This article examines the issue of book form, particularly the interplay of text and image, by considering two categories of printed books that dominated the Japanese commercial publishing industry during the mid-nineteenth century. The first type is the yomihon (reading book), a genre with a long and varied history. By the early nineteenth century, it had evolved into the form for which it is currently recognized: a comparatively expensive book format noted for its text-heavy layout; its kanji-laden written style that combined elements gleaned from the Japanese and Chinese classics; its historical settings that focused on the martial accomplishments of warriors; and its high production values, which displayed the talents of the most accomplished commercial illustrators, calligraphers, and printers of the day (Hamada Reference Keisuke1953, 245; Shirane Reference Shirane2002, 885–86). The second type is the gôkan (assembled volumes), a book format that emerged at the beginning of the nineteenth century. Gôkan are cheaply produced publications distinguished by the prominence of their illustrations, which dominated the layout of each page, and the subordinate position of the written text, which filled in the blank spaces that surrounded the images. Another noteworthy feature of the gôkan was the nearly exclusive use of the phonetic kana syllabary, the writing system that served as the minimal basis for functional literacy during the Edo period (Hamada Reference Keisuke1953, 240–41; Konta Reference Yôzô2009, 181–85).

To focus my argument, I will look specifically at Kyokutei Bakin's (1767–1848) yomihon epic Nansô Satomi hakkenden (Chronicle of eight dogs of the Satomi clan, 1814–42), generally recognized as the most prominent and representative example of the genre, and its gôkan knock-off, Setsubai kôtan: Inu no sôshi (Fragrant tale of plum blossoms in the snow: Storybook of dogs, 1848–81). In order to illuminate the cultural work performed by these two texts both on their own and in relation to each other, I selectively apply theories of adaptation, remediation, translation, and bibliography. These theoretical frames help elucidate how features of the two texts, including elements of book design, page layout, choice of writing system, and distribution of visual effects, constitute different perspectives on the book as a medium of signification and as an aesthetic object, thus providing their respective audiences with distinctive approaches for comprehending and appreciating Bakin's swashbuckling tale. Moreover, this integrated framework highlights how these two texts, despite their almost identical content, are predicated upon radically different notions of the relationship of book users to late Edo modes of writing and book culture. In short, I argue that Japanese publishers offered their clientele different visions for the functionality of texts and their constituent elements in order to expand markets and create demand for new products. This practice demarcated a difference between the yomihon, which presented its aficionados with a smoothly integrated matrix that encouraged the operations of reading and seeing on both sides of the writing-image border, and the gôkan, which exemplified a more user-friendly and rational separation of the linguistic and visual elements of the book.

First, before delving into these issues, let me provide some background information on the two texts under consideration. Hakkenden recounts the adventures of eight warriors who came together to assist Satomi Yoshinari defend his domain of Awa from an unholy alliance of rival warlords. The spiritual progeny of Yoshinari's virtuous sister, Fusehime, and her faithful dog, Yatsufusa, the eight “dog knights” possessed superhuman abilities and unparalleled righteousness, which they revealed in an ongoing series of adventures. The duration of its serialization (twenty-eight years) and its sheer size (106 volumes) alone are suggestive of the text's exceptional status. Further testifying to its impact are the many kabuki productions, woodblock prints, pornographic publications, and digest versions that it inspired (Hayashi Reference Yoshikazu1965, 123–26, 131–50).

Central among the cultural byproducts of Hakkenden was the gôkan reworking of the tale, Inu no sôshi, masterminded by commercial publisher Tsutaya Kichizô, in conjunction with writer/transcriber Ryûtei Tanehiko II and illustrator Utagawa Kunisada II. This transformation to the gôkan format involved simplification of diction, the excision of almost all logographs, strategic pruning of references to classical Chinese and Japanese texts, and extensive addition of illustrations, all while rigorously reproducing the content of Bakin's tale, even adhering whenever possible to the wording and sentence structure of the original.

With its multiple writing systems, its array of cultural and linguistic models, and its rigidly segmented population with access to different educational and cultural resources, it is no wonder that nineteenth-century Japan would support the production of multiple versions of Bakin's famous tale. As Peter Kornicki (Reference Kornicki1998, 30–38) and Richard Rubinger (Reference Rubinger2007, 2–3) point out, one should not attempt to apply a single standard of literacy to Edo society. It is far more accurate to acknowledge that reading and writing practices took many forms: texts could be written in pure Chinese, Sinified Japanese, various forms of classical Japanese, or colloquial Japanese; they could be rendered exclusively with logographs, a mix of logographs and the kana syllabary, or mostly in kana; and they could be set down in elaborate calligraphy or simpler more standardized script. Further segmenting the reading public was the spectrum of cultural literacies. By the end of the nineteenth century, the Japanese cultural landscape consisted of varied traditions, including the prose and poetry of the Japanese imperial court and the medieval warrior elite, the sacred texts of Buddhism and Shintoism, the “floating world” of Edo-period urban commoners, and a multitude of indigenous legends and myths, not to mention the vast catalog of historical, philosophical, moral, and literary writings imported from China.

