1. The stability of sarcomere lengths along single frog twitch fibres was examined, during lengthening and shortening, using a spot follower appparatus to monitor or control the length of a central segment. 2. During active shortening from sarcomere lengths beyond 2.2 micrometer the end sarcomeres shortened dramatically, while much of the fibre did not shorten at all. It is proposed that this is the cause of the tension failing to recover, after the shortening ceased, to the value of isometric tension at the shorter length. 3. During active lenghtening from sarcomere lengths beyond 2.2 micrometer, non-uniformity of stretch was seen, with the middle stretching more than the ends. Some maintained extra tension after stretch above that appropriate to the longer length was found, as were consistent changes in internal movement, and in the shape of the tension record during relaxation. 4. Measurements of stiffness during and after a lengthening suggest that no increased activation is involved. Observation of internal movement during the raised tension after a lengthening contradicts theories involving 'locked on' bridges. 5. From these and other observations, an explanation for the extra tension in terms of non-uniformity of sarcomeres is proposed. The explanation is in accord with that previously suggested for the creep phase of tension rise seen at these lengths.