In spastic paresis, stretch applied to the antagonist increases its inappropriate recruitment during agonist command (spastic co-contraction). It is unknown whether antagonist stretch: (1) also affects agonist recruitment; (2) alters effort perception. We quantified voluntary activation of ankle dorsiflexors, effort perception, and plantar flexor co-contraction during graded dorsiflexion efforts at two gastrocnemius lengths. Eighteen healthy (age 41 ± 13) and 18 hemiparetic (age 54 ± 12) subjects performed light, medium and maximal isometric dorsiflexion efforts with the knee flexed or extended. We determined dorsiflexor torque, Root Mean Square EMG and Agonist Recruitment/Co-contraction Indices (ARI/CCI) from the 500 ms peak voluntary agonist recruitment in a 5-s maximal isometric effort in tibialis anterior, soleus and medial gastrocnemius. Subjects retrospectively reported effort perception on a 10-point visual analog scale. During gastrocnemius stretch in hemiparetic subjects, we observed: (1) a 25 ± 7 % reduction of tibialis anterior voluntary activation (maximum reduction 98 %; knee extended vs knee flexed; p = 0.007, ANOVA); (2) an increase in dorsiflexion effort perception (p = 0.03, ANCOVA). Such changes did not occur in healthy subjects. Effort perception depended on tibialis anterior recruitment only (βARI(TA) = 0.61, p < 0.01) in healthy subjects (not on gastrocnemius medialis co-contraction) while it depended on both tibialis anterior agonist recruitment (βARI(TA) = 0.41, p < 0.001) and gastrocnemius medialis co-contraction (βCCI(MG) = 0.43, p < 0.001) in hemiparetic subjects. In hemiparesis, voluntary ability to recruit agonist motoneurones is impaired--sometimes abolished--by antagonist stretch, a phenomenon defined here as stretch-sensitive paresis. In addition, spastic co-contraction increases effort perception, an additional incentive to evaluate and treat this phenomenon.