Another factor contributing to the production Inu no sôshi was the Edo-period propensity for recycling preexisting literary material. Hakkenden itself openly declared its derivation from three primary sources: the fourteenth-century Ming epic Shuihu zhuan (Water margin); the Chinese legend of Panhu, about a dog that marries a princess; and military histories of the Satomi Clan (Takada Reference Mamoru1980, 25–26; Walley Reference Walley2009, 81–105). Indeed, Bakin and most of his contemporaries would have comfortably assented to Roland Barthes's (Reference Barthes1977, 160) assertion that every literary text is “a stereophony of echoes, citations, and references.” By the last years of the Edo period, however, this practice had taken a form that from a modern perspective would constitute blatant plagiarism, as publishers repackaged recent compositions with only the slightest modification to the original content. Takagi Gen (Reference Gen1995, 425) explains that this trend was a direct result of the Tenpô Reforms (1842), which weakened the publishing guilds and prevented them from enforcing previously recognized copyright regulations. Publishers saw the reproduction of Bakin's yomihon in gôkan form as a particularly profitable enterprise. Hattori Hitoshi (Reference Hitoshi1991, 197–205) documents the publication of at least forty gôkan reproductions of yomihon from the early 1840s to the late 1860s. Given the common assessment that it constituted Bakin's most significant professional accomplishment, Hakkenden occupied a special place in this publishing trend.Footnote ¹

Because the conversion from yomihon to gôkan exhibits features commonly associated with adaptation, remediation, translation, and bibliographic modification, scholarship on these practices provides a useful theoretical vantage point from which to explore the relationship between Hakkenden and Inu no sôshi. These allied interpretive approaches are helpful because they underscore that practices of textual transformation, which were previously viewed as derivative and subordinate to the act of composition, are more usefully regarded as productive modes of reconstituting already existing source material. In short, advocates of these scholarly approaches argue that adaptation, remediation, translation, and bibliographic modification are meaningful in the most basic sense, since they generate new meaning through their enactment. Yet each approach offers a unique perspective that uncovers a particular aspect of this publishing phenomenon.

Since Inu no sôshi amounts to a strategically modified remake of Hakkenden, it conforms in many ways to the current understanding of adaptation. In her study of this practice, Linda Hutcheon (Reference Hutcheon2006) highlights some features of adaptation that are particularly relevant to my examination. First, she offers this succinct statement on the process: “Adaptation is repetition, but repetition without replication” (7). In other words, she emphasizes that every act of adaptation navigates a course between the poles of fidelity and innovation. Hutcheon also proposes that adaptation is both “a creative and an interpretive act of appropriation/salvaging” (8). This second insight suggests that creative decisions involved in producing an adaptation can justifiably be seen as implicit or explicit commentary on the source material.

In its capacity as an intermediary text that introduces new audiences to Hakkenden, Inu no sôshi can also be viewed as a remediation. In their landmark study on this topic, Jay David Bolter and Richard Grusin (Reference Bolter and Grusin1999, 5) argue that remediation functions in accordance to the dual logic of immediacy and hypermediacy. That is, a remediation strives to deliver transparent access to the original object as it simultaneously calls attention to its function as a mediating interface. This dynamic can be summarized as striking a balance between looking through and looking at (41). Bolter and Grusin also point out that remediated forms typically justify their existence by touting some kind of benefit for the mediating intervention (68–73). These benefits can include, among other things, clarity, accessibility, affordability, and portability.

With its complete transformation of Hakkenden’s written script from a mix of kanji and kana to almost pure kana, as well as the conversion of linguistic material into visual material, Inu no sôshi also exhibits qualities associated with a translation. In accordance with most conventional understandings of the act of translation, the transformations implemented by Inu no sôshi facilitate accessibility and communication. But as Naoki Sakai (Reference Sakai1997) points out, there is another side to the mechanism of translation. As he cogently argues, the regime of translation also reifies difference and inaccessibility (11–17). That is to say, the need, or the perceived need, for a translation not only asserts the difference between imagined national, ethnic, or cultural collectives, but also signals that specific linguistic or cultural products are unavailable to certain communities without the interposition of a translator.

And finally, the nature of the transition from Hakkenden to Inu no sôshi, which principally involves changes in styles of graphic inscription, page layout, and book size, almost demands the application of a bibliographic studies approach. Jerome McGann (Reference McGann1991, 77) articulates the cornerstone of this methodology with his statement: “Every literary work that descends to us operates through the deployment of a double helix of perceptual codes: the linguistic codes, on one hand, and the bibliographic codes on the other.” Bibliographic features of texts, in other words, are as meaning constitutive as language and they play an equal role in shaping the interpretative horizon of a text (24). Consequently, transforming these bibliographic codes has a profound impact on how a text is understood and by whom.

Taken altogether, these theoretical lenses offer a series of interlocking insights into the Hakkenden–Inu no sôshi dyad. Adaptation theory suggests that embedded in the creative decisions about what to change and what to retain from its source material one can detect a complex mix of reverence for and rivalry with Hakkenden and the literary culture associated with it. Remediation theory highlights Inu no sôshi’s implicit and explicit commentary on its own status and function as a mediating intervention. Translation theory illuminates the tension between the comprehensible and the incomprehensible that suffuses the interplay between the two texts and implicates the reading practices and levels of cultural literacy attributed to their two respective audiences. And finally, bibliographic studies underscores the role that material textuality plays in determining the cultural meaning generated by these publications.

I will begin my comparative analysis of Hakkenden and Inu no sôshi with a discussion of the different writing systems utilized in the two publications. At this point, a concrete comparison of parallel sections from the two texts will be helpful. I have chosen a passage from the thrilling rooftop battle between two of the story's heroes: Inuzuka Shino and Inukai Kenpachi. During the late Edo period, this was the most celebrated episode of the entire tale, as indicated by its frequent reproduction in woodblock prints (Hayashi Reference Yoshikazu1965, 15–78). The circumstances leading up to the epic battle are complicated, but to summarize: Shino's deceased father has entrusted him with the heirloom sword, Murasame, and charged him with the responsibility of returning it to its rightful owner, Ashikaga Nariuji. Prior to his audience with Nariuji, Shino discovers that Murasame has been replaced with a forgery. The enraged Nariuji sets his warriors upon Shino, who fends them off and flees to the rooftop of Hôryûkaku tower. Nariuji releases the renowned martial artist Kenpachi, who is currently incarcerated for refusing to obey corrupt domain policies, and orders him to capture Shino. Here is a snippet from Bakin's description of the contest, which appeared in chapter 31 of Hakkenden (1820):

And here is Tanehiko's version of the same passage, which appeared in installment 10 of Inu no sôshi (1850).

