Archaeological bones are now routinely dated in many radiocarbon laboratories through the extraction of “collagen.” Methods for “collagen” extraction vary, and several laboratories now apply an ultrafiltration step after gelatinization to extract the higher molecular weight (usually >10 or 30kDa) fraction for dating, thereby removing low molecular weight contaminants. Ultrafiltration has been demonstrated to result in products that are easier to handle and have more acceptable C:N ratios, and in some instances can result in significantly improved (generally older) 14C dates when compared to non-ultrafiltered products from the same bone. Although it has been suggested that ultrafiltration removes potential contaminants such as short-chain degraded collagen and other peptides and amino acids, fulvic acids, and salts, there remains little published evidence to support this. This paper presents data from a pilot study investigating the most suitable techniques with which to study the products of the routine “collagen” extraction procedures employed at the Oxford Radiocarbon Accelerator Unit (ORAU) (modified Longin followed by ultrafiltration). The preliminary data demonstrates that the final product of “collagen” extraction at ORAU appears to be an aggregate consisting of a range of proteins of different molecular weights, including collagen, as well as some other organic matter and inorganic species. Ultrafiltration is removing some, but not all, of the <30kDa fraction from the samples. Further work to investigate the nature of this aggregate and how best to improve the efficiency of “collagen” extraction procedures is discussed.