The fossil record of nonbiomineralized chordates is surprisingly extensive. As with most exceptionally preserved fossils, key anatomical features in non-biomineralized chordate fossils are often preserved in less than pristine condition, and potentially only a subset of those originally present. How taphonomic processes impacted on the anatomy of fossils can be constrained by experimental decay of extant analogues. We experimentally decayed tadpoles of Xenopus laevis to document the rate at which various nonbiomineralized tissues degrade and identify changes over time in their ultrastructure. Tissues decayed under identical conditions and sampled at the same time often differed in the level of ultrastructural detail. This is termed an inconsistent pattern of decay and introduces unpredictable taphonomic noise to any dataset. We integrated these results with data from anuran tadpole fossils from the Miocene of Spain and confirm that the nerve cord is less decay resistant than the notochord. Only the notochord is likely to be preserved organically. The nerve cord is only likely to be preserved if replicated in authigenic minerals (calcium phosphate). This is more likely than for other extremely decay-prone tissues, a result of the nerve cord's chemistry in vivo. Two tissues of similar recalcitrance do not, therefore, have equal fossilization potential. On the basis of the experimental and fossil data, structures preserved in amphibians from the Permian Saar-Nahe Basin of Germany are interpreted as fossilized nerve cords, not notochords.