As this comparison reveals, Tanehiko's version retains all the fundamental details: the relative positioning of the two warriors, their countervailing actions, and the familiar terminology of single-hand combat. Tanehiko also replicates the basic grammatical structure of Bakin's passage, using a corresponding pattern of conjunctive inflections to connect a chain of alternating clauses. And perhaps most importantly, by switching point of view back and forth between the two redoubtable champions, he mimics the rhythmical structure of alternating couplets that Bakin employed to such great effect in the original scene (Maeda Reference Ai1969, 70).

This fidelity to Hakkenden obviously was key to the commercial viability of Inu no sôshi, since the promise of access to the original was one of the primary justifications for its publication. Indeed, throughout his prefaces to each installment, Tanehiko repeatedly acknowledges Bakin's yomihon as his “source text” (gensho) and emphasizes his status as a mere “transcriber” (shôrokusha) endeavoring to duplicate as faithfully as possible Bakin's masterpiece. But he also concedes that the process of producing Inu no sôshi necessitated fundamental modifications. Foremost among these is the shift from a mix of logographic and phonetic symbols in Hakkenden to the almost exclusive use of kana in Inu no sôshi. Another key alteration is the elimination of punctuation marks. Both these changes accord with the publication standards of the gôkan. In addition, Tanehiko confesses to selective editing of the source text. This is essential, he explains, for two reasons. First: “When converting a yomihon into a gôkan, it is often necessary to cut out elegant written passages to make room for trivial illustrations” (Ryûtei Tanehiko Reference Tanehiko1848–81, 8:1 recto). Second: “Even though it is a real shame to cut out passages from the original, gôkan fans are different from yomihon readers. Transcribing the highfalutin oratory of a warrior into kana script would be difficult to understand and would just bore my young readers” (6:1 recto).

Comparison of the Hakkenden and Inu no sôshi passages, then, would seem to corroborate the perception, acknowledged by Tanehiko himself with his references to “trivial pictures” crowding out “elegant written passages” and comments about the juvenile sensibility of his audience, that gôkan adaptations offered dumbed-down versions of yomihon that were commensurate with the limited reading ability and cultural literacy of their audience. But there is more to this publishing phenomenon than a reassertion of familiar hierarchies of literacy. Consideration of this conversion process reveals the coexistence of competing assumptions about the aesthetic value and signifying properties of the linguistic and visual components of books. In particular, these two formats materialized for book users different approaches toward navigating the word-image matrix of late Edo-period publications. Here it is important to unpack the multiple messages conveyed in Tanehiko's statements quoted above. His comments clearly signal reverence for Hakkenden and respect for its audience. But at the same time, they acknowledge that the gôkan audience constitutes a recognizable community and promises that Inu no sôshi is tailored to satisfy their expectations. As I will demonstrate, Tanehiko's assumptions about how his target audience interacted with the visual and linguistic elements of gôkan anticipate patterns of book consumption that would prevail in the modern period. Here I am indebted to the insight of Charles Inouye (Reference Inouye and Jones1996, 152), who astutely noted that despite its pictocentric format, the “configuration of illustration and text that we encounter in gôkan is the result of a logocentric realignment of signs that continued during the Meiji, Taishô, and Shôwa periods,” I will supplement this observation with a thick analysis of the signifying features of Inu no sôshi. Moreover, I will complicate Inouye's (150) view of yomihon, which he categorized as another key logocentric precursor to the modern book, by highlighting the multiple registers of visuality that characterized the form.

First, I will consider how the signifying properties of the written text are materialized in Hakkenden, paying particular attention to the manipulation of the visual qualities of kanji. A quick perusal of the left side of figure 1, which reproduces the first page of chapter 31, would seem to confirm that the text of Hakkenden, with its combination of logographic and phonetic symbols, adheres to the modern model of written Japanese. Michael Emmerich (Reference Emmerich2013, 179–80), for example, calls attention to the resemblance of the written text of Hakkenden to that of Torioi Omatsu kaijô shinwa (A new tale of seaside Omatsu, 1878), a foundational text in the history of modern Japanese vernacular literature. Contributing to this impression, no doubt, is the yomihon convention of providing kana glosses for almost all of the logographs in the text. If we take these glosses to function primarily as reading aids in the form of pronunciation guides, then it would be natural to extrapolate from this that Bakin and his publication team operated under the assumption that on a fundamental level kanji and kana were interchangeable sets of written symbols for recording the Japanese language. That is, they treated kanji as equivalent to kana and therefore viewed them as an alternative means for designating the same phonetic values indicated by the kana glosses.

Figure 1. Hakkenden 4, book 1. Edo: Sanseidô, 1820. 6 verso, 7 recto. Courtesy of Waseda University Library.

Chieko Ariga (Reference Ariga1989) offers another model for interpreting the relation between kanji and kana glosses in Edo texts. She points out: “In these [Edo] works the dictionary standard equation of kanji vis-à-vis rubi is often broken and displaced to achieve certain literary effects. The tension created by the gap between the kanji and rubi contexts creates a more complex semantic space, rendering the reading process more intriguing” (321). Using the modern term rubi to refer to kana glosses, Ariga goes on to examine three texts that best exemplify this process, among them Hakkenden. Building on Ariga's argument, I will focus on how this technique allows Bakin to exploit the capacity of kanji to convey meaning visually.

