This paper explores the possibility of improving the performance of\nspecialized parsers for pre-modern Slavic by training them on data from\ndifferent related varieties. Because of their linguistic heterogeneity,\npre-modern Slavic varieties are treated as low-resource historical languages,\nwhereby cross-dialectal treebank data may be exploited to overcome data\nscarcity and attempt the training of a variety-agnostic parser. Previous\nexperiments on early Slavic dependency parsing are discussed, particularly with\nregard to their ability to tackle different orthographic, regional and\nstylistic features. A generic pre-modern Slavic parser and two specialized\nparsers -- one for East Slavic and one for South Slavic -- are trained using\njPTDP (Nguyen & Verspoor 2018), a neural network model for joint part-of-speech\n(POS) tagging and dependency parsing which had shown promising results on a\nnumber of Universal Dependency (UD) treebanks, including Old Church Slavonic\n(OCS). With these experiments, a new state of the art is obtained for both OCS\n(83.79\\% unlabelled attachment score (UAS) and 78.43\\% labelled attachement\nscore (LAS)) and Old East Slavic (OES) (85.7\\% UAS and 80.16\\% LAS).\n