Assessing the Manuscripts of Sefer Ḥasidim

This article examines the content and structure of the manuscripts of Sefer Ḥasidim, engaging with ideas concerning its production addressed in Ivan Marcus’s recently published book on Sefer Ḥasidim. Marcus has argued that the book was written piece by piece and not as an integral book and further suggested that each and every manuscript of Sefer Ḥasidim should be taken as a distinct edition of the book prepared by Judah he-Ḥasid. The present study demonstrates that, notwithstanding the gradual process in which Sefer Ḥasidim was written and the great variations among the manuscripts, it is possible to reconstruct a textual process that led to the larger compilations found in the three well-known text editions of Sefer Ḥasidim, represented by MS Parma 3280, MS JTS Boesky 45, and the edition printed in Bologna in 1538. The analysis focuses on the distribution of the text in the manuscripts. While it is difficult to show linear relations among them, the different versions demonstrate a gradual process of growth and enlargement of the material around topical structures. Since most of the material is transmitted in more than one exemplar and few passages appear in one manuscript alone, it is argued that the manuscripts can be linked to show how the material grew from random collections of single paragraphs to topically ordered clusters and into the larger compilations of Sefer Ḥasidim.

Sefer H . asidim (Book of the Pious; thirteenth-century Ashkenaz) comprises a collection of ethical teachings and religious ideas from the circle referred to as "H . asidei Ashkenaz," represented by Judah he-H . asid of Regensburg (d. 1217), his father Samuel he-H . asid, and Judah's disciple Eleazar ben Judah of Worms (ha-Roqeah . ). 1 Their writings also include Bible commentaries, commentaries on the prayers, and ethical and mystical treatises. 2 Sefer H . asidim has interested scholars of medieval Jewish history since the nineteenth century, especially because the book is replete with historical realia that describe the everyday circumstances of Jews living in a Christian environment. 3 The book also serves as an important source on the religious and intellectual history of the Jews in medieval Ashkenaz.
Recently, the focus of scholarly attention has moved from the book's content to its form and structure and to the history of its transmission. Initially, Sefer H . asidim was believed to have survived in two distinct versions: the Bologna print edition of 1538 (known as SHB) and the Parma manuscript (known as SHP). 4 The launch of the Princeton University Sefer Hasidim Database (PUSHD) in 2007 provided access to all known manuscripts of Sefer H . asidim 5  that SHP and SHB embody only two possible versions of this work. 6 Ivan G. Marcus's 2018 monograph Sefer Hasidim and the Ashkenazic Book in Medieval Europe offers a new and innovative understanding of book production of Sefer H . asidim and other works produced by the H . asidei Ashkenaz. Marcus posits that the Ashkenazic Pietists did not set out to produce a single work in a final version but, rather, to collect material in notes and separate units to be compiled into larger sections with a topical structure. In this conception the various textual elements or single paragraph units often varied in their position within the larger framework. In other words, the book began as a collection of single paragraphs that only later coalesced into a book. Given the various manuscripts of Sefer H . asidim that reflect quite a diverse transmission, Marcus proposes abandoning the concept of SHB and SHP as the principal recensions of Sefer H . asidim, and that instead each manuscript should be regarded as a free-standing version of the text produced by Judah he-H . asid. 7 This article will reexamine the manuscript transmission of Sefer H . asidim, with particular focus on the distribution of material within the manuscripts, 8 seeking to establish relationships between the structure and the content of each manuscript. To that end, I closely examine the transmission of Sefer H . asidim through the known editions and manuscripts as well as the distribution of the material within these manuscripts. My argument is that despite variation within the manuscripts, SHB and SHP should still be considered as major recensions of Sefer H . asidim. Tracing the relationship between the manuscripts using the model of book production that Marcus introduced and that the copyist of MS Cambridge Add. 379 described as "gathering the berries," I will focus on the production process of the longer manuscripts, especially MS Parma 3280 (de Rossi 1133; dated to around 1300 in Ashkenaz). 9 I compare its structure and content to those manuscripts with shorter versions of Sefer H . asidim and show that the Parma manuscript is a product of growth of material-by which I mean that MS Parma 3280 represents a final stage in the process of gathering and enlarging the extant material attributed to Judah he-H . asid.

