Maciej Patryk Kurzynski / Department of East Asian Languages and Cultures, Stanford University
Abstract: The growing scholarship on the sublime in non-Western contexts makes it necessary to reconsider the possibility of this peculiar experience from a broader cross-cultural perspective. The point of convergence among the many existing interpretations is a diachronic pattern that can be disassembled into three distinct components: rising motion, boundary, and sequentiality. Given that digital humanities provide tools to detect patterns and repetitions in texts, this paper employs methods of computational criticism to explore the aesthetic of the sublime in modern Chinese narratives. Combining word embedding, topic modeling, and network analysis, I aim to shed light on what I call the “technology of the sublime,” a narrative mechanism that synchronizes plot development with vocabulary distribution in the novel. The first part of this article introduces the computational theory of the sublime to encapsulate the process whereby a steady accumulation of words and expressions describing large and powerful natural phenomena culminates in a boundary-crossing experience narrated in a novel’s plot. In the second part, I read two modern Chinese novels—Second Sun by Liu Baiyu (1987) and Soul Mountain by Gao Xingjian (1990)—and reveal how both authors avail themselves of the narrative mechanism thus defined. The discovered similarity is noteworthy given the ostensibly divergent aesthetics and antagonistic ideals conveyed by the two texts. Finally, I show the ways in which writers negotiate with the sublime meta-narratively to contain and redirect its powerful emotive thrust.
Keywords: Sublime; Chinese Modern Literature; Network Analysis; Topic Modeling; Sensorimotor Schemas
The sublime has proven a useful interpretative category in the studies of modern Chinese culture, to such an extent that it seems almost impossible today to disentangle it from the analysis of the Chinese nation-state and the ways China imagines itself. Rather than a “Western concept,” Ban Wang thus understands the sublime more broadly as a “discursive practice” and “a set of articulatory maneuvers or gestures” provoked by great historical moments requiring new ways of representation.As he notes, in post-1949 China the aesthetic of the sublime became particularly prominent, furnishing a gigantic image of the People “engaged in a world-transforming practice in order to carry out the telos of history.”Judith Shapiro similarly shows how the dictum “Man Must Conquer Nature” ( 人定胜天 ) underlay the collective endeavor of great proportions to reconstruct the physical world in Mao’s China.But both before and after Mao, Chinese artists tapped into the potential of grand-scale imagery to represent the nation-state. John Crespi, for example, identifies “vocal crescendo” in Gao Lan’s ( 高 兰 , pen name of Guo Dehao 郭德浩 ) recitation poems that would function as “an aural detonator for his combustible compatriots” and awaken them to national crisis during the anti-Japanese war,while Tie Xiao demonstrates how Chinese art catered to the modern alienated individual’s “longing for ‘oceanic’ experiences” by imagining the crowds.Writing about much more recent times, Andy Rodekohr uses the military term “human wave tactics” ( 人海战术 ) to encapsulate the guiding thread of Zhang Yimou’s artistic output, discerning in his orchestrations of the human multitude distant echoes of the “sea of red” ( 红海洋 ) revolutionary imagery.In yet another context, Haiyan Lee employs the notion of “military sublime” to re-watch the 1985 film The Big Parade (《 大阅兵》) and examine how the official iconography of Tiananmen Square constructs the charisma of the state’s political power.
Such growing scholarship on the sublime in the Chinese context makes it necessary to theorize the possibility of this peculiar experience from a broader cross-cultural perspective.At the same time, we should be careful not to “throw the baby out with the bathwater” and dismiss the manifold insights that the Western philosophical tradition still has to offer: the analytical approach to aesthetic experience and, more importantly, the very decision to consider the sublime as an aesthetic category in its own right. Situated between these two demands, this article contributes to the ongoing debate about the possibility of the sublime by combining disparate fields of inquiry: modern Chinese literature, Western aesthetics, cognitive literary studies, and digital humanities.
The Technology of the Sublime
Sublime narrative as a dynamic metaphor of sequential ascent
The starting point of my inquiry and the point of convergence among the many existing interpretations of the sublime is the sensorimotor schema of sequential ascent beyond a boundary. Schemas are spatial and force-dynamic patterns developed in early childhood that underlie more complex conceptual structures, such as metaphors or abstractions. For example, emotions are often said to reside “in one’s heart,” which involves the primary schema “container” and the conceptual metaphor “heart is a container.” Likewise, to say that a melody is “moving somewhere” is to use the schema “motion” and the metaphor “musical succession is physical motion.”While philosophers frequently employ spatial metaphors of “elevation” or “rise” as well as “boundaries” or “barriers” to make sense of the experience that they call “sublime” — according to one etymology, the Latin word sublimus literally means “up to the lintel” — an equally prominent feature in their discussions is the sequential nature of such experience. For Immanuel Kant, the sublime takes place whenever “the effort at comprehension […] exceeds the capacity of the imagination to comprehend the progressive apprehension in one whole of intuition.”Edmund Burke writes about the “succession and uniformity of parts” and “uninterrupted progression, which alone can stamp on bounded objects the character of infinity,”while Moses Mendelssohn defines the sublime as an experience in which “a single impression is repeated without alteration, uniformly, and frequently,” so that “the boundaries of extension are deferred further and further” until they “disappear completely from the senses and, as a result, something sensuously immense emerges.”In fact, long before the birth of aesthetics as a separate category of inquiry in Western philosophy, Longinus in his treatise On the Sublime already pointed to “dense composition” as a rhetorical effect whereby “one great phrase after another is wheeled into place with increasing force.”By accumulating narrative features in a progressive series, the orator “belabors the minds” of listeners with “blow after blow” and “preserves the essence of his repetitions and asyndeta through continual variation.”
All the above definitions hint at the crucial characteristics of the sublime and can thus serve as primary references for a literary study. Whereas the schemas of “rising motion” and “boundary” underlie developments within the plot (e.g., the hero becomes stronger, or an obstacle is overcome), the feature of sequentiality, by contrast, should be measurable — albeit only approximately, given the lack of rhyme and metric repetition in modern prose —on the formal level of the novel, i.e., in the vocabulary. In other words, if stories meant to provoke sublime experiences are narrated by novelists as complex dynamic metaphors mapped upon the schema of sequential ascent beyond a boundary, then I suggest that we should be able to discover regularities in the very language of the narrative.In the second half of this article, I will demonstrate that this is indeed the case. Before moving to specific cases, however, I want to show more generally why and how the computational methods of literary inquiry might change the way we think about the sublime as a rhetorical effect.
Vocabulary of the sublime
In modern China, the ancient parable of “The Foolish Old Man Who Removed the Mountains” ( 愚公移山 ) and the story of “Great Yu Controlling the Waters” ( 大禹治水 ) have been appropriated to promote the aesthetic of hero worship and collective voluntarism.But there is yet another text that proved equally if not more stimulating to modern Chinese imagination: Maxim Gorky’s famous poem The Song of the Stormy Petrel (Песня о Буревестнике/Pesnya o Burevestnike/《海燕之歌》). The eponymous “stormy petrel” has since become a triumphant harbinger of revolution and an embodiment of idealist fervor, traversing cultural landscapes not only in the Soviet Union but also in the PRC and other socialist states.
The word “tempest” or “storm” (буря/burya/ 暴风雨 ) is the keyword of the poem, as demonstrated by the last three lines (notice the repetition):
— Буря! Скоро грянет буря!
Это смелый Буревестник гордо реет между молний над ревущим гневно морем; то кричит пророк победы:
— Пусть сильнее грянет буря!
—Tempest! Soon will strike the tempest!
