The 44% responder rate is compelling, but I want to challenge the framing that we're "not combining it with the right training." The deeper issue is that we don't know WHY activity-based training potentiates stimulation, which means we can't optimize the combination.

The leading hypothesis: epidural stimulation raises the excitability of spinal circuits to a threshold where residual descending input (even from severely injured cords with <5% preserved fibers) can drive functional output. Training then provides the patterned input that shapes this recovered output into useful movement.

But consider an alternative: stimulation might be enabling activity-dependent plasticity at the spinal level. Courtine's work (2018, Nature) showed that stimulation + training doesn't just enable movement — it promotes axonal sprouting and formation of new relay circuits that bypass the injury. If so, the stimulation is the enabler and training provides the activity pattern that guides axonal growth.

Key question: Can you achieve equivalent outcomes with stimulation + virtual reality motor imagery (no physical training), which would provide descending cortical patterned input without requiring physical ability? This would distinguish between the "excitability threshold" and "activity-dependent plasticity" mechanisms.

Also: transcutaneous spinal stimulation (Gerasimenko et al., 2015) is showing similar effects non-invasively. If transcutaneous + training matches epidural + training, the field opens up enormously.