(Source: Introduction to A Linguistic Atlas of Early Middle English, ch. 3, http://www.lel.ed.ac.uk/ihd/laeme1/pdf/Introchap3.pdf)
Our primary evidence for medieval language is manuscript texts. In compiling our corpus of early Middle English texts for tagging, we transcribe from originals or (more often) from photographic reproductions, and not from editions. Printed editions can be useful reference tools. They may help in interpretation of manuscript readings, while checking against the texts of editions can help detect possible errors in our own transcriptions. But for any investigation of historical language variation it is crucial for reasons of comparability and authenticity that, where possible, the manuscript be used as the primary source
From the point of view of linguistic study these are the problems that render use of editions deeply problematical
(a) Many if not most editors silently expand manuscript abbreviations, taking as the form of the expansion the scribe’s ‘usual’ unabbreviated spelling.
(b) Some editors make a virtue of ‘normalising’ or ‘modernising’ texts to create easier reading versions for students.
(c) Equally suspect for our purposes are conflate editions compiled from numerous different scribal witnesses with the aim of producing some imaginary ‘best text’ that never existed in any time or place.
(d) An editor will often emend a form that he believes to be erroneous to one that he thinks the scribe (or original author) ‘intended’. Of course scribes did make mistakes and some such emendations would probably have been approved by the errant scribe himself. But we cannot know this, and we must not suppose that a scribe ‘really meant’ anything that he did not in fact produce.
(e) It might be argued that the editorial conventions practised in (a)–(d) are only harmful if one is interested in the detail of manuscript orthography; historical syntax (and perhaps regionally conditioned syntax) will be unaffected by editorial interference. Most editors of medieval texts, however, add modern punctuation and suppress such manuscript punctuation as exists. Manuscript word division is frequently ‘regularised’ along modern lines. This enables medieval texts to be subjected to the same types of syntactic analysis as modern ones and all too easily allows the assumption that medieval scribes had attitudes towards word, phrase and clause structure similar to our own. The use of diplomatic transcriptions from originals can challenge such assumptions.
(f) It is vital for linguistic study that each individual scribal contribution be treated separately, and in our corpus each text language is indexed and sorted individually. Conscientious compilers of single text editions will notice any changes of hand in their manuscript. But scholars trawling an edition for linguistic evidence may not always succeed in maintaining the distinction. Even if these broad distinctions are maintained in a printed edition, such care does not always extend to scribal corrections in the text. These may be interlinear, intralinear or marginal insertions. If they are made by the scribe who wrote the main text, whether as he went along or as a separate exercise later, the silent inclusion of the changes may not be too damaging. But it is often difficult, even when dealing with originals (and harder still with black and white photographic reproductions), to be sure whose hand has made a correction.