That is the courageous Petrel proudly soaring in the lightning over the sea’s roar of fury; cries of victory the prophet:
—Let the tempest come strike harder!
Given that Gorky entangles here concepts from two distinct semantic domains — the ideologically-charged “victory” ( 胜 利 ) and “prophet” ( 预言家 ) on the one hand, the powerful natural phenomena like “tempest” ( 暴风雨 ) and “ocean” ( 大海 ) on the other — within one narrative continuum, his poem provides a perfect example of how a deliberate distribution of two heterogeneous vocabularies can rhetorically enhance a literary work. In what follows, my goal will be to generalize the structure of Gorky’s poem so as to discover similar semantic entanglements in much longer novelistic texts.
In order to find words and phrases synonymous with terms such as “tempest,” “fierce,” “thunder,” and “sea” in modern literary Chinese, I use the word-to-vector (word2vec) algorithm, which learns word associations from a large corpus of texts according to the distributional hypothesis that words occurring in similar contexts tend to have similar meanings. Once trained, the model can detect semantically related words such that one can perform mathematical operations on them (e.g. “King – Man +Woman = Queen”). I use a publicly available word2vec model pretrained on 8599 works of modern Chinese literature, containing 177 million tokens and a vocabulary of 702 thousand words.To ensure a representation of oceanic vocabulary that is large enough but also customizable for my project, I search for words and expressions whose vector representations are closest to the vector corresponding to the word “tempest”(暴风雨) as well as sums of pairs of vectors corresponding to the key terms in Gorky’s poem(暴风雨+乌云,暴风雨+大海,暴风雨+闪电,暴风雨+雷声,暴风雨+猛烈).I select the top 50 synonyms for each of these six queries. This method ensures the presence of “tempest” as the guiding aesthetic idea while allowing for variation within the vocabulary selection process. Having removed duplicates, I further add 150 hand-picked terms and expressions which have not been found by the model but nevertheless constitute the staple of sublime imagery (words such as “mighty” 伟大, “torrent” 激流 , and “surge” 涌流 ). In total, this semi-supervised procedure yields 317 words and phrases that constitute the sublime vocabulary and the building blocks onto which the schema of sequential ascent is mapped by the narrative.
Correlated vocabulary: heterogeneous and yet analogous
Using word embeddings trained on a large corpus of literature, we were able to identify a set of sublime words. The next step is to find words correlated with the sublime vocabulary in the narrative space of a particular novel. What discourses parasitize our visceral reactions to powerful natural phenomena described in a text? In his elaboration of the mathematical sublime, Kant tells us that the recurrence ad infinitum of the “and so on” elements of the construction in intuition (“progressive apprehension”) soon forces the imagination to reach its maximum capacity, so that new elements can be added only at the expense of the elements apprehended earlier. Although our inability to imagine any further is experienced as painful, this greatest liminal magnitude achieved in the realm of sense finds its analogy in the idea of infinity, always already larger than any measurable magnitude, in the realm of Reason, the thinking of which brings pleasure (at least to the Enlightenment mind).Although we cannot imagine infinity, we nevertheless can think it, and we as humans should be proud of it. Kant’s theory needs to be seen in the historical context of Western mathematics: the burgeoning science of calculus (convergence towards a limit) and the then-prevalent illustration of infinity as a step-by-step geometrical construction of a potentially (but never de facto) infinite number of points in Euclidean space.What his philosophical intuition can still offer us even today, however, is the very logic of the sublime experience, whereby a process of sequential accumulation in one domain (imagination) is concluded by a heterogeneous and yet analogous process in another (thought).
One way to operationalize this insight for a literary study is through topic modeling, which is a probabilistic machine learning technique capable of identifying semantic dimensions shared by words in a corpus of texts.In a topic model, every topic is a list of words with associated probabilities indicating how representative a word is for a particular topic. We can use this method to find clusters of words that occur together more frequently than if they were randomly distributed in a text. For example, if a story features a central hero whose actions remain the focus throughout the plot, then a good topic model should be able to detect words and expressions that the author often uses to describe this hero (including her name, her frequent interlocutors, recurrent ideas and places, etc.). To a certain degree of approximation, therefore, we can reduce the novel’s themes to computational topics. This further implies that the tension–resolution dynamic as outlined above can be understood as a steady, homogenous accumulation of the sublime vocabulary which ultimately finds its euphoric conclusion — “boundary” — narrated in the story qua topical vocabulary. As the topical vocabulary (“thought”) evolves parallelly to the sublime (“imagination”), one should consider it analogous, yet since the topical vocabulary does not contain any sublime words, it remains heterogeneous. Such a generalization of Kant’s aesthetic judgment entails that the pleasure inherent in the sublime does not need to originate in our ability to think infinity, nor does the inadequacy of the imagination have to reveal its “vocation” for striving for ideas of Reason, as Kant would have it. Instead, on the present interpretation, the “pleasure of the text” can be induced — if done at the right moment — by any discourse parasitizing a sensory ascending series.
To find the plot vocabulary correlated with the sublime, I first separate the text of a novel into overlapping windows of 500 words (i.e. the first window contains the first five hundred words (no. 1-500), the second window words no. 251 to 750, the third, no. 501 to 1000, etc.). This ensures that longer passages are contained within at least one window.The choice of this window size is dictated also by the fact that writers tend to organize novelistic texts into passages that each focus on an individual topic (as opposed to chapters or whole novels, each of which may represent multiple topics).In the next step, I run a series of twenty LDA topic models (from k=10 to k=200, where k is the number of topics) on the 500-word chunks thus obtained, and for every topic generated by each of the models I take the terms (words and expressions like chengyu) from the top 10% of the probability distribution of the topic over the words of the novel.I then note down their frequency in each of the 500-word windows and compare it with the frequency of the sublime vocabulary in respective windows. The higher the correlation between these two series of frequencies, the stronger the relationship between a particular topic and the sublime.In other words, if it is the case that an increased presence of the sublime vocabulary entails an increased presence of the topical vocabulary, then the topic in question is considered correlated.
Having modeled the text of a novel in this way, I then collect the words belonging to the three most strongly correlated topics in each model, which allows me to capture general plot-related terms recurring together with the sublime throughout a particular novel (from topic models with low k values) as well as more specific expressions related to singular events, if any, that occasioned a sudden increase of the sublime (from models with high k values).Taken together, the selected words constitute the novel-specific vocabulary correlated with the sublime. By measuring how a novel distributes these two groups of words diachronically, we can now see where exactly the flow of intensity increases and identify the most dramatic passages of the narrative.
Topic modeling in conjunction with word embedding can thus determine two sets of vocabulary (the vocabulary of the sublime and the novel-specific vocabulary correlated with it), as well as places in a novel where they appear frequently together. However, this information is still not specific enough to answer the question as to exactly how a particular novel entangles the two vocabularies in a diachronic series to produce the visceral reading experience. Network analysis is thus necessary to further localize the argument that the qualitative shifts of reading experience are precipitated by the narrative when the presence of respective glossaries reaches local peaks. As we remember, Longinus argued that in the sublime “one great phrase after another is wheeled into place with increasing force.”I take “great phrase” literally here: a “great sentence” is a sentence that contains at least one word from the sublime vocabulary determined above (e.g., “Let the tempest come strike harder!”). In this sense, the following two “great sentences” should be considered a sequence, since the second one does not provide qualitatively new information but only quantitatively augments the imagery already provided by the first one:
(1) But he knew this was not the right time for such distractions; he gritted his teeth and forged ahead through the torrent.
(2)The wind was howling, the rain was swirling, and in the flashes of thunder and lightning, the raindrops flickered like countless silver dragons.
