In orthopteran insects, neural networks for joint control exhibit different characteristics due to behavioural specializations. We investigated whether these differences are generated purely by the neuronal networks, or whether characteristics of the muscles or joint architecture (musclejoint system) are also involved in these behavioural specializations. We compared the properties of the muscle system moving the femurtibia joint of the middle and hindleg of three species, Carausius morosus, Cuniculina impigra and Locusta migratoria. Four aspects were analysed for the tibial extensor muscle: (i) the frequency-dependence of motoneuronal activity in response to sinusoidal stimulation of the femoral chordotonal organ (fCO), (ii) the muscle structure, (iii) the innervation pattern of the muscle and (iv) the histochemical properties of the muscle fibres. These aspects were compared with the filter characteristics of the open-loop femurtibia control system and of the musclejoint system involved. Whereas in both phasmid species (Carausius morosus and Cuniculina impigra) the motoneuronal activity steadily increases with sinusoidal stimulation of the fCO in the frequency range 0.015 Hz, in Locusta migratoria there is a decrease in motoneuronal activity between 0.01 and 0.3 Hz. The muscle structure is basically similar in all three species, as the number of singly innervated muscle fibres (supplied by the fast extensor tibiae motor neurone, FETi) decreases from proximal to distal. The number of triply innervated fibres supplied by the FETi, the slow extensor tibiae (SETi) and the common inhibitor 1 (CI1) is maximal in the middle of the muscle, and the number of dually innervated fibres (supplied by SETi, CI1) increases from proximal to distal. Differences between the locust and the two phasmid species exist in the distal portion of the muscle. The phasmid extensor tibiae muscle contains a morphologically distinct bundle of muscle fibres, not present in the locust, which is mostly dually innervated and which is larger in Cuniculina impigra. Similar results were obtained for the histochemical characterisation of the muscle fibres as revealed from their staining for myofibrillar ATPase activity. The number of histochemically identified fast fibres decreased from proximal to distal, while the number of slow fibres increased. In Carausius morosus and Locusta migratoria, the percentage of slow fibres increased by up to 6070 % at the distal end, while this increase was to almost 100 % in Cuniculina impigra. Apparently, the larger this distal region and the higher the percentage of slow, dually innervated fibres in it, the lower is the upper corner frequency (the stimulus frequency at which the joint control system produces a movement with 70 % of its maximal response amplitude) of the musclejoint system. In summary, it appears that the upper corner frequency of the open-loop system in Locusta migratoria (<0.05 Hz) results at least in part from properties of the neuronal joint control network, but in Carausius morosus (0.51.0 Hz) and Cuniculina impigra (0.10.2 Hz) it results from the upper corner frequency of the musclejoint system.