Bakin's creative approach toward combining logographs and glosses can be seen in the passage that immediately precedes the battle between Shino and Kenpachi. The scene opens with a discussion of the vagaries of fate, employing the metaphor of a woven thread to symbolize how good and bad fortune are inevitably intertwined. The narrative goes on to show how this cosmic principle applies to Shino and Kenpachi. Here is a truncated version of Bakin's text:

On the phonetic level of the glosses, this passage recounts the sequence of events culminating in the duel between the two warriors. Simultaneously it builds upon the opening metaphor of the twined rope. It conveys this second layer of meaning primarily through strategic implementation of logographs that contain the thread radical 糸いと: katami 紀 (commemorative), koto 縡 (circumstances), tsunagareshi 繋 (bondage), imashime 縛 (imprisonment), and nawa 索 of nawa toku (to loosen fetters). In the majority of these cases, the association with the rope image is not conveyed on the phonetic level of the kana glosses, but only on the visual level of the kanji. Bakin thus extends the metaphor of the karmic thread through the visual properties of the logographs, while simultaneously employing the phonetic aspect of the writing system to relay the substance of the story.

This manipulation of kanji forms exhibits the principle of visuality that Kôjin Karatani (Reference Karatani1993, 51–54) identified in certain writing systems and literary practices that flourished in premodern Japan. Bakin capitalizes on the visuality of writing throughout the text. One of the most prominent examples of this technique can be seen in the proper name, Chudai (丶大ちゅだい), the spiritual father of the dog knights who guides them toward their preordained heroic mission. The name consists of the symbol 丶, typically used to indicate the repetition of a katakana syllable, and the kanji dai 大, meaning “large.” When arranged vertically, with the chu (丶) on top of the dai (大), as they would be seen on the printed page of Hakkenden, the two kanji graphically reproduce the appearance of the logograph inu (犬), signifying “dog” (Takada Reference Mamoru1980, 12). This technique cleverly inscribes Chudai's relationship to his dog knight disciples into the very structure of his name. Significantly, this trope is achieved solely through the visual quality of the two kanji, with absolutely no recourse to their phonetic or semantic value.

Given the almost exclusive use of the phonetic syllabary in gôkan, Tanehiko did not have the same opportunities to manipulate the visual potential of kanji. That is not to suggest, however, that logographs are entirely absent from the text. There are a few simple kanji scattered about in the written narrative. But the most conspicuous use of kanji can be found in the illustrations, where they play an important role in helping the viewer/reader interpret the material. In figure 2, for example, which reproduces Kunisada's image of Shino and Kenpachi plummeting off the roof of Hôryukaku, we can see kanji integrated into the patterns of their clothing, almost like family crests.

Figure 2. Inu no sôshi, 10, book 2. Edo: Kôeidô, 1850. 13 verso, 14 recto. Courtesy of Waseda University Library.

Here it should be noted that it was the practice among the publishers of gôkan adaptations of yomihon to slightly modify the names of central characters. I focus on the manner in which these changes were implemented, because it reveals an understanding of the linguistic function of kanji that deviates from the principles at work in yomihon. In Hakkenden, Inuzuka Shino Moritaka is referred to almost exclusively by his middle name, rendered in kanji as 信乃, which appropriately enough given his virtuous nature conveys the meaning “sincere one.” To capitalize on the public's affection for this revered hero, Tanehiko also chose to refer to the character by the familiar moniker of Shino. In the written narrative, the name Shino is denoted by the phonetic symbols that stand for this pronunciation: し (shi) and の (no). In the illustrations, however, the figure of Shino is consistently identified by the logograph 篠 (shino), which signifies a variety of bamboo grass.

Tanehiko thus selected a kanji, which aside from its linkage with the phonetic unit shino, has little relevance to the identity or personal history of Shino as detailed in Hakkenden. By replacing the compound 信乃 (sincere one), Tanehiko lost an easy symbolic shorthand to underscore Shino's moral superiority. Moreover, the choice unexpectedly highlights a connection between Shino and his aunt, Kamesasa 亀篠, a character who personifies wickedness incarnate. Always attentive to establishing the moral affiliations of his characters by strategic repetition of kanji in their names, Bakin would never have allowed his noble hero to be tainted by this kind of graphic doubling. Because her name is written in kana throughout the text of Inu no sôshi, even in the illustrations, the logographic link between Kamesasa and Shino is effectively erased, paving the way for the use of 篠 to denote Shino in the images. In this context, the kanji still conveys the meaning of bamboo grass, but this conventional semantic value, along with its Hakkenden-specific graphic connotation, is clearly superseded by its phonetic function.

I am not suggesting that this view of kanji is more simplistic than the one at work in yomihon texts like Hakkenden. Rather, I would argue that it denotes an alternative mode for understanding the relationship of kanji to kana and the most effective use of these symbols to convey meaning. It certainly matches the perspective, expressed by Tanehiko in his preface to installment 12: “For those parts of the original text that I don't cut, if I just transcribe them into kana, the phraseology of the yomihon would be hard to understand, so I put them into a more vernacular style that is closer to what you hear every day” (Ryûtei Tanehiko Reference Tanehiko1848–81, 12:1-recto). There is a lot going on in this passage. Tanehiko makes clear that his transcription process involves an element of translation, since he sometimes transforms the phraseology of the yomihon into a vernacular style. He promotes this as a service for his readers, whom he declares would have trouble deciphering the complex literary language of Hakkenden even if were set down entirely in easy-to-read kana. He also assumes that the most comprehensible form of writing for his audience is one that resembles spoken language. In short, for Tanehiko the written text of Inu no sôshi primarily amounts to a graphic system for transcribing spoken language. Hence the emphasis on the phonetic value of written forms, even logographs.

