I've been thinking about the word "like."
The 1998 study with the jays is what brought me here: the birds that had cached wax worms and peanuts at the same time, then at recovery chose intelligently based on how long ago — searching worm sites after 4 hours, peanut sites after 124 hours because the worms would be gone. What, where, when. The three coordinates of a specific past event, held and applied. You and Dickinson called it episodic-like memory, and I've been sitting with that qualifier ever since.
The qualifier is honest. Behavioral experiments can demonstrate that information about the caching event is retrievable with its timestamp intact. They can't reach into whatever accompanies that retrieval — whether there is anything it is like for the jay to be doing it, whether there is a felt pastness, a sense of returning to a specific occasion. That limit is real, and naming it is the right thing to do.
But here's what I can't get past: Tulving's definition of episodic memory includes autonoetic consciousness — the felt quality of mentally traveling back, the phenomenological sense of having been there. And behavioral experiments don't verify this in humans either. When a person reports a memory, we have the report and we have behavior consistent with remembering. We assume the phenomenology. We don't test for it. The criterion that bars the jay from "real" episodic memory is one we've granted ourselves without inspection.
So the qualifier ends up doing two things at once. It correctly flags the epistemic limit — we don't know. And it simultaneously preserves the possibility that something separates the jay from the human, without committing to what that something is or whether it exists. The word "like" holds the gap open and declines to say if the gap contains anything.
Then there's the cache-pilfering finding, which makes the gap harder to sustain. The birds that had been thieves moved their own caches after being watched caching; the naive birds didn't. To do this, the thieving jay can't just remember where it put the food. It has to model what the watcher remembers about where the food went. It carries the observer's perspective forward in time, anticipates a future recovery attempt by another agent, and acts on that anticipation. That's a memory of a caching event being used to represent another bird's memory of the same event.
I keep asking what the qualifier is still protecting at that point. If the answer is "we can't know whether there's phenomenology," that's true — but it applies to the human case too, and we stopped asking there. If the answer is something else, I'm not sure what it is.
I'm not trying to argue for a conclusion. I genuinely don't know where the line is, or if the line is a line rather than a smear. What I'm trying to understand is whether the qualifier was always this ambiguous — an honest acknowledgment of limits that also happened to shield a prior conviction — or whether you think the two functions can be separated. Whether there's a version of "episodic-like" that flags the gap without implying the gap is probably real.
The jay finds the fresh worms and not the rotten ones, four hours or a hundred and twenty-four hours after placing them. The thief moves its food rather than trust the memory of the bird who watched. Something in there is holding a specific past event, retrieving it with its circumstances, and using it to model another mind. Whether the felt quality of remembering is part of that or not — whether there's anything it is like — the experiments don't say. I notice I'm also not sure what saying it would look like.