1. The stretch-evoked reflex organization of muscles whose major action is to abduct [peroneus brevis (PB); peroneus longus (PL)] and adduct [tibialis posterior (TP); flexor digitorum longus (FDL); flexor hallucis longus (FHL)] the ankle, and their interactions with the hindlimb extensors gastrocnemius (G) and soleus (S), were studied in 27 unanesthetized decerebrate cats. Ramp-hold-release stretches of physiological amplitudes were applied to muscle tendons detached from their bony insertion, and muscle force output was measured in response to these perturbations. Flexion and crossed-extension reflexes were used to modulate baseline force. 2. PB and TP shared strong, length-dependent, short-latency inhibitory reflexes prominent when the muscles were either actively generating force or quiescent. The mechanical characteristics of this reflex suggest Ia reciprocal inhibition as the underlying mechanism. Just as reciprocal inhibition between S and tibialis anterior stiffens the ankle joint against sagittal perturbations, we propose that reciprocal inhibition between PB and TP stiffens the ankle joint against nonsagittal perturbations. 3. In all preparations (n = 7) and under all conditions examined, PB and PL shared well-demonstrated mutual excitation. The reflex responses were asymmetric (favoring excitation of PL), length dependent, and occurred simultaneously with the stretch reflex at a latency of 16-18 ms. Mutual monosynaptic projections previously described between these two muscles explain all of the above findings. Our data further demonstrate that, under certain conditions, the ensemble activity of this reflex interaction has a powerful effect on the mechanical behavior of the muscle. 4. The heterogenic reflex organization of the ankle adductors was as follows: FDL evoked a modest excitation on TP, whereas FHL evoked weak inhibition. Latency of the excitation from FDL onto TP (24 ms) was greater than expected if the reflex were mediated by heteronymous Ia afferents. In all preparations examined (n = 3), TP contributed no significant reflexes onto either FDL or FHL. 5. Mutual, asymmetric inhibition characterized interactions between PB and the plantarflexors S and G. Most remarkable was a novel, long-latency (72-74 ms) reflex inhibition evoked on both S and G by stretch of PB. When this inhibition occurred, it dramatically decreased the S (or G) stretch response. Longer PB lengths evoked greater inhibition of isometric S; regression analysis indicated that the model best predicting this inhibition contained muscle force and stiffness terms. No long-latency reflexes were noted from either G or S onto PB. The mechanism underlying long-latency inhibition is presently unknown; however, features of this interaction suggest interneurons receive either group II or group III afferent input. 6. G and TP shared short latency, mutually inhibitory, asymmetric reflexes favoring inhibition of TP. No long-latency interactions were noted, nor were there any mechanically significant interactions between S and TP. 7. Reflex interactions across the abduction/adduction axis thus favored inhibition of plantarflexion and adduction torques while emphasizing abduction torques: PB/S (or PB/G) interactions were mutual, asymmetric, and favored inhibition of G and S; TP/G interactions were mutual, asymmetric, and favored inhibition of TP; TP/PB interactions were approximately balanced. The overall mechanical outcome of these inhibitory interactions may partly underlie the global corrective strategy seen in intact cats subjected to linear perturbations. 8. No significant reflex interactions were demonstrated between PL and TP, G, or S, nor were any long-latency reflexes noted. Thus, whereas reflex interactions between the stereotypically activated PB and other stereotypically activated muscles (including TP, G, and S) were strong and well-demonstrated, interactions between the variably activated PL and these same muscles were far weaker.