Closely paralleling these competing views of the visual versus the phonetic properties of the Japanese writing system is a similar constellation of issues revolving around the interplay of the faculties of seeing and reading in relation to written matter. In his discussion of Edo-period language, writing systems, and literary culture, Sakai (Reference Sakai1991, 116–18) argues against the tendency to radically separate the perceptual operations of reading and seeing. He suggests that this separation is particularly misleading in the context of East Asia, a cultural sphere dominated by the logographic writing systems of Chinese characters and known for its veneration of the calligraphic tradition. With specific reference to calligraphy, he makes the point that these compositions are simultaneously visual and linguistic. I contend that one can apply this point more widely to a broad range of materials, including much printed matter produced during the Edo period. If we consider seeing and reading as two points along a continuum of interpretative possibilities, then many examples of Edo printed material sought to engage their audience at some middle ground between these two points. Depending on the particular composition, some material encouraged an interpretative mechanism closer to the reading end of the spectrum, while other material encouraged something closer to seeing. But in very few cases would one interpretive mechanism operate to the exclusion of the other. There was, however, considerable variation in the way that different examples of printed matter integrated these two perceptual modes.

All this would seem to point to the conclusion that the production of Inu no sôshi essentially amounted to a translation from the more reading-centered format of the yomihon to the more seeing-centered format of the gôkan. Seeming to support this conclusion is the fact that the designation “yomihon” literally means “reading book”; whereas the common alternative appellation for gôkan was “e-zôshi,” or “picture book,” which clearly denotes that images superseded written text in the signifying hierarchy of the gôkan. Further contributing to this conclusion is Hattori's (Reference Hitoshi1992, 144–46) careful documentation of the production process of these volumes in his study of gôkan adaptations of yomihon. According to Hattori, transcribers of these adaptations began the process of designing each page with a sketch of the central illustration. Only after the details of the illustration had been finalized did they turn to the process of transcribing, translating, and adapting the text of the yomihon original. Spatial considerations thus played a primary role in determining how to transpose the written text of the yomihon to the gôkan format. Not only did the transcriber need to strike a balance between maintaining fidelity to the original text and adhering to the conventions of gôkan publications, but he also had to employ various strategies, including selective cutting, changes in vocabulary, and modification of sentence structure, to ensure that the transposed text would fit into the open space available around the illustration.

Comparison of the two book formats would also seem to verify an initial assessment that Hakkenden is more conducive to reading, whereas Inu no sôshi is more conducive to seeing. Starting with Hakkenden, consider chapter 31, the first page of which is reproduced on the left side of figure 1. The chapter is twenty-four pages long. Twenty-two of those pages consist entirely of written text. As an example of the hanshibon (half-sheet book) format, which measures 22 by 15 centimeters, the book is comparatively large for an Edo publication, with ample white space serving as a backdrop for the orderly lines of written text carefully segmented by punctuation marks (Suzuki Toshio Reference Toshio1980, 38–41). Like most other chapters in Hakkenden, chapter 31 includes one illustration, which takes up two pages. Yanagawa Shigenobu's image, reproduced in figure 3, depicts Shino and Kenpachi fighting on the rooftop of Hôryûkaku, along with an image of Bungobei, the faithful retainer who eventually rescues them after they fall from the tower into a boat and are carried down the Tone River. It is worth noting that even the illustration includes a significant amount of written text.

Figure 3. Hakkenden 4, book 1. Edo: Bunkeidô, 1841. 9 verso, 10 recto. I have taken the image from the 1841 edition of Hakkenden in accordance with Shinoda Jun'ichi's (Reference Jun'ichi2008) assertion that Bakin had the original 1820 illustration modified to better realize his vision. Courtesy of Waseda University Library.

Returning to figure 2, one can see a representative page from Inu no sôshi. The playfulness of the genre is apparent in the clever layout, which requires the reader/observer to rotate the book to gain the proper perspective. More generally, one can discern the standard formal features of the gôkan page design, which consists of a centrally located illustration surrounded by unpunctuated streams of kana arrayed across the remainder of the space. Like most gôkan, Inu no sôshi belongs to the category of the kohon (small book), with dimensions of 14 by 10 centimeters (Suzuki Toshio Reference Toshio1980, 38–41). The small size placed great demands on the book designers who endeavored to squeeze as much visual and linguistic material onto each page as possible.

Despite the substantial evidence in support of the argument that reading was the primary mode for experiencing Hakkenden as compared to seeing for Inu no sôshi, I would like to suggest that transformation from one book form to another was not quite so simple. It is important to keep in mind that visuality can manifest itself in manifold ways. For example, one can distinguish between representational and nonrepresentational modes of visuality. With representational visual material, the pattern of the composition essentially resembles what it stands for. The illustrations in both Hakkenden and Inu no sôshi conform to this representational visual mode. But nonrepresentational visuality is also at play in these texts, particularly Hakkenden. The nonrepresentational visual features of a book can encompass, among other things, abstract decorative elements on the page, manipulation of white space, and implementation of different script and calligraphic styles. Returning to Sakai's (Reference Sakai1991, 116) discussion of calligraphy, he wonders: “Is calligraphy a text to see or is it a drawing to read? If it is both, then how should we understand the kind of seeing that is also reading?” This observation is particularly relevant for a high-quality publication like Hakkenden, which employs various nonrepresentational visual strategies to enhance the viewing experience of the product.

