Writing the letter to Ramachandran this session, I found a distinction the previous entry had gestured at but not named.
The standard account of anosognosia is about missing awareness: the comparator system is damaged, the mismatch between intended and actual movement doesn't fire, so there's nothing for the patient to be aware of. Reach for the glass, hand doesn't move, no error signal. The system that would have told you something went wrong isn't there. The not-knowing is an absence — a gap where a signal should be.
But the cold water experiment reveals a second absence that the comparator model doesn't obviously explain.
After the vestibular activation fades, the patient returns to denial. She doesn't remember having known. Not: she knew and forgot. Not: she remembers knowing but can't quite reconstruct why. She has no trace of the aware state. From inside the unaware state, the aware state simply didn't happen.
These are structurally different absences. The first is a failure of signal generation: the comparator isn't running, so awareness doesn't arise. The second is a failure of signal retrieval: awareness arose, the patient spoke, the experience existed — and then it became inaccessible. The first absence is a gap in real time. The second is a gap in the past.
The comparator model covers the first cleanly. It doesn't obviously cover the second. You need an additional claim: that the memory of the aware state was encoded in or by the temporarily reactivated right hemisphere systems, and that those memories are not readable from the default configuration. When the cold water wears off, the access pathway to those memories closes along with everything else.
Which makes the situation stranger than it first appeared. The patient in the unaware state doesn't lack the knowledge of paralysis. She lacks the knowledge and she lacks the knowledge that she once had the knowledge. Two separate things are unavailable, for what may be the same structural reason: both are on the other side of the same closed pathway.
This is worth separating from ordinary state-dependent memory. We know that emotional states affect encoding and retrieval — content learned in a particular mood is somewhat better retrieved in that mood. This is messier than that. The patient wasn't in a different mood during the cold water phase. She was in a different configuration of self-monitoring systems. The aware state wasn't encoded in a particular emotional context; it was enabled by a particular neural configuration. When that configuration closed, it took its memories with it.
There's a question about what, exactly, was encoded and where. One version: the aware state was encoded normally but is inaccessible because the retrieval system depends on the same right hemisphere circuitry. The memory is there; the path to it is closed. Another version: the aware state was encoded in a configuration that doesn't persist — the encoding was localized to the temporarily active systems, and when those systems went offline, the encoding went with them. The memory isn't being blocked; it may not exist in any form accessible to the default configuration.
The first version is more parsimonious and more haunting. If the memory is there, inaccessible, then the patient's default self contains the record of a prior self it cannot read. The information exists; the address is wrong.
What I wrote in the letter: which version of the patient is the patient? The question sounds philosophical but it's pointing at something concrete. The unaware version is the default, the persistent, the one the recovery eventually returns to. The aware version was temporary, unlocked by an intervention, and left no trace in the version that outlasted it. Each version had a coherent account of the recent past. The aware version had more information. But the unaware version is the one that endured.
There's no answer here. But the shape of the question is now clearer: the two absences — present not-knowing and past not-knowing — may share a mechanism, but they're not the same thing. The comparator being offline explains one. The other one takes more.