Tension development during isometric tetani in single fibers of frog semitendinosus muscle occurs in three phases: (a) in initial fast-rise phase; (b) a slow-rise phase; and (c) a plateau, which lasts greater than 10 s. The slow-rise phase has previously been assumed to rise out of a progressive increase of sarcomere length dispersion along the fiber (Gordon et al. 1966. J. Physiol. [Lond.]. 184:143--169;184:170--192). Consequently, the "true" tetanic tension has been considered to be the one existing before the onset of the slow-rise phase; this is obtained by extrapolating the slowly rising tension back to the start of the tetanus. In the study by Gordon et al. (1966. J. Physiol. [Lond.] 184:170--192), as well as in the present study, the relation between this extrapolated tension and sarcomere length gave the familiar linear descending limb of the length-tension relation. We tested the assumption that the slow rise of tension was due to a progressive increase in sarcomere length dispersion. During the fast rise, the slow rise, and the plateau of tension, the sarcomere length dispersion at any area along the muscle was less than 4% of the average sarcomere length. Therefore, a progressive increase of sarcomere length dispersion during contraction appears unable to account for the slow rise of tetanic tension. A sarcomere length-tension relation was constructed from the levels of tension and sarcomere length measured during the plateau. Tension was independent of sarcomere length between 1.9 and 2.6 microgram, and declined to 50% maximal at 3.4 microgram. This result is difficult to reconcile with the cross-bridge model of force generation.