This contribution is based on a presentation given at The Digital Orientalist’s Virtual Conference 2025 (AI and the Digital Humanities) by Lin Du (National University of Singapore). The recording of the presentation can be found here.
For a more detailed exploration of the subject matter of this article, see the author’s paper “From Information to Metaphor: Tracking Photographic Editing in Chinese Wartime Magazines through Digital Historical Forensics” Journal of Chinese Cinemas 18, 2–3 (2025): 329–74.
Two photographs appear four years apart in Chinese Communist Party wartime pictorials. They show the same young man, on the same day, in front of the same mud‑walled courtyard in rural Hebei. Yet they are not quite the same image. The first, from March 1942, carries the caption Model Families: Parents Send Their Children, Wives Send Their Husbands. The second, from March 1946, names no family at all; the young man has become “a youth,” one of “the common people.” Between the two printings lies a set of editorial gestures easy to miss and hard to undo: a caption rewritten, a name erased, a crop tightened, a fragment placed in a new discursive neighborhood.
Where in a propaganda archive does ideology actually live? And what do the dominant methods for reading such archives—textual analysis on one side, large-scale computation on the other—systematically fail to see? The argument of this piece, drawing on my recent article (Du 2025) in the Journal of Chinese Cinemas, is that it lives in precisely these small, fussy differences. I propose a digital methodology called Digital Historical Forensics (DHF) to read them: a method designed to recover the editorial micro‑decisions that propaganda’s surface works to conceal.
A magazine built out of scarcity
The Jinchaji Pictorial (晋察冀画报) was founded in 1942 in a guerrilla base area straddling three provinces in north China. Its photographer‑editors trained their own camera operators, salvaged film and chemicals under blockade, and at times carried negatives strapped to their bodies on forced marches. Print runs nevertheless reached 440,000 copies for certain issues, constituting a mass medium improvised out of material scarcity.
Scarcity and the limits of wartime printing shaped an editorial practice that would otherwise look strange. Because film and equipment were precious, a single photograph was rarely treated as a singular record. It was cropped, mirrored, split, recaptioned, paired with different articles, and reprinted across issues and special series over the span of years. Alexei Yurchak (2005), writing on Soviet late‑socialist discourse, describes this mode as manifest intertextuality: an authoritative form of communication that sustains its force by repeating recognizable “blocks” while migrating them into new contexts. For Jinchaji editors, a photograph functioned as a reusable rhetorical resource whose meaning accrued across successive acts of editorial handling.
Using scale to find the singular
Digital humanities has largely organized itself around the promise of scale. Distant reading, topic modeling, and embedding‑based clustering share an aggregative rhetoric in which more documents yield more patterns and stronger signals. The “small” is what must be washed out so that the “big” can emerge.
DHF inverts that relation. The pipeline I built with collaborators (Du, Le, and Honig, 2024) does operate at scale: it ingests thousands of printed pages and photographic prints, segments individual images from complex magazine layouts, and retrieves near‑duplicate matches with roughly 78% top‑15 accuracy. Yet the purpose of this apparatus is to locate singular variants: specific pairs or triples of near‑identical images whose differences an unaided reader would never notice, because the relevant instances sit hundreds of pages apart across different physical volumes. The pipeline does not interpret; it retrieves. It surfaces pairs that would otherwise take a researcher months of manual comparison to identify, and returns the interpretive labor to the scholar.
Propaganda aspires to smoothness; its operations are designed to disappear into the image so that the image reads as given. DHF borrows the logic of forensic analysis. It looks for the snag, the mismatch, the edit that was never quite cleaned up, and treats those moments as the places where the editorial hand gets caught in the act.
A name that goes missing
The Liu Hanxing enlistment sequence shows what this looks like in practice. In early 1942 the photojournalist Sha Fei documented a village ceremony in Pingshan County, where a young peasant named Liu Hanxing was preparing to leave for the Eighth Route Army. His parents and young wife stood with him in front of the family home; a donkey, loaded for the journey, waited in the courtyard. Sha Fei shot at least four frames, staging and restaging the group.
Figure 1. Sha Fei, contact sheet of the Liu Hanxing enlistment, Pingshan, Hebei Province, 1942. (Harvard-Yenching Library, Papers of Sha Fei.)
The editors chose one frame and published it in the pilot issue of the Jinchaji Pictorial Current News Special (March 1942) under a caption that foregrounded the family: “Model Families: Parents Send Their Children, Wives Send Their Husbands.” The wording was deliberate. Conscription in the border areas was a genuinely hard sell in rural households, where losing the strongest laborer could mean a lost harvest. The caption addresses that anxiety directly: enlistment is framed as a contribution the family itself makes, and the collective honor returns to the household.
Figure 2. “Model Families: Parents Send Their Children, Wives Send Their Husbands.” Jinchaji Pictorial Current News Special, March 1942.
Four months later, in the first official issue of the Jinchaji Pictorial (July 1942), the same photograph reappeared under a new caption: “People of the Border Areas: Join the Volunteer Conscripts in Large Numbers.” The familial register has dropped away, replaced by a political grammar in which Liu Hanxing figures as an instance of a collective subject, “the people of the border areas.”
Figure 3. Jinchaji Pictorial, Issue 1, July 1942.
