The First Official Count of the Ruins

Back in the last week of April 1906, before the refugee cottages rose in Golden Gate Park and before the Committee of Fifty turned its attention to permanent reconstruction, the work of measurement began.

Two documents arrived on the same morning in the offices of the Merchants Exchange Building, which still stood at the corner of California and Sansome Streets, its granite facade scorched but intact. The first was a mimeographed sheet from the office of City Engineer M.M. O’Shaughnessy, titled Preliminary Survey of Fire Destruction, Financial District. It listed city blocks in a neat grid, each marked with a single notation: Total Loss, Partial Loss, Standing. The second document was the Oakland Tribune of April 25, its headline running across six columns: EARTHQUAKE DEATHS MAY REACH 10, 000—Reports from Ruins Indicate Heavy Toll in Collapsed Buildings.

The engineer’s grid attributed 100 percent of the destruction to fire. The newspaper attributed an unspecified but vast portion of the mortality to the earthquake that had preceded it. Neither document cited sources. Neither explained its method. Together they established the terms of a conflict that would outlast the ruins themselves.

Counting burned acres, corpses, and dollars was never a neutral act. It was the first move in a struggle to define what had happened, who was responsible, and who would pay.

O’Shaughnessy’s survey sheet represented the city’s official position, and it was already circulating beyond San Francisco. The engineer had walked the burned district with a team of assistants, measuring street frontages and consulting fire department records to determine where the flames had traveled. His methodology was visible in the document’s very form: the grid assumed that destruction could be mapped block by block, that the fire’s path was traceable and therefore comprehensible. The earthquake appeared nowhere on the sheet. The collapsed buildings, the shattered water mains, the death of Fire Chief Dennis T. Sullivan when the California Hotel fell onto his quarters at Engine Company No. 2—where he and his wife had been sleeping when bricks crashed through the roof—these were preliminary conditions, not causes. The fire was the catastrophe. The fire could be explained.

This framing served immediate practical purposes. Insurance policies written in 1906 distinguished carefully between earthquake and fire as perils. Most standard policies excluded earthquake damage entirely; fire coverage, by contrast, was nearly universal among commercial properties. The distinction was not academic. The National Board of Fire Underwriters, which represented the industry nationally, had already begun to signal that it would resist claims attributed to seismic activity. The lower the earthquake’s profile in official accounts, the stronger the position of policyholders. O’Shaughnessy, a city employee working at the direction of Mayor Eugene Schmitz and the Committee of Fifty, understood this context. His survey sheet was a legal instrument as much as a technical document.

The Oakland Tribune’s competing estimate served different interests. The newspaper’s correspondent had spent three days in the ruins, interviewing refugees, soldiers, and relief workers. The figure of 10, 000 dead, later reduced in subsequent editions to several thousand, was not based on body counts, which were still impossible, but on extrapolation from reported building collapses. The correspondent noted that the Valencia Hotel on Eighteenth Street had buried an unknown number of guests; that the Brunswick Hotel on Mission Street had pancaked; that numerous tenements in the South of Market had simply disappeared into their own basements. These structures had failed before the fire reached them. Their occupants had died by falling masonry, not flame.

The Tribune’s higher estimate served the interests of relief fundraising. The greater the apparent catastrophe, the more urgent the appeal for outside assistance. The newspaper was actively cooperating with the California Promotion Committee, a private organization that was already managing donations from across the United States and abroad. By April 25, relief contributions had exceeded $5 million, a sum that would grow substantially in the following weeks. The Committee had every incentive to emphasize the earthquake’s role, not to minimize it. Human sympathy responded more readily to sudden seismic death than to the slower narrative of fire spreading through broken infrastructure.

Two documents, arriving the same morning, describing the same events, established a pattern that would persist for years. The city engineer’s grid and the newspaper’s headline represented different strategic investments in what the disaster would mean. The gap between them was not accidental. It was structural.

The mechanism of this divergence becomes visible in the records of the three institutions that attempted comprehensive tallies during the last week of April: the United States Army, through General Frederick Funston’s headquarters at the Presidio; the San Francisco Relief and Red Cross Funds, operating from temporary offices in the Merchants Exchange; and the city’s own Board of Supervisors, which had authorized O’Shaughnessy’s survey and was attempting to compile a parallel estimate of casualties. Each organization faced the same empirical chaos, the same smoldering ruins, the same displaced population, the same breakdown of ordinary record-keeping, and each produced a different account.

Funston’s headquarters approached the problem through military categories. The Army’s primary responsibility was order: the patrol of streets, the guarding of property, the management of refugee camps. Quantification served these ends. On April 26, Funston’s adjutant issued a memorandum requesting daily reports from all regimental commanders. The forms asked for the number of refugees in camps under military supervision, the estimated value of property protected by patrols, and incidents of disorder requiring armed response. The categories assumed that the disaster could be managed through established administrative routines. The earthquake appeared in these reports only as a disruptive event in the past tense, something that had created the conditions the Army now addressed. The fire, by contrast, was ongoing and therefore actionable. Funston’s officers tracked its final extinguishment with the same precision they applied to troop movements.

