Manuscript Studies
Medieval and Early Modern

V. Textual Analysis: Thorpe, on Principles of Textual Bibliography

P. 108f. Textual analysis: one part of the whole area of textual criticism.
-- Recension: selection of the most authoritative evidence for the text (the closest we can get to reconstructing authorial intention: must consider all available evidence, including letters, etc., and not just apply the rule of "last edition during author's lifetime").
-- Examination: determination of whether the evidence collected yields the "original" reading.
-- Conjecture: any necessary reconstruction of the "original" reading.

P. 112 Substantive and non-substantive variants: substantive variants change the substance of what is said (and, so, can include punctuation and capitalization where these can affect the meaning); substantive variants are the "material" of textual analysis.
Two types of variants:
-- Errors of transmission
-- Intentional alterations
It is usually assumed that a copy will perpetuate the errors of the version from which it is copied plus introduce new variants: this assumption is used to put copies in a chronological sequence.

Pp. 113ff. Three types of approach to textual analysis:
1. The Classical or "Standard" Approach.
The "genealogical" approach of Karl Lachman: construction of a family tree / stemma, based upon "conjunctive" errors:
-- "separative" errors distinguish one version from another;
-- "conjunctive" errors prove that one copy and another are from the same version.
This approach gives the critic much latitude; "its application to a given problem is subjective to a considerable degree" (p. 114)--this subjectivity residing in the critic's freedom to decide which reading is "error" and which is "original."

2. The "Best Text" Approach:
Associated with the name of Joseph Bédier, who expressed scepticism about the genealogical approach: most stemmata seemed to indicate that all copies ultimately descend from two copies made from some lost archetype--why should there always have been two and only two copies of the author's manuscript? There seemed to be some subjective presupposition influencing the construction of the stemma; he went on to demonstrate that, in some instances at least, equally good alternative stemmata could be derived from precisely the same evidence used by a given critic. Thus he concluded that there were fallacies in the basic "genealogical" theory. Bédier proposed, instead, to select a "best text" based upon the "coherence of sense, regularity of spelling, and form of grammar" (p. 114) without the assumption that this, necessarily, was closest to the "author's" manuscript. Some scholars consider this a repudiation of method, a counsel of despair: the value of the results depends entirely on the judgment and taste of the textual critic in determining "coherence," "regularity," etc. Further, there is a problem in distinguishing "error" from "original reading" based solely on consistency and "sense," since often several readings seem equally good (and, so, equally "authorial" without some further evidence).

3. The Statistical Approach:
Associated with the names of Quentin and Greg: this approach attempts pure objectivity, getting away from all qualitative judgments. Here, all variants are collected without prior selection, all are compared, and probability calculus is used to analyze the variants and select the "most likely to have been authorial" reading. Unfortunately, this approach has proven unproductive: no worthwhile results have yet been achieved. Further, it is laborious, and the labour increases in geometrical proportion to the complexity of the textual tradition. And, ultimately, this approach attempts to resolve a problem which is theoretically not resolvable by mathematics; moreover, probabilities are, at best, educated guesses and, so, ultimately unverifiable.

Generally, there are several objections which can be raised to past methods:
1. The possibility of "conflation" has to be admitted (too often textual critics have simply declared that conflation is assumed not to have taken place): what if the actual history of a text had the original split into two versions in the first generation but a second generation copy was "edited" from the two versions, bringing them back to unity but not authorial unity? The possibility of conflation, in fact, makes reliable reconstruction impossible, so it is easier to assume that it does not happen: however, there are known incidents, such as in the 160 MSS of St. Cyprian's De ecclesiae catholicae which give evidence that the work was "edited" from multiple versions in the ninth century and possibly several more times earlier.
2. The greater the number of manuscripts the less likely that analysis will provide a satisfactory account of all of the interrelationships; the postulation of "missing" versions multiplies, making the stemma increasingly unverifiable and arbitrary. "If human ingenuity were set loose alone on the scents of a pack of foxes--to reverse the common conditions of the hunt--it is unlikely that it would be the patriarch fox which would chance to become the one caught" (p. 122), a likelihood which decreases proportionately to the number of total foxes. Analysis / recension can leave so many questions open that, in the end, any variant from any MS might be authorial, and then stemmata and hierarchical ordering serves no purpose at all.
3. There is a danger of making assumptions which place hard problems in "forbidden territory." Paul Maas, for example, makes an assumption that conflation (contaminatio) has never taken place; he knows this not to be true, but he must make the assumption in order to procede at all, but this is hardly a satisfactory position. Similarly, the assumption of the Statistical approach that the reading appearing in the majority of MSS must be "right" because "authorial" is problematic because it begs too many questions. The common assumption that the oldest version is the closest to the authorial version has been proven, in several known cases, to have been a false assumption, so this assumption needs to be questioned. The assumption that the "consistency" and "regularity" of a "best text" mush be authorial consistency rather than scribal or editorial consistency does not stand up to scrutiny, either. These problems must be faced directly by textual critics rather than relying upon common practices of past editors, because the approach to dealing with these problems will vary from one project to another: one's response to these questions should be determined by the nature of the evidence to hand.