Is vertebral form a valid species-specific indicator for salmonids? The discrimination rate of trout and Atlantic salmon from archaeological to modern times

• We used geometric morphometrics and kNN classifications on fish vertebrae.
• We used a modern referential and archaeological bones.
• We used broken archaeological vertebrae.
• We demonstrated the usefulness of using fragmented bones.
• We found more trout than salmon at “Le Taillis des Coteaux” cave.

Salmonids, especially Atlantic salmon (Salmo salar) and trout (Salmo trutta sspp.), are ubiquitous throughout the European Upper Paleolithic archaeological context; however, as these species are morphologically similar, species identification can be difficult. Here we present (I) a species classification rate for the two species using modern reference specimens, and (II) an application on archaeological vertebrae of Salmo sp. recovered from the cave of Le Taillis des Coteaux (Vienne, France). This cave contains a rich archaeological sequence with an exceptionally well preserved stratigraphy, encompassing the Upper Paleolithic from the Aurignacian to the Middle Magdalenian. To discriminate both species, we used a geometric morphometric approach to analyse vertebral form using landmarks and sliding semi-landmarks, coupled with k-Nearest Neighbour classification method. Other quantitative approaches like Principal Component Analyses exist, but the k-NN method increases the power of these analyses. Linear Discriminant Analysis was also used; however, the k-NN method provided better results. This study presents the initial results of geometric morphometric discrimination of European salmonid bones. The success rate of reassigning these to the modern reference specimens ranged from 84 to 100%, depending on the vertebrae studied, with the data indicating that S. trutta sspp. were the dominant taxon on site. This study provides clear evidence that vertebrae morphology can be used to differentiate salmonid species, allowing archaeological specimens, even partially broken, to be identified.