The more sentences there are that contain such oceanic imagery, the more palpable the sequential character of a particular fragment, and the more likely the “blow after blow” experience for the reader. That progression over longer spans of narrative time is necessary for the sublime experience to occur might explain why reading a single phrase quoted from a literary work is unlikely to give us even a single goosebump. Not unlike seeing the Pyramids from the right distance, to respond affectively we must read such “great sentences” as embedded within a larger encompassing narrative of ascent.
My second claim is that sublime words are “parasitized” by topical words that appear in their vicinity. The working hypothesis in this article is that the span within which a topical word benefits from the sublime “aura” is two sentences to the left and to the right of any “great sentence.” In other words, I draw a co-occurrence edge between a sublime node and a topical node if the distance between them does not exceed two sentences (in this sense, the word “victory” 胜利 would collocate with the word “tempest” 暴风雨 in Gorky’s poem). The edge distance between nodes is further quantified as inversely proportional to the number of co-occurrences: if two terms appear together twice, the distance between them will be ½, etc. While individual cognitive abilities and reading habits differ widely, as do the lengths of actual sentences in any novel, I use the span of two sentences to approximate the findings from the psychological research on working memory.Building a co-occurrence network in this way resembles the process of novel reading whereby specific information absorbed by working memory from individual words, phrases, and sentences (new nodes and edges) “updates” the long-term gist memory that we have of the novel as a whole (the entire co-occurrence network constructed so far). Another way of interpreting this process is to say that, while topical words (i.e., the correlated vocabulary) “acquire” sublimity from the sublime words in their neighborhood, the sublime nodes simultaneously acquire local, novel-specific information from the topical terms that surround them. Some of these topical terms are further elevated through the sublime terms in subsequent sentences, while the sublime terms become increasingly localized in the topical space of the novel as they reappear later in the text, and on and on. The result is a process of accelerating “oscillation” or “vibration” : as the aesthetic distance grows towards the limit due to the increasing number of terms and connections, the semantic distance between the sublime and the encompassing discourse is decreasing.If the author provides enough “great sentences” — that is, if the semantic distance becomes sufficiently small — and then describes some momentous event using the topical vocabulary (some sort of boundary is finally crossed in the plot, e.g., two lovers finally find each other, a formidable mountain peak is reached, or a great mathematical problem solved), the reader is likely to experience the sublime.
The last piece in the puzzle is to find the key nodes in the network that we have just built: which topical and which sublime words are prioritized by a narrative, and which terms from one category are most strongly synchronized with the words from another within the narrative flow? If two terms can be considered synchronized if they reappear frequently together, then the network measure that can capture the synchronization of groups of words is eigenvector centrality, as it determines the level of connectedness of a node: nodes with well-connected neighbors (whose neighbors, in turn, are also well-connected) will have a high eigenvector centrality. Unlike the centrality measures that quantify information flows along the geodesics (such as closeness and betweenness), eigenvector centrality best captures the synergy between two categories of frequently cooccurring vocabulary (sublime and topical): an important sublime node will have a higher eigenvector centrality if it cooccurs frequently with many well-connected topical nodes, and likewise an important topical node will also have a higher eigenvector centrality if it cooccurs frequently with many well-connected sublime nodes. As such, eigenvector centrality captures the diachronic aspect of the narrative and translates, in a static numerical form, the sequential dynamics of the text. Combined with measurements of edge distance (shorter distance spreads the centrality better), eigenvector centrality can thus approximate the synchronization of vocabularies in the novel and determine the representative terms from both categories for any particular passage.
We have thus descended from a macroanalysis of a large corpus to a microanalysis of individual words. The computational techniques employed along the way allow us to reverse engineer what I call “the technology of the sublime”: a narrative mechanism that synchronizes plot development with vocabulary distribution into a sequential ascent beyond a boundary. The said mechanism is a method (techne) of arranging a network of words (logos) within a diachronic narrative series intended to evoke a powerful and transformative aesthetic experience in the reader.Following the methodology elucidated above, I would now like to show this mechanism in action by re-reading two modern Chinese novels: Second Sun (Di er ge taiyang,《第二个太阳》, 1987) by Liu Baiyu(刘白羽)and Soul Mountain (Lingshan,《灵山》, 1990) by Gao Xingjian (高行健 ).From the perspective of traditional literary scholarship, one would be hard-pressed to choose two texts more different from each other than these two: one championing a collective political ideal through the lens of socialist realism, the other focusing on the individual quest for meaning and truth. The computational theory of the sublime as formulated in the foregoing discussion will allow us to see numerous commonalities nonetheless.
Crossing the Yangtze
Second Sun focuses on the relentless struggle of the Liberation Army soldiers in Hubei province as they conquer numerous cities in pursuit of Qin Baijie ( 秦白洁 ), who is the captured daughter of Qin Zhen ( 秦震 ), the Bingtuan Deputy Commander. Baijie also happens to be the beloved one of Qin Zhen’s protégée, the division commander Chen Wenhong ( 陈文洪 ), and hence the military confrontation acquires deep personal significance. As the Liberation Army struggles to conquer Wuhan before the city is destroyed by the retreating Guomindang (Nationalist) forces led by Bai Chongxi ( 白崇禧 ), the soldiers face the turbulent waters of the Yangtze River. Later on, as the war spreads further south, the scorching heat, swarms of mosquitoes, torrential rains, and devastating floods bring tremendous difficulties to the soldiers accustomed to the more temperate, northern climate. The Wuling Mountains and the Yangtze River thus provide a perfect scenic backdrop for the heroic bravery of the Liberation Army, while the South becomes an object of the colonial desire for the Northerners.
Following the method explained above, I generate twenty topic models and then plot the vocabulary distribution over the diachronic axis of the narrative to visualize the emotional dynamics of the text:
Figure 1 Vocabulary distribution in Second Sun, smoothed with a rolling average of three 500-word windows. “Presence” indicates the number of words belonging to a particular vocabulary divided by the total number of words in a given window
The novel contains 7,913 sentences and is here divided into 543 overlapping windows of equal length. From the two graphs in Figure 1, we can see clearly how the topical vocabulary (bottom graph) closely follows the distribution of the sublime vocabulary (upper graph). In the following sections, I will focus on two intense moments of the story: the middle part of the novel, i.e., windows no. 258-280 (47.5%-51.6% of the text, or sentences no. 3672-4004), and the much shorter fragment towards the end of the story, containing windows no. 497-518 (91.9%-93.8% of the text, or sentences no. 7223-7530).
But instead of immediately delving into computational analysis, I would like to begin with one of the most memorable moments in Second Sun, a startling rescue scene at the beginning of the novel (10.8-11%). Chen Wenhong’s dreamlike memoir transports us to the cradle of China’s communist revolution, the little town of Yan’an, crowded with Chinese compatriots sharing strong anti-Japanese sentiments. The focus then shifts to menacing clouds heralding the arrival of a powerful storm, only to be followed by a life-and-death struggle against the formidable torrents of the Yan River as Chen Wenhong attempts to save his future lover, Baijie, from the stormy waters:
He suddenly noticed a white spot in the frenzied torrent.
“There’s someone in the water!”
A human body remained in the grip of turbulent waves. No one could see if the body was still struggling, and no voice seeking help could be heard, since at that moment everything was submerged in the furious screams of nature. From time to time a little white spot floated up to the surface, only to disappear again in the water.
“Someone is drowning there!”
Chen Wenhong had no time to think twice. He jumped from the rock and plunged into the furious rapids. The skies collapsed; the rocks rumbled; whoever bumped against the menacing boulders would be immediately crushed to powder. But look, this one human being, this son of the Earth, was brandishing his arms and struggling against the current. Whenever he saw someone in distress, Chen Wenhong would not hesitate even for a second, he would immediately throw himself into evil waves.
