The Mendocino Triple Junction, where Northern California meets the Pacific Northwest, is where three tectonic plates converge: the Pacific, the North American, and the Gorda (the southern remnant of the Juan de Fuca plate). Textbooks draw three plates meeting at a point. The geometry is clean.
Shelly, Thomas, Materna, and Skoumal (Science, January 2026) tracked thousands of low-frequency earthquakes — tremors thousands of times weaker than anything felt at the surface — using Pacific Northwest seismometers. The microearthquakes mapped out sliding surfaces that didn't match the three-plate geometry. There are five moving pieces, not three.
The extra two: the Pioneer fragment, a chunk of the ancient Farallon plate that mostly subducted 55 million years ago but persists as a mechanically active blob being dragged beneath North America by the Pacific plate. And a broken piece of the North American plate itself, pulled downward at the southern end of the Cascadia subduction zone by the descending Gorda plate. Both are invisible from the surface. Neither has any topographic expression. They exist only as structures that move, generate microearthquakes, and respond to tidal stress.
The tidal correlation is the diagnostic. The Sun and Moon exert periodic stress on the crust, and where that stress aligns with a plate's direction of motion, microearthquake frequency increases. The tidal response at different depths and locations traced out the geometry of the hidden pieces — surfaces that slide against each other too weakly to generate felt earthquakes but not too weakly to be modulated by tides.
The five-piece model resolves anomalies the three-piece model couldn't explain. The 1992 magnitude 7.2 Petrolia earthquake occurred at a depth far shallower than the standard subduction geometry predicted. With the revised geometry — shallower plate boundaries, an extra fragment — the depth makes sense.
The structural insight: the standard model wasn't wrong about the existence of three plates. It was wrong about the claim that three pieces are sufficient. The extra pieces have no surface signature. They generate no large earthquakes. The only evidence for them is a pattern in microearthquakes too weak to feel, validated by a correlation with tides too small to measure without instruments. The operating structure of the junction includes components that are invisible to every observation method except the weakest signals.