The Relief and Red Cross Funds, operating under the direction of banker James D. Phelan and his associates, developed a different taxonomy. Their ledgers, preserved in the California Historical Society’s archives, classified applicants for assistance by cause of destitution: fire, earthquake, or both. This distinction, maintained through thousands of individual interviews, created a statistical record of considerable granularity. A team of volunteer clerks, many of them women from the city’s elite families, conducted house-to-house surveys in the remaining unburned districts and in the Presidio camps. Their forms asked specifically about building damage: walls standing, roof intact, foundation cracked. The questions assumed that structural failure and fire damage were separable and that respondents could distinguish between them.

The results, compiled in a report dated May 3, revealed the difficulty of this exercise. In the Mission District, where the earthquake had caused significant structural damage before the fire arrived, 34 percent of applicants attributed their losses to both causes. In the Western Addition, where the fire had been stopped at Van Ness Avenue, 67 percent attributed losses to earthquake alone, meaning cracked walls, fallen chimneys, broken plumbing. Yet these same respondents often added, in the space for additional remarks, that they had lost everything to the flames that followed. The clerks’ categories did not capture the actual sequence of destruction. The form imposed a logic that the experience resisted.

The city’s own casualty estimates proceeded through yet another method. The Board of Supervisors directed the coroner, William H. Dimmig, to compile a death roll from hospital records, morgue reports, and military notifications. Dimmig’s office faced immediate obstacles. The city’s central morgue, located at the Mechanics’ Pavilion on Grove Street, had itself burned on April 19, destroying whatever records had been accumulated in the first twenty-four hours. The temporary morgue established at the Presidio was processing remains that arrived without identification, without documentation of where they had been found. The Army’s burial details, working in the ruins, were interring bodies where they lay when decomposition made removal impractical. Dimmig’s staff could confirm only what came to them through formal channels, and formal channels had collapsed.

On April 28, Dimmig released a preliminary count: 498 identified dead, with many more expected as records were compiled. The figure was immediately contested. The Army’s medical officer, Major Charles C. Byrne, reported to Funston that his own count of burials supervised by military details exceeded 800. The discrepancy was not merely clerical. Dimmig’s figure included only bodies that had been identified or at least counted at a morgue facility. Byrne’s included field burials, often of remains too damaged for identification, in mass graves at the Presidio and the Odd Fellows Cemetery. The city preferred Dimmig’s lower number. The Army stood by Byrne’s higher one.

The press amplified this divergence. The San Francisco Chronicle, which had resumed publication in Oakland on April 22, reported Dimmig’s 498 as the official death toll while noting in a sidebar that military authorities place the figure much higher. The Examiner, also operating from temporary quarters, led with Byrne’s estimate and added that the true number may never be known, so complete was the destruction in the worst-hit districts. The same day’s Oakland Tribune ran a headline, DEATH LIST SWELLS TO 2, 000, that cited no source at all. The numbers were becoming autonomous, detached from the counting processes that had produced them, circulating as claims in a debate that exceeded their empirical basis.

Behind these conflicting tallies lay specific institutional interests that shaped what could be counted and how. The city’s government, operating under extraordinary pressure from insurance interests and from the national business community, had strong reasons to emphasize fire over earthquake. Mayor Schmitz had already issued his famous proclamation of April 18, authorizing the shooting of looters, in an explicit effort to demonstrate that civil authority remained functional. A high earthquake death toll would suggest that the disaster had overwhelmed governmental capacity from the first moments, undermining this claim of continued competence. The Committee of Fifty, which had effectively supplanted the Board of Supervisors in day-to-day governance, was dominated by businessmen with direct stakes in the insurance question. Their preliminary reports to eastern financiers emphasized the fire’s destructiveness while minimizing seismic damage.

The Army’s institutional interests ran differently. Funston’s command had been activated without clear legal authority—the President had not declared martial law, and the Governor had not formally requested federal assistance—and Funston was sensitive to charges of military overreach. A high casualty count, attributed to conditions that preceded military intervention, strengthened the case that extraordinary measures had been necessary. The Army’s reports emphasized the chaos of the first forty-eight hours, the inadequacy of civilian firefighting, the collapse of water supply. These were arguments for the necessity of military control, and they required a substantial death toll to be persuasive.

The relief organizations occupied a middle position, responsive to both imperatives. Their fundraising required dramatic numbers; their distribution of aid required careful verification. The Red Cross ledgers show this tension in their very form: the cause of destitution column allowed for earthquake attribution, but the amount granted column was calibrated to visible need rather than to the severity of the precipitating event. A family whose house had cracked but not burned received less aid than one whose house had disappeared entirely, regardless of which disaster had initiated the destruction. The counting mechanism channeled sympathy toward certain kinds of loss and away from others.

