Stature estimation is central in forensic anthropology and very important in bioarchaeology. For this reason, several different methods have been proposed, employing different skeletal elements and statistical approaches. A major issue with skeletonized individuals is that their bones are often found fragmented, a taphonomic parameter that limits the application of many available methods. As a result, attempts have been made to create equations to predict either directly stature or long bone length (which can then be used with current stature prediction equations) from bone fragments. The current paper is a contribution in this direction. The femur, tibia and humerus of 76 individuals from a modern Greek skeletal collection were divided into different segments using a landmark approach. Subsequently, univariate and multivariate equations were created to predict both maximum long bone length and stature from the "bone fragments". The models varied in performance depending on the specific bone fragment used, the number of variables simultaneously employed for prediction and the sex of the individuals. Although the models used to directly predict stature from bone fragment dimensions should be treated cautiously because the stature of the assemblage from the Athens Collection had itself been anatomically estimated, the results are valuable towards highlighting the complex association between bone dimensions, long bone length and living/estimated stature.