By September 1945, in a supplement printed as the Second Sino‑Japanese War ended, Liu Hanxing’s name dropped from the caption altogether; he is now “common people.” By March 1946, in Jinchaji Pictorial Series I: The Eighth Route Army and the Common People, the same figure is relabeled “a youth,” a term that, through the legacy of the New Youth magazine (新青年) and the May Fourth tradition, carried a specific charge of political awakening. Over four years, the photograph undergoes no visible retouching a casual reader would notice. What changes is the verbal envelope around it, and with each change the same pictorial trace is made to mean something incrementally different: a son, a peasant, the people, the youth, and finally an image of revolutionary subjectivity almost entirely detached from the person who originally stood for it.
Figure 4. Jinchaji Pictorial Supplement, Issue 3, September 15, 1945.
Figure 5. Jinchaji Pictorial Series I: The Eighth Route Army and the Common People, March 1946.
The small as the site where ideology works
None of these individual edits is dramatic. Editorial intervention in the Jinchaji Pictorial does its work through a long sequence of minor, low‑amplitude decisions, each small enough to pass unremarked. The visible operations (cropping a figure out of frame, mirroring a scene, inserting a hand‑drawn column of marching soldiers on a cover) remain rare and occasional. The routine operations are the captions, the typographic overlays, the paragraphs printed beside the image that anchor its sense before a reader can pose a question. The small, then, is the scale at which ideological work is actually done, and the scale at which propaganda prefers to operate, because low‑amplitude interventions accumulate while attracting no defense.
Much scholarship on Chinese Communist Party visual culture has examined the doctrinal frameworks that governed cultural production: the Yan’an Talks, socialist realism, the principle of combining “revolutionary realism and revolutionary romanticism.” Such frameworks describe propaganda as a system of ideas. DHF reveals propaganda as a system of micro‑decisions: a crop here, a caption change there, a retouched wall slogan, a hand‑drawn column of soldiers. These decisions were made by specific editors working under specific constraints of limited film, scarce paper, semi‑literate audiences, and shifting political imperatives. The small is where ideology meets materiality, and where abstract doctrine is translated into concrete visual practice.
This reframing bears on what we should ask digital methods to do. The dominant rhetoric of digital humanities still promises understanding through aggregation: run the model, read the cluster, see the trend. For archives like Jinchaji, where the historically interesting object is a difference of one centimeter of crop or the swap of a single character in a caption, aggregation flattens the very evidence it is meant to recover. DHF shares with Taylor Arnold and Lauren Tilton’s (2023) “distant viewing” the premise that computer vision is itself an interpretive practice requiring its own critical methods. Where distant viewing develops a semiotic framework for reading images at scale, DHF uses scale to locate the singular and then reads it closely. In concrete terms, computation scans tens of thousands of printed pages for near‑duplicates while the historian adjudicates what a given difference means in its editorial and political context. The result is computationally-guided close reading: the machine puts the historian in front of the pages where an argument can be made, and the argument is the historian’s to make.
Against the smooth archive
Propaganda archives present themselves as smooth surfaces. That smoothness is a historical achievement, produced by editors, darkroom technicians, typesetters, and political cadres making thousands of small decisions whose traces were meant to be undetectable. Because no editorial records survive for most of those decisions, the photographs themselves, and specifically the differences between their versions, become the primary archive of the work that produced them. Legal historian Chen Li (2024, 435) has described such evidence as “Invisible Archives”: traces left by processes that were never intended to be documented. DHF offers a method for restoring friction to that surface and for treating the archive as a record of ongoing editorial labor, most of it invisible at the scale of a single page.
For scholars working on visual materials from mass‑mobilizational contexts, such as Soviet pictorials, colonial photography, and Cold War propaganda across the bloc, the lesson could generalize. It also bears on a broader methodological habit: the treatment of visual media as illustrative supplement to textual analysis. By placing photographs at the center of the methodology, not as objects to be described in words but as material artifacts whose physical transformations constitute historical evidence, DHF insists that visual media deserve the forensic attention textual DH scholars have long devoted to manuscripts, editions, and variants. The most consequential historical evidence is sometimes not a pattern across ten thousand documents but a discrepancy between two versions of a single photograph. It is the edit no one signed, the caption rewritten on a second printing, the column of soldiers drawn in at the margin. These are small things. They are also, quite often, where the history is.
References
Arnold Taylor, and Lauren Tilton, eds. Distant Viewing: Computational Exploration of Digital Images (Cambridge, MA: The MIT Press, 2023).
Chen Li 陈利, Di guo shi dai de fa lü, zhi shi yu quan li 帝国时代的法律、知识与权力 [Law, knowledge and power in the age of empire] (Beijing: Shang wu yin shu guan, 2024).
Lin Du, “From Information to Metaphor: Tracking Photographic Editing in Chinese Wartime Magazines through Digital Historical Forensics.” Journal of Chinese Cinemas 18, no. 2–3 (2025): 329–74. https://doi.org/10.1080/17508061.2025.2514410.
Lin Du, Brandon Le, and Edouardo Honig, “Probing Historical Image Contexts: Enhancing Visual Archive Retrieval through Computer Vision,” ACM Journal on Computing and Cultural Heritage 16, no. 4 (2024): 84:1-84:17. https://doi.org/10.1145/3631129.
Alexei Yurchak, Everything Was Forever, until It Was No More: The Last Soviet Generation (Princeton, NJ: Princeton University Press, 2005).
