The Chinese Typewriter

by Thomas Mullaney

  MingKwai may have looked like a conventional typewriter, yet its behavior would have quickly confounded anyone who sat down to try it out. Upon depressing one of the machine’s seventy-two keys, the machine’s internal gears would move, and yet nothing would appear on the paper’s surface—not right away, at least. Depressing a second key, the gears would move again, yet still without any output on the page. With this second keystroke, however, something curious would happen: eight Chinese characters would appear, not on the printed paper, but in a special viewfinder built into the machine’s chassis. Only with the depression of a third key—specifically one of the machine’s eight number keys—would a Chinese graph finally be imprinted on the page.

  Three keystrokes, one impression. What in the world was going on?

  What is more, the Chinese graph that appeared on the page would bear no direct, one-to-one relationship with any of the symbols on the keys depressed during the three-part sequence. What kind of typewriter was this, that looked so uncannily like the real thing, and yet behaved so strangely?

  If the fundamental and unspoken assumption of Western-style typewriting was the assumption of correspondence—that the depression of a key would result in the impression of the corresponding symbol upon the typewritten page—MingKwai was something altogether different. Uncanny in its resemblance to a standard Remington- or Olivetti-style device, MingKwai was not a typewriter in this conventional sense, but a device designed primarily for the retrieval of Chinese characters. The inscription of these characters, while of course necessary, was nonetheless secondary. The depression of keys did not result in the inscription of corresponding symbols, according to the classic what-you-type-is-what-you-get convention, but instead served as steps in the process of finding one’s desired Chinese characters from within the machine’s mechanical hard drive, and then inscribing them on the page.

  The machine worked as follows. Seated before the device, an operator would see seventy-two keys divided into three banks: upper keys, lower keys, and eight number keys (figure 6.2). First, the depression of one of the thirty-six upper keys triggered movement and rotation of the machine’s internal gears and type complex—a mechanical array of Chinese character graphs contained inside the machine’s chassis, out of view of the typist. The depression of a second key—one of the twenty-eight lower keys—initiated a second round of shifts and repositionings within the machine, now bringing a cluster of eight Chinese characters into view within a small window on the machine—a viewfinder Lin Yutang called his “Magic Eye.”6 Depending upon which of these characters one wanted—one through eight—the operator then depressed one of the number keys to complete the selection process and imprint the desired character on the page.

  6.2 Keyboard of the MingKwai Chinese typewriter

  In creating the MingKwai typewriter, then, Lin Yutang had not only invented a machine that departed from the likes of Remington and Underwood, but so too from the approaches to Chinese typewriting put forth by Zhou Houkun, Shu Zhendong, Qi Xuan, and Robert McKean Jones. Lin invented a machine, indeed, that altered the very act of mechanical inscription itself by transforming inscription into a process of searching. The MingKwai Chinese typewriter combined “search” and “writing” for arguably the first time in history, anticipating a human-computer interaction now referred to as input, or shuru in Chinese.

  Kangxi Must Die: The “Character Retrieval Problem” in Republican China and the Origins of Input

  Deferring for a moment a more detailed examination of the MingKwai’s internal mechanics—of how a three-key sequence resulted in the printing of a Chinese character—most important for us at the outset is to understand the unique historical genealogy of MingKwai as compared to other typewriters in our story—for it is in this history that we uncover the origins of input or shuru. When Lin Yutang set out to design his typewriter, he did not draw his inspiration from movable type, telegraphy, or even Western typewriting, but from a language reform debate known as the “character retrieval problem” (jianzifa wenti) taking place in China from the 1910s through the 1930s. In this debate, Lin Yutang alongside a host of Chinese library scientists, educators, and linguists fought over and experimented with novel systems of organizing Chinese dictionaries, library card catalogs, indexes, name lists, and telephone books—systems that would help Chinese speakers navigate the Chinese-language information environment more efficiently. Of all those who participated in the “character retrieval problem” debate, however, Lin would be the only one to transpose this rigorous discussion of retrieval to the domain of inscription. To understand MingKwai in particular, and input more broadly, then, we must first delve into the “character retrieval” debates of the early Republican period, and the broader history of China’s early twentieth-century information crisis.

