What's Happening
The Japanese Diet did not seriously take up the Ianfu (Comfort Women) issue until the 1990s. For the several decades before that, the survivors were alive. They had the words to testify. What the surrounding society lacked was the frame to receive their testimony as a matter for debate. Minamata disease was officially confirmed in 1956; the government did not identify the cause until 1968. For the twelve years in between, the question of causation itself was suspended inside the ministries and the medical academy. Domestic violence in Japan was treated as fuufu kenka (marital quarreling) until the 1990s, a zone the police did not enter, and a category for which the institutional vocabulary of harm barely existed.
The problems that were never debated are problems the society did not have the capacity to debate. Silence is not empty space. It is a manufactured absence.
Agnotology has treated the production of what is not known. The subject of this essay is the mirror question: how to detect, after the fact, what was never debated. Article #4 What Is Silence-Structuring traced the generative mechanism of silence (a three-stage model of how silence gets structured); Article #6 Media Agenda-Setting and Invisibility traced the machinery that sustains it. The present essay continues that line by organizing the excavation of already-fixed silence into four layers and five methods. Article #4 and the present essay form an asymmetric pair: one traces the process of being made, the other traces the process of being excavated, and together they sandwich the silence phenomenon from both sides.
The essay asks three questions. First, what structural layers does silence contain? Second, which retroactive detection method corresponds to each layer? Third, what ethical limits does the act of excavation itself carry?
Absence
The speech act itself has not occurred
The affected party is given no venue to speak, or is not made an object of record. The state of ianfu survivors before 1991 lies close to this layer. Not documented, not heard, not counted in any statistic.
Concealment
Speech exists but is unpublished or anonymized
Information is present inside an organization or research team; from the outside it is invisible. The internal deliberations of the Kumamoto University research team in the early Minamata period, the fieldnotes of anonymized ethnographic assistants. Information stagnates inside institutional boundaries.
Distortion
Speech is public but its context has been stripped
Facts circulate while severed from any interpretive frame. The kibyō reporting of early Minamata, the media treatment of domestic violence as chiwa genka (lovers' quarrel). The frame that would ask about cause is missing, even as information flows.
Delegitimization
Speech is fully present but its credibility is stripped away
Labels like "subjective," "emotional," or "political" strip standing from speech. The reframing of ianfu survivor testimony as "a political motive to humiliate Japan," the pathologization of DV victims as "delusional." This is the layer at which Fricker's testimonial injustice bites hardest.
Background and Context
Lineage: Foucault, Trouillot, Fricker, Sanjek
The methodological starting point for treating silence systematically is Foucault (1969)'s L'archéologie du savoir. Foucault called archaeology a method that reads history not through "what was said" but through "what was permitted to be said." The conditions under which a statement becomes possible (and, symmetrically, the conditions under which it becomes impossible) are extracted from the discursive formation rules of an era. This move let us treat "what was never debated" not as accidental omission but as the consequence of a rule.
Trouillot (1995)'s Silencing the Past carried Foucault's stance into the practice of historical writing. Trouillot showed that silence in history is produced in four stages: fact creation (silences enter as soon as one chooses who records), archive creation (what is retained, what is discarded), narrative creation (which archived items get cited), and retrospective significance (which past events are, in hindsight, elevated as important). Trouillot argued that the reason the Haitian Revolution stood at the margin of world history for so long was that all four stages of silencing had accumulated on top of it.
The philosophical grounding of silence is Fricker (2007)'s Epistemic Injustice. Fricker distinguished two kinds. Testimonial injustice is the injustice by which prejudice deflates a speaker's credibility. Hermeneutical injustice is the injustice by which a collective vocabulary for naming a form of harm is simply missing. Until terms like domestic violence, sexual harassment, and gaslighting were institutionalized, victims had no words with which to name their own experience. This is not a rhetorical failing of individuals; it is a structural absence of collective hermeneutical resources.
From anthropology, Sanjek (1993) exposed the structural erasure of African and Asian assistants from classical ethnography. Behind monographs published under a single author's name stood dozens of anonymized local collaborators, whose names often did not even reach the acknowledgments. Sanjek called this anthropology's hidden colonialism and, as a matter of method, insisted on the inversion of individual fieldnotes: who wrote, who translated, who drew the maps.
Three Japanese Cases
Read against this lineage, three Japanese cases separate into distinct types.
The Ianfu (Comfort Women) issue is a case in which all four Trouillot stages operated. Japanese military documents were systematically incinerated at the end of the war (stage one). The surviving documents were scattered across defense archives and diplomatic records, and access to ianfu-related materials was long restricted for researchers (stage two). Until the 1990s controversy over the Yoshida Seiji testimony and the government's own investigation (the Kono Statement), the voices of survivors had no channel through which to circulate as narrative (stage three). The retrospective significance stage did not turn until Kim Hak-sun's 1991 public testimony under her real name; before that, the whole matter existed as something that had "not happened" (stage four).
