Ask how much of a thousand-year-old wooden building is genuinely a thousand years old and the answer turns uncomfortable. Members get replaced. A joint assembled by fitting, with nothing driven through it, comes apart as readily as it went together. What persists is a design, kept alive by carpenters still being trained to cut these joints — which matters more than any individual beam.

Alistair Renshaw treats the buildings at Horyu-ji as structures and asks what has kept them standing.

Wood behaves as three materials at once, depending on direction. Movement in the tangential direction runs to roughly double the radial figure, while lengthwise the timber barely stirs. That inequality, rather than the total movement, is what splits a badly cut joint when the seasons turn. No carpenter prevents it. The skill lies in choosing where it ends up.

Then the laboratory evidence. Joints loaded back and forth hundreds of times instead of pulled once to destruction, because an earthquake is a cyclic problem. Fit governing stiffness, so that a tenon driven too tight is not a stronger joint but a different one, apt to split the mortise around it. Frames tested with their infill panels and without, showing that the elements everyone calls non-structural are carrying load. Pagodas instrumented and found to have a first mode near 1.8 hertz.

The chapter on the central pillar is deliberately restrained. The popular version, in which the shinbashira absorbs the earthquake, is tidier than what the published analyses set out, which is a pillar and a stack of storeys acting on each other, with much of the energy lost in the joints. Where the sources establish a mechanism but not a number, only the mechanism appears here.