Character of Sefer H . asidim as a "Book"
Sefer H . asidim consists of paragraphs (simanim) or short units, some of a narrative character (maasim/stories), arranged in a topical order, but lacking a systematic structure. This is reflected in the manuscripts themselves. Not one of the surviving manuscripts of Sefer H . asidim is identical to another. All the manuscripts differ in scope, content, and often also in structure. 10 Marcus has suggested that Judah he-H . asid collected these small textual units into clusters and may have compiled them into a "book." (This also could have been done by one of his disciples.) 11 Such a process is reflected in use of the terms mah . barot, liqut . im, or quntras for notebooks, excerpts, or folios of Sefer H . asidim transmitted in manuscripts. 12 What this indicates is that Sefer H . asidim is not a book in the traditional sense, that is, there is no Urtext, and it is likely that there never was an original version written by Judah he-H . asid. It evades the classical characterization as a single work produced by a single author at a specific date. The book was not written in one piece, but was collected and compiled from textual units into clusters in various versions (quite similar to modern book production in the age of copy and paste). All these different forms are found in the manuscripts.
In his recent book, Marcus claims to be able to reconstruct no less than fourteen editions of Sefer H . asidim based on the material and its distribution in the manuscripts. 13 These editions can consist of topical notebooks in various combinations, but they can also contain single paragraphs collected and copied, sometimes referring to the same subject, sometimes gathered without any apparent connection. The smallest unit in Sefer H . asidim is the single paragraph (siman). 14 These paragraphs were collected and compiled, some-10 See the description of the manuscripts in Marcus, Sefer Hasidim and the Ashkenazic Book, 90-112. 11 Marcus, Sefer Hasidim and the Ashkenazic Book, esp. chap. 1; idem, "Recensions and Structure"; Johann Maier, "Rab und Chakam im Sefer Chasidim," in Das aschkenasische Rabbinat: Studien über Glaube und Schicksal, ed. Julius Carlebach (Berlin, 1995), 37-118. 12 Marcus, Sefer Hasidim and the Ashkenazic Book, 17-20; sometimes it even says, "another Sefer H . asidim," ibid., 20. 13 Marcus, Sefer Hasidim and the Ashkenazic Book, 19, 29-30. 14 Johann Maier already stated this in his "Rab und Chakam im Sefer Chasidim," 42. times according to a topical order in a booklet (mah . beret). These booklets were then combined into bigger compilations of Sefer H . asidim represented in the longer manuscripts. Marcus describes this process using the image of an inverted pyramid, with the point on the bottom denoting the later editions and the broad upper section representing the many different paragraph units that appear in the manuscripts. 15 In his recent work, Marcus returns to an argument he made in 1978 that there are five compilations of topical booklets found in the various versions of Sefer H . asidim (SHB I, II, III and SHP I and II). 16 From this starting point, he has now defined several combinations of these topical booklets, represented in the longer manuscripts-MS Parma 3280, JTS Boesky 45, MS Cambridge Add. 379, and MS Oxford Add. 34, as well as the edition printed in Bologna in 1538. 17 The remaining manuscripts contain only single paragraph collections, often with parallels in the larger compilations, but sometimes including original material not found in the other manuscripts. Yet, the majority of the Sefer H . asidim material is transmitted as single paragraphs in collections and as part of a larger textual unit-a booklet or a compilation of booklets.
This very fluid corpus of material is arranged in different combinations in the manuscripts. Comparing parallel paragraph collections in different manuscripts, we find that on occasion individual paragraphs are missing or added to these collections. Marcus calls this phenomenon the "independent circulation of individual paragraphs." 18 In what follows I will show that the individual paragraphs do not always circulate completely haphazardly, but rather that there was a process of structuring and growth of the material. This process is reflected in the fact that few simanim exist in only a single manuscript and that Parma 3280 encompasses the textual corpus of Sefer H . asidim almost in its entirety. 19

The Production of MS Parma 3280
Parma 3280 represents the largest textual corpus of Sefer H . asidim. The manuscript was written in Ashkenaz around 1300, meaning that the cumu-lation of the textual units of Sefer H . asidim transmitted in the various other surviving manuscripts took place within eighty years after Judah he-H . asid's death in 1217. This supports the view that the book was not written as a single work at its inception but, rather, was composed of combinations of textual units and clusters eventually compiled into one long version, the one represented in Parma 3280. While Judah he-H . asid himself may have produced the full manuscript from his mah . barot, the production process, as reconstructed below, suggests that his disciples or followers have carried out this task. 20 One fact that is clearly reflected in the transmitted manuscripts of Sefer H . asidim is that within a relatively short period of time following Judah he-H . asid's death, the text went through a process of repeated expansion until it became the comprehensive version found in Parma 3280.
