The mechanism of the active tension response of insect fibrillar muscle to step changes and small oscillations of length was re-investigated, following White's demonstration (1983) that the high relaxed stiffness evidently persists during activation and cannot be neglected as had previously been assumed. White's result makes earlier explanations of the small-signal response untenable; the experimental and theoretical studies described here lead to a new class of explanations at the cross-bridge level. The response of an activated muscle to a fast stretch consists of a synchronous tension increase that is followed first by a rapid decay of tension and then by a delayed rise ('stretch activation'). It was shown in glycerinated fibre preparations from the water bug and the bumblebee that subtraction of the relaxed tension response from the active response results in a prominent undershoot of the tension level preceding the step, before the delayed rise of tension. The responses of the same fibres to sinusoidal oscillations, in the frequency range 1-150 Hz, showed an equivalent behaviour, with the active locus circling the relaxed locus in a Nyquist plot, as described by Machin & Pringle (1960). Stiffness was determined during the tension response to a large step (of 1%) by recording the immediate change of tension to a small test step (0.2%), applied at various times after the conditioning step. In the majority of preparations stiffness remained constant or increased during the undershoot of tension. Step and sinusoidal responses with the above features cannot be explained at all by an active component resembling a simple exponential delay. We show, however, that such features are predicted if certain small-signal effects of cross-bridge distortion (previously and erroneously assumed negligible in insect flight muscle for the small-signal case) are incorporated in models of the cross-bridge cycle. Two alternative hypotheses for the effects of distortion are examined: (i) distortion-induced detachments and (ii) distortion-modulated transitions among multiple attached states (Huxley & Simmons, 1971). For the first we also show that the results do not differ qualitatively whether one assumes strain, interfilament displacement or 'bridge recruitment' as the physical correlate of stretch activation. Both of the above explanations account, at least qualitatively, for the observed rapid decay and undershoot of tension following a step increase of length, and for the circling of the active Nyquist-plot loci about the passive locus. The explanation based on distortion-induced detachments, however, appears to be inconsistent with the stiffness measurements.(ABSTRACT TRUNCATED AT 400 WORDS)