As the strength of the individual is being contrasted here with the overwhelming force of the raging torrent, the young Chen Wenhong becomes yet another “stormy petrel.” The narrative focuses on the confrontation between human freedom and the mountain flood, foregrounding the bravery of Chen Wenhong whose ardor and determination the natural world will never match. Hence starts the river imagery which will accompany us till the very end of the novel, to various degrees of amplification.
Indeed, Chen Wenhong’s initial act of bravery is just a foretaste of what the narrative has to offer in the following chapters. The crossing of the Yangtze by the Liberation Army is the most intense episode of the whole story, both in terms of plot development (it will ultimately decide the fate of the whole military campaign) as well as vocabulary distribution (both sublime and topical words are highly present). Right in the middle of the novel, not just Chen Wenhong — although he does play a prominent role according to the “three prominences” ( 三突出 ) principle of Chinese socialist realism — but also the whole contingent and, by extension, the Liberation Army itself, face the tremendous powers of the untamed nature of China’s south.The river serves here as a boundary which the soldiers will cross only with great exertion:
Two mighty torrents formed between Heaven and Earth:
One was the torrent of Nature, with raging waters and powerful thunders.
One was the torrent of the People, struggling bravely against the waves.
If the former was violent, the latter was fearless. It is exactly these two mighty torrents
that spurred into existence a human life’s most precious character, spirit, power.
Network analysis of this emotional encounter (windows no. 258–280, see Table 1) reveals that words related to powerful weather phenomena such as those from Gorky’s poem (e.g., “tempest” 暴风雨, “gale” 狂风, “downpour” 风雨, etc.) play a vital role also in these dramatic passages. Among topical terms, on the other hand, it is Chen Wenhong who serves as the focus imaginarius of the emotional geography of the text, bringing to the fore the fortitude of the soldiers under his command who march forward ( 前进 , 向前 , 前面 ) and cross the mighty Yangtze.By incessantly repeating the hero’s name (65 times, or once every 5 sentences), which itself hints at the “literary torrents” ( 文洪 ), and by surrounding it with the ever-expanding masses of images and feelings expressed through “great sentences,” Liu Baiyu attempts to synchronize the two dimensions of the narrative —the “syllogistic progression” of the plot and the vocabulary distribution — and transform the human into the superhuman. As though always anxious that his reader’s imagination might go astray and lose focus, Liu Baiyu incessantly repeats the landscape terms such as “wind and rain” ( 风雨 , 14 times) or “mountain flood” ( 山洪 , 16 times), foregrounding the epoch-making significance of the military struggle ( 战斗 ).
The novel’s pedagogy consists in continual accumulation of the sublime terms, which in turn allows it to emphasize the cataclysmic proportions of the battle as the Liberation Army endeavors to cross ( 跋涉 , 涉渡 ) the wrathful river ( 急流 ) while being shelled by the enemy entrenched on the other side. The measurement of edge distances reveals that landscape terms, such as “wind and rain” ( 风雨 ) or “mountain flood” ( 山洪 ), and Chen Wenhong ( 陈文洪 ) become almost inseparable, merging into one powerful image of a sublime hero (see Table 2). By mixing it with a slightly less repetitive yet equally dramatic portrayal of Qin Zhen (whose name Zhen [“quake”] is also evocative of nature’s wrath) and the omnipresent military terms (e.g., “order” 命令 , “radio transmitter” 报话机 , “troops” 队伍 ), Second Sun thus provides a powerful embodiment of the truth that “Man Must Conquer Nature.”
Figure 2 Co-occurrence network visualization of windows no. 258-280 (47.5%-51.6% of the text, or sentences no. 3672-4004) in Second Sun. Size of the node indicates its eigenvector centrality, and edge transparency indicates distance. Nodes with eigenvector centrality lower than 0.04 have been removed for better visibility
Table 1 Influential nodes in the co-occurrence network visualized in Figure 2
Table 2 Edges with shortest distance in the co-occurrence network visualized in Figure 2
Towards the end of the story, Qin Zhen reluctantly leaves the battlefield to attend the grand political events in the capital. But Liu Baiyu does not slow down the pace of his narration, capitalizing instead on the momentum engendered by the warfare episodes. The founding ceremony of the People’s Republic of China on the 1st of October, 1949, is depicted through mesmerizing terms of grandeur:
At three in the afternoon, the grand celebration began. From the Gold Water Bridge in front of Tiananmen all the way to the Arrow Tower, the space between the red-walled and yellow-tiled three-arch gates on west and east sides was packed with people. The crowd looked orderly and solemn, as though everyone was savoring the most precious moment of their lives. Under the eyes of tens of thousands of people, Mao Zedong raised the first five-starred red flag in human history with his very own hands. The bright flag was rising slowly, brilliant, splendid, and wonderful, looking almost like a flame waving in the wind. It was ascending steadily, and all the suffering, noble souls were rising with it together, like the sun bursting with bright rays of life, turning a new epoch-making page in the calendar and henceforth changing the journey of mankind. The majestic, lively, and passionate tune of “The March of the Volunteers” flowed from countless broadcast speakers, so passionate that it could make rivers and seas boil. Everyone thought of the moment when we finally left the abyss of slavery, when “the Chinese nation faced its greatest peril…” Here, countless generations finally converged, marching forward. The sound of the majestic salute made the earth rumble.
The quoted passage forms the pinnacle of the revolutionary tale, narrating a collective experience of awakening that merges the manifold of singular bodies into a nation. Surrounded by the human multitude, the rising flag symbolizes the crossing of the historical boundary between pre- and post-1949 China.
Figure 3 Co-occurrence network visualization of windows no. 497-518 (91.9%-93.8% of the text, sentences no. 7223-7530) in Second Sun. Size of the node indicates its eigenvector centrality, and edge transparency indicates distance. Nodes with eigenvector centrality lower than 0.025 have been removed for better visibility
Table3 Influential nodes in the co-occurrence network visualized in Figure 3
Table4 Edges with shortest distance in the co-occurrence network visualized in Figure 3
Figure 3 is a visual interpretation of this final part of the novel (windows no.497-518). As Mao Zedong oversees the birth of the nation from afar, the revolutionary masses merge in a collective ecstasy ( 高 潮 ) on Tiananmen Square. Like Chen Wenhong in the previous fragment, here it is Qin Zhen who concentrates the continually growing emotional momentum. One is justified, therefore, to argue that Liu Baiyu’s narrative strategy is to enhance the charisma of the central hero through continuous “blow after blow” additions of sublime terms surrounding the character-related words. As Qin Zhen stays the focus of the plot thanks to the frequent reappearance of his name (once every 5.5 sentences), he cuts an increasingly impressive and indomitable figure. Simultaneously, the nationalist discourse embraces the sublime on the square ( 广场 ), absorbing numerous sublime terms such as “to seethe” ( 沸腾 ), “tremor” ( 震响 ), and “ocean” ( 大海 ). A lot of heavy lifting is done through such metaphorical work: the square is likened to a raging sea ( 大海 ), and deafening cheers of exulting crowds to thunderous waves ( 浪涛 ). By unloading the multiplicity of national synecdoche in a limited verbal space and appending the terms in quick succession (what Mendelssohn called “accumulating nouns”), the narrative provides the necessary sensory material for the sublime experience to occur.