  In Too Much to Know, Ann Blair uncovers many of the ways in which present-day notions of “information society” and “information overload” have a much deeper history than we often assume. Well before the advent of computing and the Internet, purveyors of knowledge and truth in early modern Europe spoke of their age in terms of overwhelming information overload. An array of bibliographers, publishers, and others developed novel approaches and technologies designed to help gain control over the burgeoning information environment that constantly threatened to burst its containers and leave humanity in the ironic condition of over-informed ignorance. In the East Asian context, the work of Mary Elizabeth Berry has revealed parallel trends in early modern Japan. Tokugawan cartographers, bibliographers, compilers, editors, and others deployed a repertoire of repurposed and novel technologies—including maps, digests, encyclopedias, and travel guides—to help them stay afloat in an ever-rising sea of information.7

  Late imperial and early Republican China was yet another site of “information crisis.” With the population boom of the late imperial period, the rise of new forms of state surveillance, and the introduction of novel information technologies such as telegraphy, Chinese elites were growing increasingly anxious about their capacity to keep pace with this new data environment. The history of “information overload” in early twentieth-century China stands apart from the European and Japanese contexts explored by Blair and Berry, however. In China, what vexed elites most were not primarily questions of information quantity or velocity, nor even the country’s engagement with new informational genres, but debates over the capacity of the Chinese language itself (or “incapacity,” as many argued) to handle “modern information.”

  Circa the 1920s, banal essentials of the modern information environment became a persistent source of anxiety: how to develop a modern Chinese telephone directory, magazine index, archival index, name list, and indeed all forms of reference materials where one might need to locate some form of information encoded in Chinese characters. In one study from the era, it was found that research subjects using one of the many new experimental character retrieval systems were able to locate Chinese characters between one-tenth of a second and one full second faster than when using China’s leading dictionary.8 Still others determined that Chinese character look-up was consistently slower than its alphabetic counterparts.

  While the difference of mere seconds hardly seems to justify the sounding of alarms, it appeared possible to some language reformers that tiny delays such as these might be significant factors within China’s broader challenges during the modern age. If information encoded in Chinese took slightly longer to find than in alphabetic look-up, what might this mean when aggregated across the wider textual landscape: indexes, phone books, name lists, concordances, passenger manifests, encyclopedias, commodity inventories, and library card catalogs? It stood to reason that the Chinese-language corpus as a whole was in a very real sense further removed from the average Chinese user by thousands of additional minutes, hundreds of additional hours, even added days, as compared to the English-language corpus and its respective user base.9 Aggregated across the Chinese populace as a whole, it would seem that China’s bac
kwardness might indeed be the macrohistorical extrapolation of these countless microhistorical lag times. China was functioning in slow motion.10 As one language reformer claimed, a more advanced system of organizing and retrieving Chinese characters would save every literate Chinese person a full two years in his or her lifetime (assuming a working career of forty years) simply by helping them find information faster.11

  Propelled by the growing sense of information crisis, the early Republican period witnessed an explosion of experimental organizational methods for Chinese characters. The crisis would come to be known as the “character retrieval problem” (jianzifa wenti), and would attract participation by such luminaries as long-time president of Peking University Cai Yuanpei, one-time head of Shanghai Commercial Press Gao Mengdan, pioneering Chinese library scientist Du Dingyou, and dozens of others.12 By its peak in the early 1930s this “crisis” would give rise to no fewer than seventy-two experimental systems by which to reorganize the Chinese-language information environment—seventy-two new “alphabetic orders” for Chinese script. Sprouting like “bamboo shoots after the rain,” as one observer put it, these systems were varied and diverse—and yet they cohered around a growing consensus. The Kangxi Dictionary and its attendant radical-stroke taxonomic system were unworkable for the purposes of modern information systems.13 If Chinese-language information were to be every bit as modern as alphabetic information, Kangxi would have to be overthrown.