Marcus argued that Parma 3280 consists of two compilations of topicallyordered booklets. 21 The first part of the manuscript ( § § 1-1385 = SHP I) consists of a compilation of five blocks concerning Shabbat, books, Torah study, charity, and honoring one's parents. The second part of the manuscript ( § § 1386-1983 = SHP II) comprises sixteen blocks of various contents, some of which, such as Torah study and books, are also discussed in SHP I. Interestingly, such passages in SHP II are introduced with the words: "Subject X is also discussed here." Other passages address subjects not included in SHP I, such as oaths, prayer, and damages. It therefore appears that the second part of Parma 3280 (SHP II) consists of text clusters on various topics that were collected and added to the compilation represented in the first part of the manuscript.
This assumption is corroborated by another manuscript: former JTS Boesky 45, now in private hands (Italy, fifteenth century). 22 This witness of Sefer H . asidim shows the same textual structure as Parma 3280 and comprises the same material in the first compilation of booklets ( § § 1-1385 = SHP I), differing from the Parma manuscript only in the numbering of paragraphs. When it comes to the second compilation (SHP II), in comparison to Parma 3280, the Boesky manuscript has less material and a considerable number of paragraphs that are absent entirely, as shown in table 1.
The material presented in Parma 3280 § § 1388, 1391, and 1392, for example, finds no counterpart in the Boesky manuscript. Furthermore, as can 20 No manuscript mentions Judah he-H . asid as the author of Sefer H . asidim. The book was only later ascribed to him; see Joseph A. Skloot, "Printing, Hebrew Book Culture, and Sefer H . asidim" (PhD diss., Columbia University, 2017), 60-69; and the contribution by David Shyovitz to this volume. 21     The distribution of the text in the manuscripts shown in the table supports the hypothesis that SHP II is built of material compiled in stages until it reached its most comprehensive version as reflected in Parma 3280. But during this process of growth the text clusters were not simply appended to the end of the text. The structure of the manuscripts shows that new passages were inserted into the existing topical structure of the compilations. 24 One can think of a skeleton of text clusters on various topics that was fleshed out with further material on the same subject. 25 Viewed from this perspective, the manuscripts containing less material may represent earlier stages in the compilation of Sefer H . asidim than the longer versions. Of course, these observations need to be supported by a full comparison of the contents of the manuscripts. 26 Imposing Marcus's metaphor of the inverted pyramid on the process of producing SHP II by means of the accumulation of material as reconstructed here, we offer the following model: In light of this reconstruction, Oxford Opp. Add. 34 would witness an early, first collection of the material, already structured around certain subjects. Then, the corpus is enlarged and extended by adding further material as evidenced in Cambridge Add. 379. The same process is repeated to reach the stage represented in JTS Boesky 45, which already contains about eighty percent of the material of Parma 3280. We can actually observe Judah he-H . asid and his disciples, or later compilators, working on the text, adding more and more material, until arriving at the final step of Parma 3280.

Manuscripts of Sefer H . asidim Bologna
Turning to the version of Sefer H . asidim printed in Bologna in 1538, it becomes clear that there is a problem. No surviving manuscript of Sefer H . asidim represents the full scope of this printed edition. 27 Marcus defined four compilations within this version: SHB 0 and SHB I, II, and III. Several manuscripts belonging to SHB contain solely the first part (SHB 0 = § § 1-152), which as Marcus and Haym Soloveitchik have demonstrated was not a part of the Sefer H . asidim, but of a different work, referred to as Sefer ha-H . asidut and produced in Northern France. 28 Soloveitchik further argued S. DÖNITZ  30 and then change into the compilation of SHP II that was analyzed above. These findings corroborate Marcus's opinion that the text of Sefer H . asidim was transmitted in more than one version-certainly in more than the versions presented in SHB and SHP-but in so-called "mixed" ones as well. 31 Furthermore, it proves that besides SHB 0, SHB I is also represented in the manuscripts, leaving the question of SHB II and III (according to Marcus's classification) open, as they exist in Bologna 1538 but are not found in known manuscripts of Sefer H . asidim. 32 A closer look at Oxford Opp. Add. 34 (table 4) shows that, as we found in the case of the SHP II block in the manuscripts, this manuscript does not 29 Marcus, "Recensions and Structure," 145, table II. 30 Cambridge Add. 379 shows the same numbering of the paragraphs as in the print Bologna 1538. Even when looking at the readings in the paragraphs, in most cases the text in Cambridge Add. 379 indicates a close connection to the printed version. 