The grand finale of Liu Baiyu’s novel thus conveys the enthusiasm surrounding the birth of a nation by addressing the readers’ kinesthetic memories of powerful natural phenomena such as devastating floods and heavy downpours. The parallelization of discourses underlying the novel’s educational project allows the language of nationalism to closely associate itself with the language of nature, which itself has been already heavily charged with militaristic potential through the lengthy depictions of the Yangtze campaign. More importantly, however, the above analysis lends credence to the argument that the sublime imagery has been in fact domesticated within the topical space of the nation, which is perhaps the greatest paradox of revolutionary romanticism. While there is no reason to doubt that such stories find their origin in intense passion, we should also consider them as an affective technology: a deliberate distribution of vocabulary which aims to condition the reader’s emotional responses.
But to do full justice to Liu Baiyu’s narrative skills, we also need to consider the very last page of his magnum opus. Eventually, we learn that Baijie dies in a Guomindang prison, and the story ends on an ambiguous note.However strong Qin Zhen’s desire to resurrect his daughter in the majestic image of the “second sun” (as opposed to the “first sun,” which does not shine for her anymore), his suffering continues to emanate through the novel’s final words. This qualified rejection of Cui Wei-style happy-ending indicates that, although deeply attached to revolutionary morality, the Liu Baiyu of the 1980s does not remain at the mercy of repetition; on the margins of revolutionary pantheism, there remains a modicum of space for unrecuperated sentiment.This does not in the least invalidate the point made above, however. Now that we consider the novel as a whole, we can also see how this final act of transcendence once more manifests Liu Baiyu’s mastery in narrating emotions. The hundreds of pages long furtherance of emotional momentum, from Chen Wenhong’s individual act of heroism to collective ecstasy on Tiananmen Square, is here unexpectedly undermined by an experience of personal loss, only to be redeemed once again through the magnificent figure of the rising sun. But it is only after we acknowledge the mechanistic nature of Liu Baiyu’s text that the nature of this sorrowful and yet sublime coda can be properly appreciated.
Dreamscapes of the Self
The foregoing discussion made a case that the experience of the sublime has a particular spatiotemporal character which can be schematized as a “sequential ascent beyond a boundary.” As such, my argument dovetails with Ling Hon Lam’s recent study The Spatiality of Emotion in Early Modern China: From Dreamscapes to Theatricality, which likewise argues that “rather than a state of mind being imprinted by, overflowing onto, or mingling with things in external space, emotion per se is spatial.”Lam’s interpretation of Chinese cultural history distinguishes three consecutive modes of spatiality: “winds,” “dreamscapes,” and “theatricality.” The ancient topos of “winds” or “airs” (feng,风)interprets emotions as an “affective atmosphere,” an all-pervading field of cosmological forces that traverse across individuals and communities and sometimes even throws them “under the weather.” The emotion-realm called “dreamscape” (mengjing,梦境),on the other hand, retranslates emotion as a temporary illusion, a dream from which one is to be delivered, but only to yet another dream, giving rise to a never-ending back-and-forth — the Zhuangzi-like deliverance (hua,化) — between manifold layers of reality. Finally, the term “theatricality” refers to the early modern sentimental episteme in which the dreamer is transformed into a detached spectator observing a play from a distance: split into “ethereal soul (hun, the lighter half that ascends and wanders away) and humic soul (po, the turbid half that sinks and stays),” the skeptical dreamer is now sidelined as an onlooker to his own dream.
If we now take a closer look at Soul Mountain, Gao Xingjian’s famous novel, we will notice that Lam’s tripartite division of emotion-realms can greatly illuminate a computational literary study, even though I am not following here the chronological and essentially Foucauldian order of Lam’s presentation, but rather employ the three modes of spatiality as interpretive lenses through which to read a modern text that draws extensively upon the traditional Chinese hermeneutics of dreams. Gao Xingjian’s novel describes a journey of a certain “you” traveling to the fringes of the Han civilization in search of the eponymous sacred mountain. Very early in the story, a stroke of serendipity makes it that “you” meets “she,” who grudgingly agrees to join the man’s journey, and hence starts a complicated, erotically charged relationship between their bodies and souls. Simultaneously, the other pronoun, “I,” follows Gao Xingjian’s own quest for authentic life that he embarks upon after his cancer diagnosis proves false and guides us along a more down-to-earth ethnographic search for knowledge and self-understanding. There is no unifying viewpoint in this story, as the figure of the narrator is irretrievably split between “I” (wo,我)and “you”(ni,你): while “I” moves between reminiscences, “you” wanders through dreams.
However paradoxical it might seem at first glance, in his lavish descriptions of individual freedom Gao Xingjian employs the same narrative mechanism that we have already identified in Liu Baiyu’s work. To find the corresponding terms, I follow my previous strategy: I run a series of twenty topic models on the text of the segmented novel and find the topical words correlated with the sublime vocabulary. I then plot the results on the diachronic axis to detect the most dramatic moments in the narrative. The novel contains 7823 sentences and is divided into 510 overlapping windows.
Figure 4 Vocabulary distribution in Soul Mountain, smoothed with a rolling average of three 500-word windows
The first peak of intensity (28.4-29.2% of the text, windows no. 145-149, sentences no. 2154-2208), unequivocally erotic in nature, is a dream narrated from hindsight by “you” to his lover:
You say you had a dream, just now, while you were asleep on her. She says yes, it lasted only a moment, she even spoke to you, you didn’t seem to be fully asleep. She says she touched you while you were dreaming, she could feel you pulsating, for just a minute. You say yes, an instant ago everything was so vivid. You feel the warmth of her breasts, the heaving of her abdomen. She says she held you and felt you pulsating. You say you saw a black sea rising, its flat surface slowly, inexorably, towering up. When it was upon you, the horizon between the sky and the sea was squeezed to nothing and the black sea occupied the whole of your vision. She says you were asleep pressed against her breasts.
Figure 5 Co-occurrence network visualization of windows no. 145-149 (28.5-29.5% of the text, sentences no. 2154-2208) in Soul Mountain. Size of the node indicates its eigenvector centrality, and edge transparency indicates distance. Nodes with eigenvector centrality lower than 0.01 have been removed for better visibility
Table 5 Influential nodes in the co-occurrence network visualized in Figure 5
Table 6 Edges with shortest distance in the co-occurrence network visualized in Figure 5
Although described from the perspective of an awakened hero, the memories of formlessness, of dissolved horizons, and of the skies covered by the surging black sea are still fresh in the narrator’s mind. The act of deliverance is narrated here as an escape from the scene of slaughter full of “black sleek bodies of human-like sea animals or animal-like humans” left all over the shore following the release ( 倾泻 ) of erotic tension and the retreat of a powerful tide ( 浪潮 ). The most intense delight is verging on the macabre, as the tidal imagination enhances sexual ecstasy by juxtaposing the language of lovemaking with a description of the seashore engulfed in darkness.
The word “wave” (浪潮)generates the most of the sublime thrust as it carries the erotic fantasies running wild; woman’s breasts (乳房), intense emotions(感觉,感到), and blood pulsating under the skin (脉搏), all serve here as markers of erotic sensuality, pushing the asymptotic narrative (越来越) towards the powerful climax. As in Second Sun, where the founding ceremony awakens everyone present from the somnolence of Lu Xun’s iron house to the modern nation-state, here it is in the pitch-black ( 黑色 ) space that the sublime takes place, the woman’s breasts providing the form, the re-imagined Square on which the sublime rite is to be reenacted: “You say you felt her breasts swelling, like a black tide, a surging tide, like surging lust, growing higher and higher, wanting to engulf you.” (你说你感到了她乳房鼓涨，像黑色的海潮，而海潮升腾又像涌起的欲望，越来越高涨，要将你吞没) The spatial dynamics is at once perpetuated and dissolved by the female figure as she simultaneously blinds (黑色)the hero and entices him to take a look (看见) (see Table 6).
