Archaeological animal bone assemblages are often highly fragmented, meaning that for over 70% of the recovered bone fragments we do not know what animal (or human) species they belonged to. This is especially problematic in Paleolithic contexts, when both humans and carnivores repeatedly occupied the same caves and rock shelters. Identifying bone fragments from these dual occupation contexts through Zooarchaeology by Mass Spectrometry (ZooMS) can provide additional insights into both carnivore and human behavior. In this paper, we apply ZooMS to the majority of morphologically unidentifiable bone fragments larger than 20mm (n=840) recovered from the 2012-2013 excavated Châtelperronian layer of Cassenade (France). Collagen was extracted using an ammoniumbicarbonate (AmBic) buffer and over 99% of the sampled bone fragments could be identified taxonomically. While the proportion of Equidae is similar in both ZooMS and zooarchaeological components, Bos/Bison is represented by a threefold increase in the ZooMS fraction (50.5% vs. 16.6%). Conversely, Ursidae, the dominant taxa in the morphologically identifiable remains (36.4%), only formed 7.3% of the ZooMS fragments. Carnivores are also present, but in low numbers (0-2%), and include Hyaenidae, Panthera and Canidae.

In the ZooMS fraction, only few bones show traces of human activity (ca. 2%), which is most likely related to low bone surface readability. These show that human groups at Cassenade were processing Bovinae and Equidae, as well as Ursidae and rhinoceros. Conversely, traces of carnivore activity are abundant, and we were able to taxonomically identify 334 bone fragments that were digested by carnivores (as indicated by acid etched surfaces). While large proportions of the Rhinocerotidae (63.79%), Elephantidae (52%), Equidae (48%), and Bos/Bison (45%) remains were digested by carnivores, this is only the case for 1.7% of the Ursidae ZooMS fragments. Three dimensional data are available for all the ZooMS-identified fragments and confirm the near-exclusive presence of cave bear in the lower part of the sequence. Further, the ZooMS spatial data identified a restricted presence of mammoth in the middle part of the sequence and a diverging presence of reindeer and Cervid/saiga remains at the bottom and top. Overall, this study illustrates the added value of integrating zooarchaeological and ZooMS datasets to obtain additional insights into past ecologies, changing site use, carnivore diets, and human subsistence practices.