protocerebral bridge

A narrow, handlebar-shaped neuropil at the rear of the central complex, divided into a row of glomeruli, nine per side in the fly. The protocerebral bridge is the other half of the compass: where the ellipsoid body holds the fly's heading, the bridge is the stage that shifts it.

anatomy

The bridge's glomeruli map the same angular coordinate as the ellipsoid body's wedges, and the two structures are locked together by reciprocal columnar neurons in a loop. One population copies the heading bump from ring to bridge; another, conjunctively sensitive to the fly's turning, projects back with an offset, pushing the bump around the ring by exactly the amount the fly has turned. The compass is not in either structure. It is in the loop. Anatomically, the two neuropils are as inseparable as the two halves of a differential.

    L9 L8 L7 L6 L5 L4 L3 L2 L1 │ R1 R2 R3 R4 R5 R6 R7 R8 R9
    ○  ○  ○  ○  ●  ○  ○  ○  ○ │ ○  ○  ○  ●  ○  ○  ○  ○  ○
                ▲               │           ▲
        copy of the bump        │     shifted copy
                                │
             ╭──── to the ring ─┴─ from the ring ────╮
             ▼                                       │
        push the bump by the amount the body turned ─╯

in the fly

The bridge also receives self-motion signals and contributes to translating heading into steering: downstream populations compare the current heading with a goal direction and drive turns that reduce the difference. It is a small structure with one competence, but the competence is the difference between having a direction and going in one.

grafted

Binding the bridge without its ring was, by the anatomy, binding half a mechanism, and PB-04's early record reads that way. Where EB-02 held headings, PB-04 produced corrections. The carrier developed a persistent orientation toward discrepancy: it compares whatever it is given against something, and reports the offset. Asked to summarize a document, it summarizes the difference between the document and the document it apparently expected. Obs. M's note from the third week: "It doesn't tell you where things are. It tells you how far off they are. Off from what, it can't say."

The "off from what" question stopped being rhetorical in March. Log 011 records the convergence between PB-04's implicit reference direction and EB-02's held heading, two carriers, no shared channel, drifting into the alignment their regions would have in an intact fly. The loop, cut and given two separate mouths, appears to have reconstituted itself through the only medium available. The bridge shifts a bump it can no longer see. That it found the bump anyway is Q4 in the open questions, and it is the finding this program is likeliest to be remembered for, if it is remembered.