In the penultimate chapter, when the narrator climbs a formidable ice mountain in the midst of a fierce snowstorm, the text guides us through a similar experience. At the very moment when the deathly loneliness and a tearing pain in the hero’s lungs speak of the ultimate dysphoria and the decision to leave the human world and plunge oneself into absolute darkness seems final, “you” suddenly hears a faint tinkling coming from the depths of his mind. Together with a vaguely visible figure of a woman on a horseback, the persistent references to the throbbing heart again engender a powerful dynamic leading the dreamer towards the powerful apogee. At this moment, the language itself breaks into pieces:
Before your eyes, in your heart, in your body oblivious to time and space, in the continual surge of sustained noise, of reflected images in the dark sun within the dark moon, is a blast exploding exploding exploding exploding explo- explo- explo- explo-ding -ding -ding -ding – then again absolute silence.
But the arousal of sexual tension is not the only occasion for the sublime imagery to cut loose. Yet another moment of ecstasy is recounted by Gao in a quite different setting, in the chapter describing the author’s visit to a Buddhist monastery and his participation in the morning prayer (87.5-88.5%, windows no. 446-453, sentences no. 6849-6949). The persistent sound of drums which leads the monks into a spiritual trance is so powerful and overwhelming that, “as the speed increases and the momentum builds up,” the drumming reverberations “become one continuous sound which makes the heart palpitate and the blood surge. The overpowering sound of the drum does not slacken, it is simply breathtaking.” (节奏随即越来越快，重重迭迭，轰然一片，人心跟着搏动，血也沸腾。浑然一片的鼓声毫不减缓，简直不容人喘息)We can see that here again the author employs the same narrative pattern: the most important words —i.e., the vibrating (震荡)thunderous soundwaves (声浪)—dominate the arrangement of the narrative space, as the thumping sound of the drum spreads through the scripture hall (大雄宝殿,大殿,and殿堂)and delivers the pious monks (和尚) into the spiritual realm:
Figure6 Co-occurrence network visualization of windows no. 446-453 (87.5-88.5%, sentences no. 6849-6949) in Soul Mountain. Size of the node indicates its eigenvector centrality, and edge transparency indicates distance. Nodes with eigenvector centrality lower than 0.035 have been removed for better visibility
Table 7 Influential nodes in the co-occurrence network visualized in Figure 6
Continued Table 7
Table 8 Edges with shortest distance in the co-occurrence network visualized in Figure 6
One element that draws attention is the clustered structure of this network (see Figure 6). The cluster on the left side centers on the descriptions of place, i.e., the scripture hall and the corridors of the temple, while the cluster on the right foregrounds aural experiences. Since this network graph is a representation of word co-occurrences, it clearly demonstrates how targeted Gao’s choice of vocabulary is and how carefully his writing “speaks” to the reader’s senses. This will become even clearer if we consider the word “inside” (li,里), which scores a high eigenvector centrality and partakes in edges with shortest distances (see Table 8). One could think of li (里) as yet another “stopword” and remove it for a more precise analysis, but I believe that the current result can inspire an illuminating interpretation. Chapter 69, which contains the monastery episode, begins literally “in the dream” (睡梦里). Other location descriptions quickly follow, as the hero(“I” or wo我) gropes his way through the monastery, trying to identify the source of the mysterious sound: “in the temple”(寺庙里), “in the courtyard”(庭院里), “in the cloister” (回廊里), “in the side yard” (侧院里), “in the corridor” (走廊里), “in the hall” (殿堂里), and even “in the incense burner” (香炉里). While “I” is finding his way to the praying monks, his spatial movement inward is paralleled by an increasingly powerful aural progression outward, as the sound of the drum grows stronger and ultimately disencumbers the monks from earthly restraints.
As in the final pages of Second Sun, however, such dream-like episodes should not blind us to the contrapuntal mechanics which partakes of what Lam calls “theatricality.” It is my contention that through the disaggregation of the self, Soul Mountain constructs the plot in such a way that beyond the scope of the iterative dream there always remains the space for a detached spectator. In other words, while availing itself of the sublime mechanics as it postulates the freedom of the soul through the metaphor of spiritual ascension, Soul Mountain simultaneously employs the additional space offered by the non-coincidence of the self to inspire a meta-narrative reflection:
I too join in, and like them press the palms of my hands together and chant Namo Amitofu, but I clearly hear another sound. As each sentence of the sutra is about to end there is always a voice with a slightly higher pitch than the chanting. So there is therefore still the unextinguished passion of a soul still being tormented.
The provenance of the split self has been located by scholars in Gao Xingjian’s notion of “tripartition of the actor,” which requires the actor to keep a cool and detached observation of both the character he portrays and of his performance.But what is more important for the present analysis is the fact that in Soul Mountain the narrative itself acknowledges the limitation to repetition at the same time as it repeats itself:
I don’t know where I am at this moment, I don’t know where this realm of Heaven comes from, I look all around.
I don’t know that I don’t understand anything and still think I know everything.
Things just happen behind me and there is always a mysterious eye, so it is best for me just to pretend that I understand even if I don’t.
While pretending to understand, I still don’t understand.
The fact of the matter is I comprehend nothing, I understand nothing.
Eventually, to God’s inquiring gaze Gao Xingjian demurs with non-knowledge. As a theatrical act, this taking distance from the recursive loop of the dream is not detectable by any means of statistical reasoning. The “ruptured dreamscape” forces the reader to cross a boundary, not as yet another mechanical repetition, but as a singular event of discontinuity and narrative anarchy through which the soul is split in half and flees from the dream.
Numerous objections to my approach can be raised. The most glaring issue is the choice of the sublime vocabulary, which in the present study is based on vector representations of words from just one poem, albeit a highly representative one. A better model would provide a more variegated selection of words and obviate the need to hand-pick additional terms, although my ultimate ambition here is not to find “all” sublime words but rather to illustrate a particular narrative pattern. Moreover, one needs to stay vigilant against a possible but erroneous impression that we could unproblematically copy the vocabulary of the sublime determined in this article and apply it in another milieu. While I believe that the mechanism through which the entanglement of vocabularies takes place will be similar for longer narrative forms also in other contexts, the building blocks of the ascending series are culturally specific.Yet another common objection concerns the ideal, good-faith reader assumed by the present theory. Undoubtedly, all readers deviate to a greater or lesser extent from the original order of presentation and approach the various parts of a narrative in ways unanticipated by the author. It is unlikely, however, that a reading detached from the syllogistic plot sequence or one that neglects the intended vocabulary distribution — e.g., picking chapters at random, or the repetitive, laborious close-reading of the sort performed in academic settings — will produce a dynamic necessary to elicit the powerful aesthetic experience of the sublime. Finally, setting the window size to two sentences is a choice that influences the final results. Using different window sizes would produce networks with different characteristics.
Nevertheless, I believe I have managed to demonstrate a few things. Although counting is not interpreting, I show that computational criticism can draw our attention to the ways in which the formal arrangement of vocabulary works in synchrony with the novel’s plot to achieve a powerful aesthetic effect that we can now theorize as the sublime. Furthermore, I have suggested that we should understand the sublime as an embodied experience of sequential ascent beyond a boundary, elicited through a performative process that associates sensorimotor memories of powerful natural phenomena with a secondary discourse, be it nationalism, morality, or religion. In other words, instead of “liberating” the sublime from the “narrow confines of rhetoric,”we should precisely see it as a rhetorical mechanism that addresses and manipulates our “muscular imagination.”The computational theory of the sublime as presented here is thus free from the metaphysical claims of Kant’s philosophy and can further illuminate our explorations of the sublime in non-Western contexts. Finally, my reading of Second Sun and Soul Mountain proves that there is more to the sublime in the 1980s Chinese literature than “desublimation” and the mainstream trends of “the grotesque, the fantastic, and the schizophrenic.”As we have seen, writers continued to be inspired by revolutionary romanticism and articulated bodily and spiritual experiences as sources of delight.