31 In the description of the manuscripts in PUSHD both manuscripts are attributed to the socalled "mixed" group. 32 How the Bologna version in its full scope found its way into print remains an open question. During recent discussions, we repeatedly are faced with the situation that the early printed texts do not represent the extant medieval manuscript traditions. There must have been some process of selection and rearrangement in the making of models for printing in the early modern period. In other words, texts were produced in a new version, differing from the medieval manuscript tradition; see Skloot, "Printing, Hebrew Book Culture, and Sefer H . asidim"; Daniel  It is therefore possible to deduce the following: As in the process analyzed above for SHP II, SHB I was compiled from single paragraph units and text clusters to which more material was added according to content and subject. 33 Interestingly, at a certain moment in time this block (SHB I) was restructured and reordered to become SHP I in Parma 3280. SHP I and SHB I are similar in content, but not identical in structure. According to Marcus, SHP I is actually constructed of SHB I and additional material. 34 If this is correct, then Cambridge Add. 379 and Oxford Opp. Add. 34 represent manuscript witnesses to SHB I before new material was added, and the text units were reordered to become SHP I. Again, Oxford Opp. Add. 34 appears to represent the first draft, supplemented with further material found in Cambridge Add. 379. Then, the compilation was restructured, such that 33 It would also be legitimate to assume that Oxford Opp. Add. 34 represents an excerpt from Cambridge Add. 379. Of course, this could be the case. From that point of view, one would have to ask if there are any indications as to why the copyist chose only a selection of paragraphs for his copy-a study that exceeds the framework of this article. See also the discussion of the manuscripts with no correlation to SHB or SHP below. 34 Marcus, "Recensions and Structure," 148-50. SHB I changed into SHP I as represented in JTS Boesky 45 and Parma 3280. Again, we can see the author or compilator working on the text, collecting, and adding material from variant sources and combining them into textual blocks, sometimes restructuring the passages into different versions of the text.
Analysis of the Sefer H . asidim manuscripts enables us to reconstruct the process by which the editions SHP II and SHB I were compiled, based on the assumption that more and more material was added to notes Judah he-H . asid and his disciples wrote down on separate sheets that were later combined into topically arranged blocks and finally arranged in different versions of Sefer H . asidim.

Manuscripts That Do Not Follow the Structure of SHB or SHP
Up to this point, the analysis has focused only on manuscripts that follow the topical structures of either SHB or SHP. Yet, there are also Sefer H . asidim manuscripts that do not follow either of these structures but, rather, are collections of single paragraphs. 35 The first such examples are MS Oxford Opp. 614 and MS Oxford Or. 146 (both Ashkenaz, fourteenth century). 36 Although both manuscripts contain eleven paragraphs each, they do not correspond to the order in SHB or SHP. Interestingly, they only comprise material from SHP I. Accordingly, it could be assumed that these manuscripts represent early collections of notes for the compilation of SHP I that were later rearranged into a larger version with a topical structure and paragraph numbers, similar to the process reconstructed above for SHP II and SHB I. 37 However, we might ask whether these manuscripts comprise collections of single paragraphs or, alternatively, were selections motivated by interest in particular topics made from a more complete source. In another manuscript of Sefer H . asidim it was possible to discern such an interest in a special topic from within the collection, i.e., an interest in prayer. 38 However, a similar intention could not be detected in the two manuscripts discussed here; they contain larger or smaller, seemingly unsystematic, collections of paragraphs. 39 36 H . asidim"). The term liqut . im is An even more interesting, but puzzling, example is the manuscript pair of MS Vatican 285 (Byzantium, fifteenth/sixteenth century) and JTS Reel 2499 (Italy, fourteenth/fifteenth century), 40 both of which contain the same 158 paragraphs. 41 The majority of these paragraphs also appear in SHP, albeit in a different structure. Some seventy-five percent of the paragraphs (116 in number) belong to the compilation of SHP I. These manuscripts contain no material from SHP II. The remaining forty-two paragraphs (approximately twenty-five percent of the total) are not transmitted in any other manuscript of Sefer H . asidim. Other manuscripts, such as Frankfurt Heb. Oct. 94 and Zürich Heidenheim 51, also contain randomly selected paragraphs plus some material not found elsewhere. 42 The question as to what kind of material is not included in the larger recensions of SHB and SHP and why is beyond the scope of the current article but should be addressed in future research in order to get a broader picture of their production and their relation to Parma 3280.