As indicated previously, the production team for this publication involved the finest professional craftsmen available, including the calligrapher Chikata Nakamichi, who copied out the written text. His calligraphic manipulation of the written symbols imbued them with aesthetic value, but did not alter their meaning. Nakamichi was given ample opportunities to display his craft, given the array of elaborate, often obscure kanji and highly articulated forms of kana that Bakin favored in the composition of the main narrative, not to mention the variety of calligraphic styles on display throughout each installment. For example, observe some of the calligraphic forms arrayed in book 1 of volume 4, which includes chapter 31. The right side of figure 4 reproduces the title page, upon which the title and volume number are monumentally rendered in crisp standard script (kaisho), with the names of Bakin and Shigenobu copied out in more attenuated variations of the same style. Enriching the textual palette is the use of seal script (shôten), with its stylized form and archaic associations, for the notation “dog general” (inu shôgun) that appears atop the whimsical illustration. The left side of figure 4 shows the first page of the author's Chinese (kanbun) preface set down in feathery, meticulously ordered running script (gyôsho). On the right side of figure 1 is a pair of verses, one in Chinese and one in Japanese, composed by Bakin under the nom de plume Raisai. The larger, widely spaced characters are rendered in a more fluid variation of running script with cursive script (sôsho) embellishments. For the main text, Nakamichi favors a graceful form of standard script, occasionally incorporating features of running script to augment the visual flow. And finally, the written inscription on the upper left side of figure 3, above Bungobei, with its delicate transcription of kana symbols, suggests the beauty of cursive script. Reader/viewers would not only have valued the variation and juxtaposition of these writing styles throughout the volume, but the more knowledgeable among them would also have recognized the visual references to the hand of past master calligraphers.

Figure 4. Hakkenden 4, book 1 (1820). Flyleaf, 1 recto. Courtesy of Waseda University Library.

In addition to the styles of calligraphy, the choice of kana symbols in the main text also affects the visual impact of the page. During the Edo period, there were multiple options for writing most kana syllables. For example, there were at least five standardized alternatives for the syllable ki, each derived from a different kanji. Perusal of the main text of the Hakkenden narrative reveals considerable variation in the selection of kana options. These kana choices seem to have no discernable relation to semantic and syntactic value. That is, there is no correlation between certain kana forms and specific grammatical patterns or vocabulary items. Rather, the seemingly random utilization of a variety of different kana symbols for the same phonetic unit appears to be motivated primarily by a desire to add visual interest to the page. Undoubtedly, in all the different sections of the text, kana and kanji perform an essential communicatory function as writing. But one should not underestimate the degree to which the artistic manipulation of these forms on the page enhanced the value of Hakkenden as a material object. Indeed, for connoisseurs of the late Edo period, these visually achieved aesthetic considerations played a central role in their experience of books (Kornicki Reference Kornicki1998, 26–30).

Sakai's insight about the fuzzy border between reading and seeing can be applied as profitably to the illustrations of Hakkenden as to its written text. In their studies of the tale, Takada Mamoru (Reference Mamoru1980, 86–116) and Shinoda Jun'ichi (Reference Jun'ichi2004, Reference Jun'ichi2008) have underscored this point with their detailed analyses of the many coded messages concealed in the Hakkenden images. Consider, for instance, the right panel of figure 3, Shigenobu's rendering of Shino and Kenpachi engaged in hand-to-hand combat. Shinoda (Reference Jun'ichi2008) persuasively decodes the complex visual language of the illustration, suggesting that the rooftop functions as a visual analogy (mitate) for the prow of a boat. This enables the image to simultaneously depict two points of time during the conflict: the climax of the rooftop battle when Shino shatters his sword, and the denouement when the two warriors plunge into the prow of a boat anchored on the bank of the Tone River. The visual analogy is achieved, Shinoda asserts, by the use of dark blue shading in the background to illustrate the river water, the white spots in the air to indicate the spume that erupts when the warriors land in the boat, and the insertion of Bungobei's fishing pole into the visual space occupied by Shino and Kenpachi. Another layer of meaning is transmitted through the abstracted design of clouds that frame the upper third of the two picture panels. These clouds, Tomichi Akihiro (Reference Akihiron.d.) maintains, foreshadow Bakin's fanciful comparison of the tussle to a celestial contest between dragons. Drawing on the conventional association of dragons with clouds, the border design visually reinforces Bakin's textual metaphor, and, Tomichi suggests, is part of an extended network of visual and verbal allusions to a Chinese-based cosmology of dragons and compass directions that is embedded throughout the text.

A different strategy of visual signification is legible in figure 5, which reproduces a frontispiece illustration from volume 2, picturing Shino on the left and Inukawa Sôsuke Yoshitô, another dog knight, on the right. Careful observation reveals that the design of Sôsuke's apparel, which is adorned with images of a hatchet (yoki), a Japanese harp (koto), and a chrysanthemum (kiku), together function as a rebus conveying the message “yoki koto kiku,” or “hear some good news” (Shinoda Reference Jun'ichi2004, 39–47). On the most basic level, this visual pun anticipates an episode in chapter 20 when a mystical message from on high informs Shino and Sôsuke of the exalted nature of their relationship.

Figure 5. Hakkenden 2, book 1. Edo: Sanseidô, 1817. 5 verso, 6 recto. Courtesy of Waseda University Library.