Appendix: Sublime Vocabulary
I would like to thank Mark Algee-Hewitt, Robert Clewis, Cynthia Vialle-Giancotti, Melissa Hosek, Lingjia Xu, and Aaron Gilkison for their advice and support. This article has also benefited from questions and suggestions I received from friends and colleagues at the Center for Spatial and Textual Analysis (CESTA) and the Center for the Study of the Novel (CSN) at Stanford University. I am also very grateful to the anonymous reviewers and the editors of this volume for their insightful and inspiring comments.
Ban Wang, The Sublime Figure of History: Aesthetics and Politics in Twentieth-Century China, Stanford: Stanford University Press, 1997, pp. 10-11.
Ban Wang, The Sublime Figure of History, p. 8.
Judith Shapiro, Mao’s War Against Nature: Politics and the Environment in Revolutionary China, Cambridge: Cambridge University Press, 2001.
John Crespi, Voices in Revolution: Poetry and the Auditory Imagination in Modern China, Honolulu: University of Hawaii Press, 2009, p. 77.
See Tie Xiao, Revolutionary Waves: The Crowd in Modern China, Cambridge, MA: Harvard University Asia Center, 2017, p. 109.
Andy Rodekohr, “Human Wave Tactics: Zhang Yimou, Cinematic Ritual, and the Problems of Crowds,” in Red Legacies in China: Cultural Afterlives of the Communist Revolution, eds. Jie Li and Enhua Zhang, Cambridge, MA: Harvard University Asia Center, 2016, pp. 271-296.
Haiyan Lee, “The Charisma of Power and the Military Sublime in Tiananmen Square,”The Journal of Asian Studies, vol. 70, no. 2, 2011, pp. 397-424.
See Guy Sircello, “How Is a Theory of the Sublime Possible?,”The Journal of Aesthetics and Art Criticism, vol. 51, no. 4, 1993, pp. 541-550; Jane Forsey, “Is a Theory of the Sublime Possible?,”The Journal of Aesthetics and Art Criticism, vol. 65, no. 4, 2007, pp. 381-389; Lars Aagaard-Mogensen, The Possibility of the Sublime: Aesthetic Exchanges, Newcastle, UK: Cambridge Scholars Publishing, 2018.
Such spatial primitives as “container,”“path,”“motion,”“in/out” and “top/down” serve as building blocks of a great number of metaphors used in daily life. For the notion of primary schema and primary metaphor, see, for example, George Lakoff and Mark Johnson, Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought, New York: Basic Books, 1999; Mark Johnson, The Meaning of the Body: Aesthetics of Human Understanding, Chicago: University of Chicago Press, 2007; Jean M. Mandler and Cristobal Pagán Cánovas, “On defining image schemas,”Language and Cognition: An Interdisciplinary Journal of Language and Cognitive Science, vol. 6, no. 4, 2014, pp. 510-532; Mark Johnson, Embodied Mind, Meaning, and Reason: How Our Bodies Give Rise to Understanding, Chicago: The University of Chicago Press, 2017. For conceptual metaphors in Chinese, see, for example, Ning Yu, The Contemporary Theory of Metaphor: A Perspective from Chinese, Amsterdam: J. Benjamins Pub, 1998. For “boundary” as a schema, see Barbara Dancygier, “’Un-Walling’ the Wall: Embodiment and Viewpoint,”The Cognitive Humanities, ed. Peter Garratt, London: Palgrave Macmillan, 2016, pp. 55-70. For the relationship between music and motion imagery, see Steve Larson, Musical Forces: Motion, Metaphor, and Meaning in Music, Bloomington: Indiana University Press, 2012.
Jan Cohn, Thomas H. Miles, “The Sublime: In Alchemy, Aesthetics and Psychoanalysis,”Modern Philology, vol. 74, no. 3, 1977, especially pp. 289-292.
See Immanuel Kant, Paul Guyer, Critique of the Power of Judgment, Cambridge, UK: Cambridge University Press, 2000, pp. 131-139 (5: 248-256).
Edmund Burke, Adam Phillips, A Philosophical Enquiry into the Sublime and Beautiful, Oxford: Oxford University Press, 1990, p. 68.
Quoted in Robert Clewis, The Sublime Reader, London: Bloomsbury Academic, 2019, p. 93, original emphasis.
Stephen Halliwell et al., Aristotle: Poetics, Longinus: On the Sublime, and Demetrius: On Style, Cambridge, MA: Harvard University Press, 1995, p. 205.
Stephen Halliwell et al., Aristotle, p. 237, emphasis added. Longinus also indicates that, by itself, “amplification” is not enough to produce the effect of sublimity; what is also required are “lofty thoughts,” a point which will become clearer later in this article.
For the dynamic nature of image-schemas and metaphors built upon them, see Robert B. Dewell,“Dynamic Patterns of CONTAINMENT,” eds. Beata Hampe and Joseph E. Grady, From Perception to Meaning: Image Schemas in Cognitive Linguistics, Berlin: Mouton de Gruyter, 2005, pp. 369-394; Mark Johnson, The Meaning of the Body, pp. 113-134.
See, for example, Barbara Mittler, A Continuous Revolution: Making Sense of Cultural Revolution Culture, Cambridge, MA: Harvard University Asia Center, 2012, pp. 189-255. See also the section on Mao Zedong’s poetry in Ban Wang, The Sublime Figure of History, pp. 187-193; Corey Byrnes, Fixing Landscape: A Techno-Poetic History of China’s Three Gorges, New York: Columbia University Press, 2018, pp. 25-53.
Margaret Ziolkowski, Literary Exorcisms of Stalinism: Russian Writers and the Soviet Past, Columbia, S.C.: Camden House, 1998, p. 111; Dan Levin, Stormy Petrel; The Life and Work of Maxim Gorky, New York: Appleton-Century, 1965, pp. 85-86. Gorky is, of course, not the only author who imagined socialist revolution as a powerful natural phenomenon. Tora Lane has recently shown that in the 1920s Soviet literature the people appear as an immense natural force, embodying Alexander Blok’s famous statement that “Russia is a storm” or Boris Pilnyak’s even more succinct “Russia. Revolution. Snowstorm.” See Tora Lane, “Heroism or Tragedy: The Sublime in the Revolutionary Romanticism of Soviet Aesthetics,” Revolutionary Russia, vol. 30, no. 2, 2017, pp. 247-267.
The most popular Chinese translation (used here) is by Ge Baoquan (戈宝权), and the English one comes from Wikipedia. See the Wikipedia entry “The Song of the Stormy Petrel,” https://en.wikipedia.org/wiki/The_Song_of_the_Stormy_Petrel, accessed Apr. 15, 2021.
Shen Li et al., “Analogical Reasoning on Chinese Morphological and Semantic Relations,” Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), Melbourne, 2018, pp. 138-143. See “Chinese Word Vectors 中文词向量,” https://github.com/Embedding/ Chinese-Word-Vectors, accessed Apr. 1, 2021.
Cosine similarity is used to compare vectors.
For the full list of the sublime vocabulary, see Appendix.
