Sensory prosthesis research has advanced at a rapid pace, yet standard stimulation pulse shapes have remained largely unchanged. Most prostheses typically utilize symmetric pulses to deliver electrical stimulation to peripheral nerves. Recent neurophysiological studies, however, have shown that these symmetric pulses do not always reliably evoke firings in stimulated afferents, thus impeding optimal restoration of sensory inputs to the brain. One promising strategy is the use of alternative pulse waveforms---specifically, asymmetric pulses, characterized by a brief, high-amplitude cathodic phase followed by a long, low-amplitude anodic phase. Testing alternative waveforms for sensory prostheses can be challenging due to lack of simple, objective measures in animal models and the limited number of implanted patients. Here, we leveraged the unique properties of the vestibular system---the readily quantifiable reflexes that stabilize gaze and posture by generating eye and head movements---to investigate whether alternative pulse shapes are beneficial to prosthesis performance. We applied vestibular prosthetic stimulation using symmetric and asymmetric waveforms and quantified the evoked reflexes by measuring eye and head movements. Simultaneously, we recorded both evoked potentials and single-unit responses in the vestibular nuclei using high-density silicon probes. In comparison to symmetric pulses, asymmetric pulses evoke markedly stronger reflexive eye and head movements. Correspondingly, asymmetric pulses resulted in larger evoked potentials compared to standard pulses. Single-unit recordings revealed a strong correlation between neural recruitment and enhanced behavioral performance. A simple population model suggests asymmetric pulses are more effective in recruiting afferent fibers that are further away from stimulating electrodes. Taken together, our findings indicate that asymmetric pulses can increase vestibular prosthesis performance. We speculate that these beneficial effects will prove translatable to other sensory prostheses and, eventually, clinical practice. Critically, our single-unit recordings and population model results will be useful for validating and improving existing biophysical models of peripheral nerve stimulation.