Electronically available corpora have inevitably enhanced access to historical texts, but building historical corpora can often, unintentionally, distort historical language. This paper discusses how issues of faithfulness and sampling can impact the reliability of a historical corpus-based analysis, with recommendations provided for both corpus users and corpus builders. First, by comparing the Beowulf manuscript with the language found in a historical corpus, this paper shows how seemingly innocent editorial changes such as spacing, punctuation, and spelling can have linguistic consequences for phonological, morphological, syntactic, and socio-historical analyses. The removal or addition of elements is argued to have both theoretical and methodological implications. Second, this paper discusses how sampling issues can lead to erroneous conclusions by corpus users about historical language. Recommendations are provided regarding supplementation and mixed-method approaches.
James M. Stratton (Sun,) studied this question.