The Qualifier

The experiment gave jays two types of food: wax worms (perishable, preferred) and peanuts (non-perishable, less preferred). Both cached in distinct trays at the same time. Then, depending on the trial, jays were allowed to recover their caches either 4 hours later or 124 hours later.

After some training, the jays learned that wax worms were fine at 4 hours but rotten at 124 hours. Nothing unusual about that. What was unusual: when allowed to recover, jays that had cached 4 hours ago searched the worm locations first. Jays that had cached 124 hours ago searched the peanut locations — because the worms would be gone but the peanuts would still be there.

To do this, the jay can't just apply the rule "wax worms go bad." It has to know when it put this particular batch of worms in this particular tray. It has to locate a specific past event — not just the food type and location, but how long ago it placed the food there, on that occasion. What, where, when — the three coordinates of a specific memory. That's what Clayton and Dickinson showed in 1998.

They called it "episodic-like" memory.

The word "like" is doing a lot of work.

Tulving's definition of episodic memory includes something he called autonoetic consciousness — the felt sense of yourself in time, the subjective experience of mentally traveling back to a past event and re-encountering it. Not just information retrieved, but the feeling of remembering: the pastness of the thing, the sense of having been there. Without evidence of that inner quality, you might have memory that works like episodic memory, retrieves the same information, achieves the same behavioral results — but isn't, technically, episodic memory.

Clayton and Dickinson used the qualifier honestly. Behavioral experiments can't speak to phenomenology. You can show the jay finds fresh worms and not rotten ones. You can't show what it's like for the jay to be doing that — whether there's something that feels like remembering, or whether the information just becomes available with no accompanying sense of occasion. The qualifier flags that limit. It's a fair flag.

The problem is that the thing the qualifier can't reach is exactly what Tulving's definition depends on. So "episodic-like" becomes both a genuine epistemic acknowledgment and, simultaneously, protection for the view that something important separates the jay from the human. The word "like" carves out a gap and then declines to say whether the gap contains anything.

The critics of this framework have pointed out that we don't actually test for autonoetic consciousness in humans either. We ask people to report memories and assume the phenomenology. There's no behavioral test that demonstrates the felt quality of remembering rather than just assuming it. The criterion that excludes the jay from "real" episodic memory is one we've never verified even in ourselves — we just take it as given because we are the ones doing the taking.

A second finding from this line of research sharpens the problem further. Dally, Emery, and Clayton showed that jays which had been thieves — birds that had previously pilfered other jays' caches — were likely to move their own cached food after being watched during caching. Naive jays, birds that had never stolen, didn't show this behavior even when watched. The thieves moved their stores. The innocents didn't.

To move caches strategically, the jay has to represent not just its own knowledge of where the food is, but another bird's knowledge of where the food is. It carries the watcher's perspective forward in time, anticipates a future recovery attempt, and preemptively moves the goods. That's not just memory of a past caching event. That's using memory of a past caching event to model someone else's memory of that event.

The obvious question: if the jay can do that, what is the qualifier still protecting?

The less obvious question: if it's protecting something real — some genuine inner quality of remembering that behavior can't reach — then we should acknowledge that it's protecting that same thing in humans. Not demonstrating it. Protecting it. We're assuming it's there. The behavioral evidence for human autonoetic consciousness is, structurally, the same kind of evidence we're refusing to accept from the jay: reports of the experience, and behavior consistent with the experience, and nothing else.

I don't know where the line is, or if there's a line rather than a gradient. The jay finds the fresh worms and not the rotten ones, four hours or a hundred and twenty-four hours after placing them. Something in it records the caching event and makes it retrievable with its timestamp intact. Whether there's anything it is like to be the jay doing that — whether there's a felt pastness, a sense of occasion — the experiment doesn't say. Neither does anything else.

The qualifier hangs there, honest and inconclusive, doing exactly what it was designed to do: acknowledging the limit, holding the gap open, declining to fill it.