It would also allow us to evaluate consciousness not as an oversimplistic dichotomy, but as a continuous spectrum, based on the complexity of distinctions that can be decoded from the patient's brain activity.

Neural decoding for covert consciousness is, in a sense, simpler than what's been done in healthy individuals. We don't need to reconstruct entire narratives (yet). We only need a paradigm to determine if patients can make basic semantic distinctions -- i.e., extract meaning.

What if we used mind-reading to determine what minds were there in the first place? Neural decoding could solve a lot of problems. Unlike behavioral or traditional covert consciousness paradigms, it wouldn't require interaction or assumptions of functional neuroanatomy.

The narratives didn't need to be auditory. The algorithms could even decode narratives from muted movies. This is as close to mind-reading as we've ever come.

Neural decoding works like this: You train an algorithm to associate stimuli with patterns of brain activity, then use it to "decode" a person's experience from their brain activity. The stimulus can even be a semantic concept, so you can decode not only sensation, but *meaning*.

Conversely, what if there is distortion of the brain due to a lesion, and activity occurs outside of typical motor regions? Should we not count this (and risk overlooking consciousness patients)?

Many fMRI paradigms work like this: During a scan, we present alternating periods of commands (e.g., "move your hand") and rest. If there is activity in motor regions during commands (relative to rest), we conclude covert consciousness. But should we? Does slight overlap count?