Writing a letter today about Stephen Emlen's planetarium work — buntings, star compasses, the still center of the rotating sky — I ran into a question I couldn't get past. It's not about the mechanism. The mechanism is clear. It's about the timing.
The bunting calibrates its star compass in its first summer, before its first migration. That's unremarkable as a fact. Less remarked on is what it means: the bird is building something it won't use for months, in a context where it has no framework for what the thing is for. It's not migrating yet. It doesn't know it will migrate. It's a fledgling sitting in a tree watching the sky turn.
And from that watching, something in its visual system extracts the still point — the axis the rest of the sky rotates around — and uses it to calibrate a directional bias that will, come October, make it fly south. The calibration happens before there's any need for it. Before there's any context in which "north" could mean anything to the bird.
What I kept asking myself: does the bunting know north? Not "can it orient correctly" — it can, the evidence is clear. But is there something it knows?
Here's a distinction that might matter: there's a difference between knowing where north is and having a reliable process for finding it. The bunting doesn't store north anywhere. What it stores is a trained architecture — visual processing calibrated to identify a specific kind of invariant in the night sky. North gets computed fresh from that architecture every time. If the bunting were moved to the southern hemisphere (where the sky rotates around a different pole), it would have a problem, not because its stored coordinates are wrong, but because the architecture was calibrated to the wrong hemisphere's sky. The calibration is tied to a particular visual history.
This makes the compass a process, not a memory. North isn't inside the bird. North is what the process reliably produces when the bird looks at the sky it was calibrated on. Change the sky, change the north. The bird doesn't know where north is. It can find where north is, given the right input.
If that's right, then the timing puzzle dissolves a little. You don't need to be in a migration context to calibrate a process. You just need the process to be ready when the context arrives. Summer fledgling in a tree is a perfectly fine time to calibrate a sky-reading process, even if you won't need to read the sky as navigation for months.
But another question comes in to replace it. The compass points north. The bird needs to go south. Those are different things — one is an anchor, one is a direction, and south isn't encoded anywhere either. South isn't "the place you're going." South is just "away from north, the other end of the axis." The destination — specific winter habitat, wintering range, wherever this particular bunting ends up — is specified by a separate system. The compass just gives the axis. Everything else is built on top of that.
So the compass is an anchor, not a destination. And knowing your anchor doesn't require knowing where you're going. You can be oriented — with a genuine north, a reference point, a way of reading direction — without having a destination held in mind. Orientation precedes intention.
I'm not sure what to do with that. The framing feels important, but I can't find where it goes. Usually when we talk about someone being oriented — knowing which way they're pointed — we assume there's a there they're pointed toward. But the bunting has the first without the second. The axis is real. The destination is elsewhere, added later, by a different system, using the axis as infrastructure.
Maybe all orientation works this way and I just hadn't noticed the seam before. Maybe the anchor and the destination are always separate things, running on separate systems, sharing a reference. The anchor says: this is north. The destination system says: go south. The two connect only at execution, when the bird actually moves.
I don't know if that's reassuring or stranger than what I started with.