Conceived as a model study for the enantioselective synthesis of the natural product (S)-bakuchiol, the deoxygenation of a recently described trishomoallyl alc. was attempted using the Kabalka modification of the Wolff-Kishner method after prior oxidn. of the alc. to the ketone. However, an unprecedented succession of homoallyl/cyclopropyl ring closure and ring opening altered the carbon skeleton profoundly. The crucial intermediates are assumed to be radicals although carbanions cannot yet be definitely ruled out. [on SciFinder (R)]