The objective of this study was to identify the most important clinical variables determining gait speed in persons with stroke. Sixteen chronic stroke subjects (mean age, 47.9 (±15.6) yr; mean time post-stroke, 43.9 (±36.5) mo) able to walk independently without a brace participated in the study. The impairments in motor function, sensation of the paretic lower limb, and balance were evaluated with the Fugl-Meyer Assessment. A spasticity index was used to assess the muscle tone of the plantarflexors. The maximal strengths in plantarflexion and hip flexion were measured with a Biodex dynamometric system. Cinematography and foot-contact data collected on the paretic side were used to determine the comfortable and maximal gait speeds. The level of association between gait speeds and the clinical variables were first examined with Pearson's correlation coefficients and, then, with multiple linear regression analyses using the stepwise method. Results revealed that the motor function of the lower limb, balance, and hip flexion strength were significantly related to comfortable and maximal gait speeds (0.5 <r< 0.88;P< 0.05). For the comfortable gait speed, the regression analysis selected only the hip flexor strength as a significant variable (R2= 0.69). For maximal gait speed, the variables retained were hip flexor strength, sensation at the lower limb, and plantarflexor strength (R2= 0.85). The present results suggest that strength and sensation at the lower limb are important factors to consider in determining the gait capacity of chronic stroke subjects.