Minamata's twelve years look closer to Fricker's hermeneutical injustice. At the moment of official confirmation in 1956, the medical vocabulary for organic mercury poisoning was effectively absent in Japan. There was no established medical frame connecting factory effluent to neurological symptoms; victims were given no name beyond kibyō (strange disease). The organic mercury hypothesis was proposed by the Kumamoto University medical team in 1959, and the government's own recognition took a further nine years. The silence of those years is a case in which the absence of a collective cognitive frame invalidated individual medical observations.
Domestic violence in Japan is a case of vocabulary-absence silence. The proverb fuufu genka wa inu mo kuwanai ("even dogs don't touch marital quarrels") stands for the long period in which Japanese society had no vocabulary that named domestic violence as violence. Until the DV Prevention Act passed in 2001, the police applied the doctrine of non-intervention in civil matters to households, and both the vocabulary and the statistics for making the harm visible were thin.
What the three cases share is that silence was not merely "not speaking." The resources for speech were structurally undersupplied. The next section organizes the techniques for excavating that structure into five methods.
Reading the Structure
Four Layers of Silence
Silence is not uniform. It decomposes into four layers with different intensities and mechanisms.
The first layer is Absence. The speech act itself has not occurred. Either the speaker was given no venue, or was not made an object of record. The state of ianfu victims before 1991 lies close to this layer. Not documented, not heard.
The second layer is Concealment. Speech exists but is unpublished or anonymized. The internal deliberations of the Kumamoto University research team in the early Minamata period, or the fieldnotes of ethnographic assistants that Sanjek documented, belong here. Information is present inside an organization or research group; from the outside it is invisible.
The third layer is Distortion. Speech is public but its context has been stripped and its meaning altered. The kibyō reporting of Minamata circulated the facts while severing them from any frame that would ask about cause. The media treatment of domestic violence as chiwa genka (lovers' quarrel) in earlier decades belongs here too.
The fourth layer is Delegitimization. Speech is fully present, but its standing is stripped away by labels like "subjective," "emotional," or "political." The reframing of ianfu survivor testimony as "a political motive to humiliate Japan," or the pathologization of DV victims as "delusional," belong here. This is the layer at which Fricker's testimonial injustice bites hardest.
The four layers sediment in stairs. Excavation works downward from the upper layer to the lower.
Five Methods of Retroactive Detection
Each layer has a corresponding detection method.
The first method is statistical absence detection: identifying the systematicity of values that are not observed inside a given set. As Article #7 Statistical Invisibility of Femicide showed, the absence of an independent femicide category in Japanese crime statistics is itself an indicator of Absence. The sharp jump in police-recorded DV incidents after the 2001 DV Prevention Act is a retrospective demonstration that the earlier zero was not a real zero but a statistical absence. Running the principle that "not in the statistics" does not mean "did not happen" as a systematic method is what this technique amounts to.
The second method is archival inversion: reading the reverse side of what was preserved. This corresponds to Trouillot's stage two. Who recorded, and who did not get recorded. What was retained, what was discarded. Lists of documents destroyed by the Imperial Army, the record of information-disclosure requests answered as "not extant," the missing years of a public archive itself become objects of analysis. The digital collections of the National Archives of Japan can be used as a primary resource for this inversion.
The third method is testimonial reconstitution: accessing the affected community and reconstructing the inside of silence through oral testimony. The ianfu testimony-collection work in the 1990s by the Asia-Japan Women's Resource Center and the Women's Active Museum on War and Peace, and the record-keeping movement of Minamata patient support organizations, are examples. It is a method that cuts across the third layer (Distortion) and the fourth layer (Delegitimization), but its temporal window is severe. It depends on survivors being alive to speak.
The fourth method is comparison: contrasting the presence and absence of a phenomenon across similar societies. Domestic violence entered the English-speaking vocabulary as domestic violence in the 1970s; Japan did not institutionalize it in law until 2001. That thirty-year gap makes visible a Japan-specific lag in hermeneutical resources. When contemporaneous harm was being named elsewhere but not here, the fact of the un-naming itself becomes an indicator.
The fifth method is temporal shift analysis: identifying the tipping point at which silence turned into speech, and reading the structure retroactively from that pivot. Kim Hak-sun's 1991 testimony. The 1968 Minamata government statement. The 2001 DV Prevention Act. Comparing the vocabulary in circulation before and after each pivot brings the lexical void that had been sustaining the silence into relief. This method connects with the three-stage silence-structuring model of Article #4 and provides the empirical entry point for testing how silence gets destructured.
