Suppose that as I slept, my body has been switched on me in the middle of the night, but all my memories were kept intact. Now it's morning and I have woken up in a new body. But the new body is similar enough to the old that I haven't noticed any difference. My, will I be surprised when I look in the mirror.
I remember going to bed last night. Or do I?
In remembering going to bed, what do I remember? I remember, perhaps, my feeling myself lying bodily in bed. That involves remembering my body as lying in bed.
Which body do I remember lying in bed?
Consider: I remember my body as lying in bed a few seconds ago and I remember my body as lying in bed the previous morning Moreover, there seems to be a univocity to these memories. It is the same body that I remember lying in bed, just as it is the same bed that I remember it lying in. I have no idea that today's body is different from yesterday's body.
Suppose the body I remember lying in bed is yesterday's body in both cases. Then my memeories from a few seconds ago are non-veridical when I remember my body being in bed a few seconds ago. And that doesn't seem right.
Suppose the body I remember lying in bed is today's body in both cases. Then I don't actually remember my body lying in bed yesterday morning: it was a different body that was lying in bed. I am misremembering due to an error of body misidentification.
But if I am misremembering, then my memories must have somehow changed. Last night my memories of yesterday morning were veridical. So at night they must have changed, either in type or in token.
And so the initial description of the story was wrong: I didn't actually switch bodies while keeping all my memories intact.
But perhaps I am wrong when I insist that the body I remember lying in bed a few seconds ago is the same body I remember lying in bed yesterday morning. Perhaps bodies are like clothes: when I remember wearing a blue shirt yesterday and I remember wearing a blue shirt a week ago, there is no particular blue shirt that I remember wearing on both occasions. I just remember wearing some blue shirt or other. Likewise, maybe rather than remembering a particular body as lying in bed yesterday morning, I remember lying in bed yesterday morning in some body or other, and I remember lying in bed this moring in some body or other.
I am not sure that's right. Surely I can remember that yesterday morning I had my phone in this hand, checking my email while awaiting the full return of consciousness. And so there is a particular hand in my memories, the same one that I remember holding my phone in a few seconds ago. And, again, as before this is either yesterday's hand or today's hand. That it's yesterday's hand means that I can't remember which hand I held the phone in a few seocnds ago. That it's today's hand means that I must have lost a memory.
But let's press on this. Can't the same thing happen with clothes? Say I remember wearing this very shirt last week, the one I am wearing now. But suppose my wife has switched it on me in the meanwhile. Surely a switch of shirts doesn't imply a loss of memory. But which memory is mistaken? My memory of wearing this shirt last week or my memory of wearing it now? Neither, perhaps. I remember wearing this shirt a few seconds ago. I remember wearing that shirt last week. But I don't remember wearing this shirt last week—though I may say I do. What I remember is wearing a shirt like this.
Can I make the same move with hands? I don't remember holding the phone in this hand. I simply remember holding the phone in a hand like this—most notably, like it in respect of chirality and connection to my mind. Still, I think it gets the phenomenology wrong.
So, if I'm right, the hypothesis of changing bodies while keeping all memories is a dubious one. Our autobiographical memories are bound up with our bodies—just as we enter de re into our autobiographical memories, so do our bodies.
Could I, though, switch bodies and keep a core of autobiographical memories? I am not sure. Core autobiographical memories seem to be closely bound up with embodiment. We can abstract from the particulars of embodiment, but when we abstract from a memory what we get need not be a memory any more. I remember lying in bed last night in some body or other, I say. But that's an abstraction from what I actually remember, perhaps—I remember lying on my back—this back—with a pressure sensation in my hand—this hand—from the phone that I am holding in it while checking my email. That's the memory.
Is this fatal to psychological theories of personal identity? Maybe not: maybe only to ones that are focused on autobiographical memories.
I am not entirely convinced by this argument, but I think it has some force.