  In Search of the Dao

  Contenders for the throne of Kangxi entered the arena in droves. In 1912, Gao Mengdan—successor to Zhang Yuanji as the chief editor at Shanghai Commercial Press—proposed his Merged Radicals Method (Guibing bushou fa) premised upon a sharp reduction in the number of Chinese radicals.14 In 1922, Huang Xisheng proposed his Chinese Character Retrieval and Stacking Method (Hanzi jianzi he paidie fa). Soon after, library scientist Du Dingyou proposed his Chinese Character Retrieval Method (Hanzi jianfa) and Chinese Character Arrangement Method (Hanzi paizi fa). More well known was the Numerical Retrieval System (Haoma jianzifa), later reformulated as Sijiao haoma (or “Four-Corner Code”), by Wang Yunwu. It was within the context of this early twentieth-century “character retrieval problem” that Lin Yutang began to develop and propose his own systems. In 1918, he published his Index System for Chinese Characters (Hanzi suoyin zhi),15 and in 1926 reentered the debate with two new retrieval systems: the New Rhyme-Based Indexing System (Xin yun suoyin fa) and the Last-Stroke Character Retrieval Method (Mobi jianzifa).16

  The sheer number and diversity of experimental retrieval systems, all created within years of one another, raised a startling question: Even at this late date in history, had the fundamental essence and order of Chinese script yet to reveal itself? Unlike English and French, in which the sequence of alphabetic letters had long ago been established and stabilized, could it be that the Chinese script remained a wide-open terrain of uncertainty, possibility, and even revelation as late as the third decade of the twentieth century? How was it possible for so many of the leading minds of China to disagree so fundamentally with one another over the Chinese script? Moreover, which of these seventy-plus experimental systems was the true order, if any? Which was most efficient? Which would achieve dominance? Which offered the solution to China’s information crisis?

  One observer, Jiang Yiqian, took this proliferation of experimental systems as unmistakable evidence that the “fundamental method” (genben fangfa) of Chinese had yet to be discovered—and what is more, that Chinese script was clearly more complex than its alphabetic counterparts. “Chinese writing,” Jiang wrote, “is not alphabetic, and does not have a set order. As a result, within the history of Chinese culture there have been all sorts of different character retrieval systems.”17 Drawing upon Darwinian metaphors, Jiang portrayed all of the many competing and divergent character retrieval systems as akin to species that, in their interaction, competition, survival, and extinction, together constituted a churning, collective, evolutionary process that would eventually reveal the essential truth of Chinese script.18 Among the dozens of systems that had been proposed, he argued,

  Table 6.1 Retrieval systems invented between 1912 and 1927 (partial list)

  the question of which is best awaits comparison and corroboration, as well as a long period of use. For the time being, it’s impossible to say—and, in fact, there’s no need to do so. Over the course of a substantial period of time, and by process of elimination, the best one will naturally rise to the top. … The best method will also necessarily be the fundamental method of Chinese characters. Since Chinese characters possess a structure, so too then must they possess a way (dao) by which to organize them. Phrased differently, since there must be a system, so too must there be a fundamental method of arrangement.19

  When discovered, the “fundamental method” would be to the Chinese language what the alphabet was to English: the unambiguous, rational, transparent system by which the Chinese language would finally realize its true order, and in which the swirl of novel experimental systems would finally achieve quietude—its own end of history.

  Jiang Yiqian may have been satisfied to await the victor of this evolutionary struggle, but others were less patient. For Lin Yutang, Wang Yunwu, Chen Duxiu, Du Dingyou, and others, the character retrieval crisis was both a professional and deeply personal contest. To the victor went the spoils, not only in economic terms, but perhaps primarily in terms of symbolic capital. Each strove to promote his system within and among government departments, such as the ministries of education and communication. Each strove to develop relationships with China’s centers of print capitalism, such as Commercial Press. Above all, however, each participant in the “character retrieval problem” competition harbored a dream that his system would be the one to depose Kangxi and to accede to the throne.

