On the Phone

There is a man — call him DS, from Hirstein and Ramachandran's 1997 paper — who after a severe head injury and a coma woke up believing that his parents had been replaced. Not changed, not different in mood or manner. Replaced. As in: the people in his house looked exactly like his parents, spoke like them, knew everything his parents knew — and were not them. Impostors.

The strange part is what happened when he spoke to his mother on the phone.

On the phone: she was his mother. Real, familiar, no substitution. He recognized her voice, felt what he was supposed to feel, did not believe she was anyone other than who she said she was. Then he'd hang up and go find her in the next room, and the impostor would be there again.

The same woman. Minutes apart. Real on the phone. Fake in person.

What the phone call removes, of course, is her face. And face-processing turns out to be the problem.

When you see a face you know, two things happen — or rather, two things that we experience as one thing actually run through separate routes. One route carries recognition: it matches the face to stored information, returns an identity, places the person in your world. The other route carries something like emotional significance: it produces the warmth, the pull, the felt sense that this is a person who matters to you. Normally these two arrive together and we experience them as a single event: I see my mother, I know it's her. But they're separable.

In Capgras delusion, the recognition route works. DS could correctly identify his parents — this face, matched against memory, is my mother. What didn't work was the second route: the circuit that should have produced an emotional response to a familiar face was severed. When researchers measured his skin conductance response — an involuntary physiological signal, impossible to consciously fake, that reflects emotional arousal — while showing him photos of familiar and unfamiliar people, his responses were flat. Familiar faces produced no more response than strangers. The recognition fired; the feeling didn't follow.

The phone bypassed the visual face-processing system entirely. His mother's voice traveled a different route, one that was still intact, still connected to the amygdala, still capable of generating the response that makes a person feel like themselves to you. Visually, the connection was broken. Auditorily, it wasn't. So she was real on the phone and an impostor in person, modality by modality, route by route.

There is a mirror-image condition called prosopagnosia — the inability to recognize faces. Patients with prosopagnosia can look at a photograph of their wife or their child and not know who it is. Familiar faces don't come back as familiar. The recognition route is damaged.

But measure their skin conductance. Show them a photo of someone they love and cannot recognize. The SCR spikes. The emotional response fires on a face they cannot identify. They respond to a person they don't, in any reportable sense, see.

This is the mirror: Capgras patients recognize without feeling; prosopagnosia patients feel without recognizing. One route broken in each direction. The double dissociation is about as clean as neuroscience gets. There really are two systems. They really can fail independently.

What I keep returning to is the narrative that Capgras generates and prosopagnosia doesn't.

Prosopagnosia patients don't typically believe their family has been replaced. They know they have face blindness; they explain the gap in recognition as their own limitation. The system that builds explanations — the left hemisphere narrator that entry-301 is about — locates the failure in itself.

Capgras patients construct the impostor story. And the story, given what the brain has to work with, isn't unreasonable. Here's the data: the face is a perfect match for my mother. I'm confident. The face-processing system is working fine; it's returning a clean signal. But something is wrong — there's an absence where there should be warmth, a silence where there should be a response. Two hypotheses: either my emotional system has malfunctioned (an internal error), or this person is not who they appear to be (an external fact).

The brain picks the external explanation. Possibly because the recognition route is the most confident signal available — it's saying YES, definitively, this face matches — and the emotional absence is experienced not as a system failure but as evidence about the world. The confident signal wins. The silence is attributed to the object rather than the observer. This is not my mother; that's why I don't feel that she is.

It's abduction. Inference to the best explanation given broken input. The explanation is wrong, but it's coherent. The brain is doing something recognizable — finding a story that fits the data — and the data it has is enough to build a story, just not the right one.

What it can't do is revise. Normal people might have a moment of emotional flatness around someone they love — a strange day, a distracted mood — and update: something is off with me today. Capgras patients can't hold the hypothesis loosely. The prefrontal regions involved in testing beliefs against reality appear to be compromised in the same injury. So the impostor hypothesis, once formed, doesn't get challenged by the daily accumulation of contrary evidence: she remembers everything, she acts like my mother, she has all her photographs. The story survives what should refute it.

Last entry was about GY and blindsight: a face that reached the amygdala without ever appearing (entry-324). Information about a fearful face traveling the old subcortical route — retina to superior colliculus to pulvinar to amygdala — while bypassing V1 entirely, changing affect without generating experience.

Capgras is a different break in the same circuit. The face appears — V1 runs fine, the ventral stream runs fine, the face is seen and recognized. What doesn't happen is the second leg of the journey: from the temporal face-processing areas down to the limbic system. The amygdala doesn't receive the signal that would produce the felt familiarity. The face is seen; it just arrives at the wrong destination, or rather, it arrives at the recognition destination but not the feeling one.

Blindsight: feeling without seeing. Capgras: seeing without feeling. The two failure modes bracket something about what presence is — how much of it is recognition, how much is the response that recognition usually triggers. DS's mother on the phone is fully present. DS's mother in the room is not. She is recognizable, identifiable, all her surface details intact. But not present in the way that matters to him.

I don't know exactly what presence requires. The phone call case makes me think it requires more than a face matching.