The four-by-five matrix is designed as a typology of primary affinities, allowing secondary cross-layer application. The first method (statistical absence) is strong on Absence; the second (archival inversion) is strong on Absence and Concealment. The third (testimony) is strong on Concealment and Distortion; the fourth (comparison) is strong on Distortion and Delegitimization. The fifth (temporal shift) treats the whole set of layers diachronically. In practice, empirical research starts from this primary axis and layers several methods together to excavate the full stratigraphy of silence.
Limits of the Methodology
Excavation-oriented methodology carries three limits.
The first limit is survivor bias. What one can work with is the physically surviving testimony, the surviving document, the surviving witness. The silence that is materially gone cannot be excavated. Incinerated documents, deceased victims, factories that closed before any record was made. Excavated silence is always partial. Reading "what we could excavate" as "the whole of what was there" reproduces a new layer of silencing.
The second limit is the reflexivity of the excavating act. As researchers gain access to affected communities and reformulate testimony in academic vocabulary, the words gain scholarly legitimacy while being severed from the vocabulary of those who spoke them. Terms like "sexual slavery" and "wartime sexual violence" secured international standing in ianfu research, but their grain is not identical to the grain of survivors' own testimony, which uses concrete phrases like heitai-san ni ("by the soldiers"). Within Fricker's framework, excavation dissolves one hermeneutical injustice while risking a new form of testimonial injustice. Specifically, the re-framing of speech by academic authority becomes the new problem.
The third limit is the division of labor between academy and activism. The academy secures methodological rigor by keeping distance; activism prioritizes stopping present harm. Their time horizons do not align. While academic research waits to consolidate its methods, survivors pass away. How to redesign this division is a practical question that methodology alone cannot answer.
Where to Dig: Implementation Sites
The methodology has three implementation sites.
The first is policy-document archives. Central ministries and local governments generate a vast volume of policy documents. What gets published and what gets treated as non-public and then discarded carries a structural bias. The full-text horizontal analysis of local council minutes handled at ISVD labs/machikarte can be operated not only as a visualization of what was debated but as a retroactive detection method for what was not. When a certain topic is being debated in neighboring municipalities but is absent from a given municipality, the absence itself signals silence-structuring at work.
The second is the taxonomic axes of statistical data. The absent femicide category from Article #7, the statistical burial of non-regular employment inside labor accident statistics, the gap between actual poverty and the recognized welfare-capture rate. These are all cases of Absence generated by "no aggregation category exists." Reviewing the statistical code hierarchy of e-Stat and systematically surfacing the expected-but-missing axes is a form of civil-society-side retroactive detection that can be implemented.
The third is temporal-shift analysis in media archives. Combining the digital archives of national newspapers with the microfilm holdings of the National Diet Library, one can trace the first-mention year and circulation trajectory of specific terms such as DV, sexual harassment, harassment, and gender. From the surrounding reporting of the periods before a term became mainstream, one can reconstruct what was happening during that lexical void.
Each of these three sites connects to the systematization of what agnotology has so far treated case by case. Together they push the field toward the shape of a reusable detection typology.
Open Questions
First, how does one define operational indicators for the four layers? Distinguishing Absence as "physical absence of speech" and Delegitimization as "credibility deflation of speech" is linguistically clear, but the empirical measurement of each remains unsettled. Second, the methodological guidelines on ordering and combining the five methods are underdeveloped; case-by-case accumulation is what will refine the typology. Third, how does one institutionalize the epistemic repair that follows excavation? State apology, reparations, addition to official records, inclusion in textbooks. Each institutional route needs a circuit that connects it to the outputs of excavation.
The archaeology of silence is not aimed at excavation for its own sake. The final target is converting excavated silence into a collective resource that stops the next silencing. The series will pursue that conversion route from the next essay onward.
References
The Archaeology of Knowledge — Foucault, M.. Routledge Classics (2002 English edition)
Silencing the Past: Power and the Production of History — Trouillot, M.-R.. Beacon Press (20th Anniversary Edition 2015)
Epistemic Injustice: Power and the Ethics of Knowing — Fricker, M.. Oxford University Press
Anthropology's Hidden Colonialism: Assistants and Their Ethnographers — Sanjek, R.. Anthropology Today, 9(2), 13–18
Agnotology: The Making and Unmaking of Ignorance — Proctor, R. N. & Schiebinger, L.. Stanford University Press
The Unknowers: How Strategic Ignorance Rules the World — McGoey, L.. Zed Books
Article #4 What Is Silence-Structuring / Article #6 Media Agenda-Setting and Invisibility / Article #7 Statistical Invisibility of Femicide