DNA appears to decay by random chain scission resulting in a predictable range of fragment lengths. Collagen decay has also been modelled in this same way, although it has become increasingly evident that collagen decay does not follow this same pattern. Radiocarbon and stable isotope analysis now use ultra-filtration to isolate large fragments (>30% of original polymer length) even in Pleistocene bone. How then does collagen decay? This study contrasts experimentally degraded samples with collagen extracted from forensic, archaeological and fossil bone. In experimentally degraded bone, values for amino acid and elemental (C:N) composition, bulk δ 13C, δ 15N, and aspartic acid racemisation (AAR) changed very little until 99% of the collagen was lost, suggesting that the collagen triple helix and polypeptide chains remained remarkably intact. This suggestion was demonstrated directly by examining the integrity of individual polypeptide chains using cyanogen bromide (CNBr) cleavage followed by SDS-PAGE electrophoresis. In ancient samples, AAR values remain remarkably stable and the pattern of CNBr-cleavage was only replaced with a smear of smaller polypeptides in the oldest (Pleistocene) bones investigated. Smearing may reflect both modification of the methionine resides (the sites of CNBr-cleavage) and/or partial hydrolysis of the collagen molecule. The findings reveal why it is not usually necessary to worry about collagen diagenesis; it is mostly intact. However, evidence of partial deterioration of the oldest bone samples suggests that alternative purification strategies may increase yields in some samples.