Since the order of the paragraphs in these manuscripts does not match the larger compilations of Sefer H . asidim, and since there is no evidence of topical interest that might have inspired copyists to pick them from a larger collection of material, I am inclined to assume that these manuscripts present early stages in the collection of material that was later sorted according to topics and copied into other, more comprehensive collections. In other words, these witnesses in particular offer a clue regarding the first stages in the organization of Sefer H . asidim, as it grew from single paragraphs in Judah's notebooks to the most comprehensive version reflected in Parma 3280. 43 Consistent with recent notions on book production within medieval European Jewish communities, this hypothetical reconstruction of the compilation process of Sefer H . asidim material enables us to organize the manuscripts hierarchically. Manuscripts containing paragraphs lacking any systematic or topical order as in the pair of Vatican 285/JTS Reel 2499 probably reflect the earliest stage in the process of collecting the material. After this stage material was sorted by topics as illustrated in Oxford Opp. Add. 34. Additional relevant material was then added at different points in the text, as represented in Cambridge Add. 379. Yet another restructuring of the material must have taken place, as seen in the varying order of paragraphs in the first parts of JTS Boesky 45 and Parma 3280. Finally, all the material was compiled into the larger compilations of JTS Boesky 45, Parma 3280, and the Bologna print edition of 1538.
This process probably began in the lifetime of Judah he-H . asid and ended sometime around 1300, when Parma 3280 was produced. While the work likely was begun by Judah he-H . asid, in contrast to Marcus's idea that Judah himself produced all these parallel versions of Sefer H . asidim, it is equally possible that larger compilations that appear in SHB and SHP were the work of his disciples or later copyists and scholars. 44 What is new is the idea that there never was one original version of a book produced by one single author, but rather that the text initially consisted of collections of material and notes that were later compiled into a book, sometimes even generations later by a scholar (or scholars) who collected and synthesized all the material together, as illustrated in Parma 3280 of Sefer H . asidim. 45 This active style of redacting a text seems congruent with the idea that medieval Jewish book production involved private initiatives and individual scholars who copied the books for their own use. 46 In contrast to copies by son said that his father wrote two pages (shnei dappim) before his death; see Marcus, Sefer Hasidim and the Ashkenazic Book, 18. 44 Marcus emphasizes the point that Judah himself produced all the existing versions of Sefer H . asidim, even though sometimes he also refers to his disciples or his followers who may have compiled them; see Marcus hired scribes, user-produced copies written by individual scholars show many deliberate interventions with the text and redactional activities of learned copyists who revised and reformed texts in light of their personal interests. We find this aggressive style of redaction in the Ashkenazic copyists' treatment of the hekhalot corpus as well as in many other works associated with H . asidei Ashkenaz circles. 47 It is also a characteristic of the Ashkenazic piyyut commentary tradition. 48 Collecting and compiling works that way, however, is not solely an Ashkenazi phenomenon, since the same manner of compilation is also found in the early works of the Kabbalists in Provence and Spain, as in the creation of the Zohar, the central kabbalistic work. 49 Perhaps this type of book production was a part of the individualistic character of bookmaking in medieval Jewish culture in general rather than to practices in particular geocultural areas.

Conclusion
This examination of the distribution of material within the manuscripts of Sefer H . asidim certainly supports Marcus's idea concerning the work's compositional process. The various manuscripts represent fluid textual versions of Sefer H . asidim, and it is not possible to reconstruct any definitive original Urtext. Instead, it appears that the manuscripts show the various stages of compilation of Sefer H . asidim's textual units from indiscriminately collected material (sometimes in single folios) to smaller textual units arranged in topical order. In the following stage, more material was inserted peau à peau until finally flowering into the collection of paragraphs in Parma 3280 some eighty years after Judah he-H . asid's death. All these stages are equally represented in the manuscripts. However, not all of them follow the structure of SHP. Given that the material in the larger compilations of Sefer H . asidim is structured according to either SHB or SHP, it is difficult to agree with Marcus's suggestion that these compilations be completely ignored and that each and every manuscript should be viewed as a parallel authoritative version of Sefer H . asidim. The foregoing analysis leads to the more likely assumption of a hierarchical order in the production of the teachings of the H . asidei Ashkenaz titled Sefer H . asidim that involved a process of growth and expansion of relevant material into compilations topically arranged by individual copyists and editors. To fully reconstruct the history of this intriguing book, future research should be dedicated to the life and growth of the topical clusters and inquire into the reasons copyists felt compelled to add or sometimes also remove material in Sefer H . asidim. 50