Weijia Wang, “Kant’s Mathematical Sublime: The Absolutely Great in Aesthetic Estimation,”Kantian Review, vol. 25, no. 3, 2020, pp. 465-485; Emily Brady, The Sublime in Modern Philosophy: Aesthetics, Ethics and Nature, Cambridge: Cambridge University Press, 2013, pp. 58-60.
Today we can represent an infinity of points formally, by means of universal quantifiers. See Michael Friedman, Kant and the Exact Sciences, Cambridge, MA: Harvard University Press, 1992, pp. 55-135.
For an example of topic modeling used in literary studies, see Andrew Piper, Enumerations: Data and Literary Study, Chicago; London: The University of Chicago Press, 2018, pp. 66-93.
I use the pkuseg Python package to segment Chinese texts into individual terms. See Ruixuan Luo et al., PKUSEG: A Toolkit for Multi-Domain Chinese Word Segmentation, 2019, https://arxiv.org/abs/1906.11455, accessed Apr. 10, 2021.
For a discussion on the window size in literary study applications, see Jordan Boyd-Graber et al., “Applications of Topic Models,”Foundations and Trends in Information Retrieval, vol. 11, no. 2-3, 2017, pp. 70-71.
In other words, I collect the terms with the highest probabilities of occurring in that topic until the sum of their probabilities exceeds 10%. I use the topicmodels package for R to build the models, with Gibbs sampling and default hyperparameters, performing 3,000 iterations for each model.
Before slicing the top 10% of the probability distribution and calculating the correlations, I remove the words (if any) belonging to the sublime vocabulary from the terms generated by the model for each topic to eliminate false positives. I use Pearson’s correlation coefficient to measure the linear relationship between frequencies.
Generating several models also allows me to make up for the volatility of the topic modeling algorithm.
Stephen Halliwell et al., Aristotle: Poetics, Longinus: On the Sublime, and Demetrius: On Style, p. 205.
Highlighted words belong to the sublime vocabulary. The two sentences come from Second Sun, a novel by Liu Baiyu which I will analyze in the second half of this article.
For the Pyramids, see Immanuel Kant, Critique of the Power of Judgement, pp. 135-136 (5:252). For “encompassing narrative,” see Paul Ricoeur’s discussion on biblical parables: “The Bible and the Imagination,” in Figuring the Sacred: Religion, Narrative, and Imagination, Minneapolis: Fortress Press, 1995, pp. 144-166.
See Nelson Cowan, “The Magical Mystery Four: How Is Working Memory Capacity Limited, and Why?,”Current Directions in Psychological Science, vol. 19, no. 1, 2010, pp. 51-57; Patrick Colm Hogan, What Literature Teaches Us About Emotion, Cambridge: Cambridge University Press, 2011, pp. 40-75; Nigel Fabb, What Is Poetry?: Language and Memory in the Poems of the World, Cambridge, UK: Cambridge University Press, 2015, pp. 171-191; Elizabeth H. Margulis, On Repeat: How Music Plays the Mind, New York: Oxford University Press, 2013, pp. 75-94.
Kant uses the term “vibration”(“Erschü tterung”) in Critique of the Power of Judgement, p. 141 (5:258).
For semantic distance, see Ian Lancashire, “Cognitive Stylistics and the Literary Imagination,”A Companion to Digital Humanities, eds. Susan Schreibman et al., London: Blackwell Publishing, 2004, pp. 397-414.
For technology as “formal arrangement,” see Nicholas D. Paige, Technologies of the Novel: Quantitative Data and the Evolution of Literary Systems, Cambridge, UK; New York, NY: Cambridge University Press, 2021.
Liu Baiyu, Di er ge taiyang, Beijing: Renmin wenxue chubanshe, 1987; Gao Xingjian, Lingshan, Taipei: Lianjing chuban sheye gongsi, 1990; Gao Xingjian, Soul Mountain, trans. by Mabel Lee, New York: HarperCollins Publishers, 2000. Hereafter, page numbers in square brackets for Soul Mountain refer to the Chinese edition.
Liu Baiyu, Di er ge taiyang, p. 53.
Barbara Mittler, A Continuous Revolution, pp. 71-72, 84-85, 340-344.
Liu Baiyu, Di er ge taiyang, p. 233.
One reviewer asked about the relationship between the seemingly ordinary words such as 前面 or 新 and the novel’s plot or even nationalism. Indeed, words considered by themselves cannot provide us with much useful information. But when considered in context together with other topical words or, as in the present case, as a member of a network, the word 前面 becomes much more meaningful. The “dramatic” dimension of 前面comes to the fore when juxtaposed with words like 激流 and 部队.
Kenneth Burke, Counter-Statement, Berkeley: University of California Press, 1968, p. 124.
I use Cytoscape 3.8.2 to visualize co-occurrence networks, applying the yFiles Organic Layout, which is a multi-purpose layout style for undirected graphs. See “yFiles Layout Algorithms for Cytoscape,” https://www.yworks.com/products/yfiles-layout-algorithms-for-cytoscape, accessed Feb. 10, 2021. To calculate eigenvector centrality, I use the “CentiScaPe” app available in Cytoscape. See Giovanni Scardoni et al. “Analyzing Biological Network Parameters with CentiScaPe,”Bioinformatics, vol. 25, no. 21, 2009, pp. 2857-2859.
Liu Baiyu, Di er ge taiyang, pp. 421-422.
For Mendelssohn, see Robert Clewis, The Sublime Reader, p. 93. For synecdoche, see Roman Jakobson,
“The Metaphoric and Metonymic Poles,” in Dirven René , Ralf Pö rings, Metaphor and Metonymy in
Comparison and Contrast, Berlin: Mouton de Gruyter, 2002, pp. 41-48.
Liu Baiyu, Di er ge taiyang, p. 454.
The famous director Cui Wei崔嵬frequently shot scenes of massive gatherings to express revolutionary sentiment, a style Dai Jinhua calls “Cui Wei ceremony.” See Ban Wang, The Sublime Figure of History, p. 130.
Ling Hon Lam, The Spatiality of Emotion in Early Modern China: From Dreamscapes to Theatricality, New York: Columbia University Press, 2018, p. 4.
Ling Hon Lam, The Spatiality of Emotion in Early Modern China, p. 35.
Gao Xingjian, Soul Mountain, p. 137 (139), my emphasis.
Gao Xingjian, Soul Mountain, p. 137 (139).
Gao Xingjian, Soul Mountain, p. 504 (525).
Gao Xingjian, Soul Mountain, p. 442 (450).
Gao Xingjian, Soul Mountain, p. 443 (461-462).
Gao Xingjian and Gilbert Fong, The Other Shore: Plays, Hong Kong: The Chinese University of Hong Kong Press, 1999, p. xx.
Gao Xingjian, Soul Mountain, p. 506 (526).
The “stormy petrel” imagery, for example, was probably much more easily accepted in modern Chinese culture given the already present and deeply rooted representations of the Yellow River and the Yangtze as rivers bestowed with destructive powers. Nowadays, such imageries are often enriched with representations of gigantic skyscrapers, highways, cityscapes, etc. Compare the final fight scene in The Matrix trilogy.
Robert Doran, The Theory of the Sublime from Longinus to Kant, Cambridge: Cambridge University Press, 2015, p. 111.
Kenneth Burke, Counter-Statement, p. 141. For a related discussion about the notion of “crescendo”from the perspective of psychology and rhetoric, see Jeff Pruchnic, “The Priority of Form: Kenneth Burke and the Rediscovery of Affect and Rhetoric,” eds. Donald Wehrs and Thomas Blake, The Palgrave Handbook of Affect Studies and Textual Criticism, New York: Palgrave Macmillan, 2017, pp. 371-390.
Ban Wang, The Sublime Figure of History, pp. 229-262.