  All of this raises an obvious question: As Lin Yutang and his contemporaries set out to uncover the “fundamental method” of Chinese script, where did they believe it would be found? When Lin and other participants in the “character retrieval problem” designed their symbolic systems, what governed this process? How, moreover, would they know if this fundamental order had been discovered?

  As we examine early Republican debates over character retrieval, we find two primary forces or motifs that shaped this contest: a pursuit to discover the fundamental, orthographic essence of Chinese script, and a parallel pursuit to discover the fundamental cognitive or psychological essence of Chinese people. In the first of these pursuits the question was, in a sense: How do Chinese characters fundamentally want to be found? What was the essential nature of the Chinese character, and its fundamental building blocks? If Russian, Hebrew, Greek, English, and Arabic all enjoyed fundamental, agreed-upon, unambiguous orders, what was the Chinese equivalent? In the second of these pursuits, the question focused on people rather than script: How do Chinese people fundamentally want to search? In their efforts to develop a “transparent” system—one that could be used by “anyone and everyone”—Republican-era language reformers ventured into a site of political contestation: namely, the battlefield in which political elites of different stripes vied to define the Chinese “masses” themselves. As part of trying to create systems that “everyone” could use, Republican-era language reforms were thus equally invested in defining who the “Chinese everyone” was—in terms of capacities, limitations, tendencies, and instincts. In setting forth their respective retrieval systems, each system constituted a competing theorization of both Chinese script and the “average Chinese user.”

  How Ancient China Missed the Point: Character Retrieval and the Essence of Chinese Script

  To begin our discussion with the first of these poles—character retrieval as the search for the dao of Chinese script—an illustrative case is that of Chen Lifu (1900–2001). Throughout the Republican period (1911–1949), and into the postwar era, Chen Lifu was deeply involved in the political management of culture and education. In the late 1920s, he was elected a membe
r of the Guomindang Central Executive Committee, and was later named Guomindang secretary general in 1929. He went on to become president of the Central Politics College, a member of the Standing Committee of the Central Executive Committee, director of the Military Affairs Department’s Bureau of Statistics, and the minister of education, among many other posts.

  Chen Lifu also invented a character retrieval system known as the “Five-Stroke” or Wubi retrieval system. Five-Stroke was born on the battlefields of 1920s China, when the country was entering its second decade of political fragmentation. Vast territories of the former Qing empire were now governed by military elites who offered only paper allegiance, or none at all, to the central government of a defunct republic. The Guangzhou-based Nationalists and their Communist allies launched a northward military campaign whose goal was to defeat or co-opt militarists and reunite the country under a single capital and regime.

  Chen Lifu’s Five-Stroke system was fashioned and field-tested during this campaign. Charged with managing confidential materials, including telegrams and papers, Chen and his team processed approximately 150 documents daily, always under an acute sense of time-sensitivity and urgency. “Time was short and Chiang was always impatient,” Chen recalled in his memoir, referring to Nationalist leader Chiang Kai-shek. “Whenever he wanted a certain paper, he required us to find it for him immediately. This challenge forced me to conceive a classification method for Chinese characters to organize our files.”20 Specifically, Chen developed Five-Stroke retrieval as a way to categorize the Nationalist military’s growing collection of enemy combatant telegraph code books, presumably organizing them according to the Chinese character surname of the enemy combatant officer, or perhaps the name of their regime or alliance.21 The Guomindang faced a multitude of military actors, each employing different numerical transformations to encrypt messages transmitted using the four-digit telegraph code we examined in chapter 2. Upon interception, decryption was time-sensitive and depended not only on identifying the code being used, but also on quickly finding the code books themselves from within the army’s library.