Studia Caroliensia: Papers in Linguistics and Folklore in Honor of Charles E. Gribble

Macro-Level and Micro-Level Structural Analysis of Mixed Content Miscellanies

Macro-Level and Micro-Level Structural Analysis of Mixed Content Miscellanies' David J. Birnbaum Abstract. The present report employs cluster analysis to examine structural correlations in the text of the series of apocryphal "Stories about Abraham" in twenty-two manuscripts. This analysis confirms many of the conclusions that scholars had drawn previously about the textual transmission of these stories using traditional methods; it also identifies new relationships among the manuscripts. Philological Background The series of apocryphal "Stories about Abraham" (henceforth "Abraham") contains up to nine component texts, each of which may appear in up to three redactions (Miltenova 2005)2 The fact that this work consists of discrete and identifiable subcomponents means that from a structural perspective it can be considered a sort of internal fixed-content miscellany, or miscellany within a miscellany, comparable to the many mixed-content or fixed-content miscellany manuscripts that are well known from the middle Bulgarian period (Miltenova 1982)3 "Abraham" is closer in form to a fixed-content miscellany than to a mixed-content miscellany in that the contents are constrained by the need to refer to the Old Testament legend of the patriarch Abraham, and the order of the constituent parts is partially constrained by logical considerations (e.g., we expect that the "Story of Abraham" will precede "Ishmael" and "Isaac" as fathers precede sons chronologically; we expect that "Ishmael," 1 An earlier version of this paper was delivered at the March 2005 Midwest Slavic Conference in Columbus, OH. I am grateful to Sibelan E. S. Forrester, M. A. Johnson, Predrag Matejic, and Anisava Miltenova for comments and suggestions. 2 "Story of Abraham" (two redactions), "Story of Sarah," "How Sarah Advised Abraham" (two redactions), "Ishmael," "Isaac" (two redactions), "Melchizedek," "Abraham and the Holy Trinity" (two redactions), "Death of Abraham" (three redactions), and "Samuel." 3 Other micro-level miscellany-like work clusters include the "Erotapokriseis" (Miltenova 2004), the "Fiziolog," and the "Tales of Evil Women" ("Zlye zeny"). Robert Rothstein, Ernest Scatton, and Charles E. Townsend, eds. Studia Caroliensia: Papers in Linguistics and Folkore in Honor of Charles f. Cribble. Bloomington, IN: Siavica, 2006, 41 - 56. 42 DAVID J. BIRNBAUM when present, will precede "Isaac" textually as Ishmael preceded Isaac chronologically; etc.). A survey of twenty-two manuscripts that contain "Abraham" reveals patterns of structural agreement, which is to say that the appearance of certain "Abraham" stories in specific redactions in a particular manuscript is strongly correlated with the appearance of certain other"Abraham" stories in the same manuscript. An analysis of these correlations shows that these patterns of agreement can be used to divide the corpus into natural groups according to structural similarity. Mixed-Content Miscellanies Miltenova (1982 and elswhere), building on the work of Istrin (1922), Speranskij (1963), and others, has argued that coincidences in the presence or absence and in the order of texts within a miscellany may imply shared textual transmission, and her conclusions have recently been corroborated by computational analysis (Birnbaum 2003, Dubin and Birnbaum 2004). Until now, however, most arguments of this sort (and the supporting computational analysis) have been made on the "macro level," that is, on the level of examining the presence or absence of entire specific works within a mixed-content miscellany manuscript. In fact, however, if we view works like"Abraham" as micro-miscellanies (that is, as miscellanies embedded within miscellanies), it should be possible to apply the same type of analysis at the micro level, as Miltenova has done for"Abraham" (2005) and the erotapokriseis (2004), and to perform the same sort of computational validation. Micro-Level Clustering The present study is based on a micro-level cluster analysis of the "Abraham" texts in twenty-two manuscripts (see list in appendix A and detailed information in Miltenova 2005). Six clustering methods4 were applied both with and without length-normalization, yielding twelve dendrogrammatical visualizations (appendix B). The small number of component texts in "Abraham" suggested that length-normalized similarity calculations were likely to be 4 Cluster analysis builds family trees by grouping items according to degree of similarity. The present study used the following six methods: agglomerative single linkage, average linkage, complete linkage, Ward's linkage, weighted linkage, and divisive linkage. Similarities were calculated according to an algorithm first presented in Birnbaum 2003 and described from a clustering perspective in Dubin and Birnbaum 2004. Dendrograms were generated with the R software system. Structural Analysis of Mixed Content Miscellanies 43 most revealing, and...