within the myosin heads must have some unexpected properties.