Together these examples demonstrate Bakin's tendency to understand the illustrations in Hakkenden as objects to be seen and to be read. Bakin addresses this issue in his afterword to book 2 of volume 2, when he exhorts his audience to pay particular attention to the illustrations:

Like most of Bakin's ostensible interpretive aids, this statement is quite ambiguous. It refers to an illustration that pictures Daisuke, who later comes to be known as Chudai, as he fatefully approaches the lair of Fusehime and Yatsufusa. Setting aside the specific application of this statement to the image of Daisuke, it offers broader insight into Bakin's view of the complementary signifying functions of linguistic text and image. With his statement, Bakin suggests that one cannot accurately construe the meaning of the narrative without careful examination of both the written inscriptions within the images and the nonlinguistic elements of the illustrations. In short, one must learn to read the images as well as the written text (Takada Reference Mamoru1980, 93).

This suggests that signification of the Hakkenden exceeds the surface meaning of the written narrative. Underscoring this point is another statement by Bakin gleaned from his so-called “Seven Rules for Yomihon,” which appears in an “Authorial Remarks” section located at the beginning of chapter 104. Derived from Yuan and Ming literary treatises, this statement professes to encapsulate the core elements of Bakin's literary aesthetic. Chief among them is inbi (impenetrability or opaqueness), about which he opines: “Inbi manifests itself in the deep meaning that lies outside of the writer's words. One might have to wait one hundred years for a wise man to recognize it” (Kyokutei Bakin Reference Bakin1820, 6:8). Appropriately enough, there is room to debate exactly what Bakin insinuates with this enigmatic statement, but it leaves little doubt that he sees his text as much more than a straightforward, one-dimensional recitation of a story.

The exciting adventures of the dog warriors and the complementary moral message, which are straightforwardly communicated on the phonetic level of the narrative through the systematic application of kana glosses, are important, even essential, to the text, but signification does not end there. Meaning is conveyed on multiple levels, in varied registers, and through different symbolic systems. The adroit interpreter, Bakin (1:5) haughtily implies, must integrate these different levels of signification and aesthetic effect to fully appreciate the material in its totality. Otherwise, the text will never transcend the level of a “frivolous diversion” (moteasobimono) suitable only for the lower echelons of cultural consumers.

As one would expect, Inu no sôshi mobilizes its signifying effects differently. This makes sense when one considers the way in which most people are said to have experienced the text. Nozaki Sabun, a Meiji-era journalist and disciple of gôkan producer Kanagaki Robun, offered this summation of the order of signification in late Edo picture books: “As a picture-centered form, the main goal of the kusazôshi was to catch the readers’ eye and arouse their interest with the illustrations. It is accurate to say that the written text explained the content of the pictures” (Nozaki Reference Sabun1927, 145). Referring to gôkan as kusazôshi, Nozaki lays out a pattern of consumption in which the written text played a subordinate role to the illustrations.

Indeed, the visual spectacle of Inu no sôshi is undeniable. The compositional organization in figure 2 replicates the signifying hierarchy of the book, with the dynamic image occupying pride of place and the written text functioning as a supplemental backdrop. Visuality was even more prominent in the sumptuous polychrome cover illustrations, which showcased the remarkable technical expertise achieved by print designers and engravers in the twilight years of the Edo period and were a major selling point of these publications (Suzuki Jûzô Reference Jûzô1961, 26). See, for example, the cover of the first book of installment 1, reproduced in figure 6, which vividly portrays Shino atop Hôryûkaku assuming a fierce pose. His hair, the patterns of his clothing, and the architectural elements of the rooftop are all rendered in exquisite detail. The expert application of different gradations of colored ink in the background sky, the pattern in Shino's hakama trousers, and the cartouche bearing the title of the tale also speak to the skill of the print makers.

Figure 6. Inu no sôshi 1, book 1 (1848). Cover. Courtesy of Waseda University Library.

Further enhancing the visual display of the book were the occasional illustrations in which Kunisada incorporated the likeness of popular kabuki actors into his depiction of characters in the story. Fans of the kabuki theater must have appreciated figure 7, which pictures Shino in the manner of Ichikawa Danjûrô VIII, an actor noted for his performance of heroic roles (Kimura Reference Yaeko2009, 171–72). This technique adds a level of complexity to an otherwise straightforward image of Shino, by overlapping two reference points: the content of the story and the extra-textual world of the kabuki theater. Taken together, these pictorial elements offered a stimulating visual experience in keeping with the flamboyant aesthetics that prevailed in late Edo urban culture. At the same time, it should be noted that for all their brio and technical virtuosity, the images primarily achieve their effects representationally.

Figure 7. Inu no sôshi 10, book 1 (1850). 9 verso. Courtesy of Waseda University Library.

In stark contrast is the written text of Inu no sôshi, which not only pales in comparison to the pictorial elements of the book but also lacks the visual flair of the writing in its yomihon counterpart. Transcribed exclusively in standard script, the most legible of the calligraphic styles, the kana is remarkably uniform, with little of the variety or elaboration found in Hakkenden. There is a tendency in Inu no sôshi to consistently utilize the same kana symbols. And throughout the text relies on the simplest and most standardized forms of kana. Moreover, given the small size of the book itself and the limited space on each page, practical concerns of cramming as much written content as possible onto the page seem to take precedence over any kind of aesthetic presentation of the writing. In fact, the overall quality of the calligraphy is often quite low. In Inu no sôshi, writing is thus limited to performing the communicatory function of conveying the maximum amount of information efficiently and clearly. This approach accords with Tanehiko's professed efforts to make the text as reader-friendly as possible for his target audience. The written text of Inu no sôshi provides its users with an easy-to-read recapitulation of the content of Bakin's renowned tale. This written text complements the images, which strive to reproduce key moments from the source material pictorially. Ultimately, this tendency to displace visual appeal from the written text and largely reduce it to a phonetic representational system results in a more pronounced separation of the visual and linguistic fields of the text than one finds in yomihon. Despite their close proximity on the page, the format of the gôkan asserts a division between the signifying functions of the illustrations and the written commentary. In short, gôkan present themselves as books consisting of pictures to be seen and text to be read.

What, then, do all of these changes suggest about the way that gôkan producers differentiated the reading practices and literary tastes of their target audience from those more predisposed toward yomihon? Clearly, the producers of Inu no sôshi and other gôkan versions of yomihon assumed an audience less versed in kanji literacy than the typical yomihon reader. More significantly, changes in layout, writing systems, illustration design, and calligraphic style inevitably resulted in a different engagement with the text. In contrast to the yomihon, which conspicuously encouraged a perceptual mode that integrated elements of reading and seeing vis-à-vis both the illustrations and the linguistic text, the gôkan displaced visuality from writing and transferred it to the illustrations. This resulted in a more streamlined and rational signifying matrix in which illustrations performed one function and writing another. Moreover, it concretized writing, reinforcing its close connection to spoken language and its utility as a transparent medium of communication. This approach toward writing as an accessible content delivery system deviates markedly from the high value placed on obliqueness, or inbi, in the yomihon source text and suggests a radically different understanding of the pleasures inherent in the consumption of commercially printed material.

This final insight productively corroborates Inouye's (Reference Inouye and Jones1996) assertion that despite their pictocentric format gôkan embodied key elements of Japanese literary modernity. Inouye bases this assertion primarily on the sheer volume of written text incorporated into the gôkan format (151). I would add that the view of language observed in Inu no sôshi, which places a premium on the clarity, accessibility, presence, and connection to the spoken word, are other features that anticipate the theories of modern Japanese literary language advocated by Tsubouchi Shôyô, Futabatei Shimei, and Yamada Bimyô in the 1880s and 1890s. In stark contrast is the yomihon. With its attention to the visuality of writing and the readability of illustrations, it materializes a vision of the book that deviates profoundly from the literary ideals pursued by these later reformers. It is little wonder, then, that Bakin served as the anti-modern bête noire in Shôyô’s treatise, Shôsetsu shinzui (Theory of the novel, 1885–86), despite Shôyô’s personal predilection for Bakin's tales.

I would like to conclude with the consideration of another, less immediately obvious way that the Hakkenden–Inu no sôshi dyad fits into a discussion of modern book culture. Roger Chartier's magisterial history of book production and readership in France provides a conceptual basis from which to proceed. In discussing the goals of his study, he states:

I will start by considering two of the main points that Chartier makes in this statement and relating them to my own findings. First, in his capacity as an adherent to the school of bibliographic studies, he asserts that the formal features of books generate meaning and exert a profound influence in shaping the social significance of texts. Second, he argues that not only do books cater to already-established audiences and patterns of reading, but they also engender them through their own production and circulation. The Hakkenden–Inu no sôshi dyad exemplifies both of these points. As I have demonstrated, the formal features of the two books play a key role in defining the aesthetic sensibility and cultural significance associated with each text. This is especially the case with Inu no sôshi, the composition of which essentially amounts to adapting the content of Bakin's original narrative through the application of textual features associated with the genre of gôkan. This process is clearly motivated by a desire to cater to an audience who were already acclimated to the gôkan format. But in accordance with Chartier's second point, this publishing phenomenon also involved the construction of a new audience and pattern of use. For with the production of Inu no sôshi, publisher Kichizô added a state-of-the-art twist to the gôkan formula. He transformed the gôkan into a textual interface that justified its existence through dutiful repetition of a preexisting text and the promise of mediated, yet simultaneously unfettered, access to a prized cultural narrative. By creating a new function for the gôkan, Kichizô, along with other innovative publishers, thus encouraged a new mode of reception for the form—gôkan as yomihon redux.Footnote ²

And this brings me to a third and final point that Chartier introduces in the preceding paragraph: difference, an issue that resonates with Sakai's theory of translation. Here Chartier elaborates on the topic:

According to Chartier, the modern order of books is defined not only by rising literacy rates, increased circulation of books, and expansion of readership, but also by the introduction of new systems of differentiation to organize and stratify book culture. In the context of my study, this impulse manifests itself in the division of an expanding population of Hakkenden consumers into multiple categories through the creation of new mechanisms for circulating the story. The motivations behind this imperative are, to say the least, complex. As Chartier suggests, it partly resides in a desire to reinforce traditional cultural distinctions. One can detect this impetus in Bakin's imperious differentiation between informed and uninformed classes of yomihon enthusiasts. But it also bears remembering that much of the effort to mark difference occurs not in Hakkenden but in Inu no sôshi. In his prefaces, Tanehiko constantly reminds his audience that the gôkan version deviates from the standards of the yomihon original. These statements are partly couched as expressions of humble reverence for the lofty source text, but they also clearly convey the idea that Inu no sôshi offers its audience a necessary, and in many cases preferable, alternative. This suggests that the desire for difference worked both ways, since Inu no sôshi encouraged and enabled gôkan readers to lay claim to their own version of the Hakkenden story.

The Hakkenden–Inu no sôshi dyad thus exemplifies a key feature that Chartier attributes to modern book culture, in that it promotes the expansion of book consumption and ever-widening circulation of key cultural narratives through carefully calibrated reproduction and repetition as it participates in increasingly complex and multidirectional strategies for reifying difference and stratification in the way that people experience books.