The insurance industry, though not directly present in these counting exercises, exerted decisive influence through its representatives on the Committee of Fifty and through its national organizations’ communications with the press. The National Board of Fire Underwriters issued a bulletin on April 27 that was widely reprinted: The great San Francisco fire, originating from causes incident to the earthquake, has resulted in losses unprecedented in American history. The earthquake itself caused negligible damage to properly constructed buildings. The bulletin cited no evidence for this claim about construction quality. It established a narrative template that would govern subsequent litigation: the earthquake was a triggering event, not a destructive one; the fire was the proper subject of insurance coverage; policyholders who had suffered structural damage without burning had no claim.

This template directly contradicted the experience of many policyholders, particularly in the residential districts south of Market and in the Western Addition. The relief ledgers contain numerous entries describing houses that had fallen off foundation or walls cracked beyond repair without ever catching fire. These families faced a double loss: uncompensated structural damage and, often, denial of temporary housing assistance on the grounds that they had not been burned out. The counting systems that separated earthquake from fire thus created a stratified population of the displaced, sorted by the category of their destruction into different queues for aid and different prospects of insurance recovery.

The mechanism of this sorting becomes visible in a comparison of two specific cases from the Red Cross records.

On April 29, a team of volunteer interviewers visited the Presidio refugee camp to assess applications. The first case, recorded at 10:15 a.m., involved a family of four who had occupied a frame house on Shotwell Street in the Mission District.

The interviewer noted that the house burned completely, the family escaped with clothing only, the cause was fire following earthquake, and the amount granted was $50 temporary relief with referral for permanent housing. The second case, recorded at 10:47 a.m., involved a family of three from a brick apartment building on Valencia Street.

The interviewer noted that the building was standing but cracks in walls rendered it unsafe, the family was removed by military order, the cause was earthquake damage, and the amount granted was $15 temporary relief with no referral. The difference in treatment, $50 versus $15, permanent housing assistance versus none, derived entirely from the categorical assignment of cause. Yet both families were homeless, both had lost the use of their residences, both faced identical prospects of winter weather in canvas tents.

The Red Cross clerks who made these assignments were not acting arbitrarily. They were following instructions developed by the Committee of Fifty’s subcommittee on relief, which had determined that fire refugees should receive priority in permanent housing reconstruction while earthquake refugees should be encouraged to repair existing structures. The distinction served the city’s broader interest in rapid rebuilding: burned districts offered cleared land for new construction, while damaged-but-standing buildings required individual negotiation with owners. The counting mechanism channeled resources toward the first category and away from the second, shaping the physical reconstruction of the city through its categorical distinctions.

By the first week of May, these various counting exercises had produced a characteristic pattern of official numbers: low and precise for earthquake deaths, high and general for fire destruction, variable and contested for total casualties. The State Board of Health, which had begun its own independent survey on April 24, reported on May 5 that the death toll from direct earthquake effects cannot at this time be determined with accuracy, but is believed to be less than five hundred. The same report estimated total deaths from all causes, including fire and subsequent exposure, at not less than three thousand. The range, 500 to 3, 000, with the lower figure attributed specifically to the earthquake, would persist in official discourse for decades.

The mechanism that produced this range was not merely empirical uncertainty, though uncertainty was genuine and extensive. Different institutions selected and emphasized different aspects of the disaster according to their functional requirements. The city’s government needed a contained, comprehensible catastrophe to maintain legitimacy and attract investment. The Army needed a substantial death toll to justify extraordinary intervention. The relief organizations needed dramatic numbers for fundraising and manageable categories for distribution. The insurance industry needed a clear causal separation to limit liability. Each requirement shaped the counting process, and the resulting numbers reflected these shaping pressures more than they reflected any underlying demographic reality.

The consequences of this process extended far beyond the immediate emergency. The official death toll of approximately 3, 000, later reduced in many accounts to fewer than 500 earthquake deaths, entered historical memory as established fact. It supported the city’s subsequent narrative of rapid, successful recovery: a community that had suffered a terrible fire but avoided the devastating mortality of comparable disasters. It shaped the legal environment for insurance litigation, where courts generally accepted the fire as proximate cause theory and limited recovery for earthquake damage. It influenced building codes and urban planning, where the emphasis on fireproof construction addressed the visible catastrophe while neglecting seismic resilience.

The alternative numbers, the higher estimates that circulated in the first weeks and were subsequently forgotten, left traces in less visible records. The Army’s burial logs, preserved in the National Archives, document approximately 1, 400 interments in the Presidio cemetery through May 1906, many of them unidentified. The coroner’s office maintained a supplementary list of presumed dead, persons reported missing whose bodies were never recovered, that exceeded 1, 000 names. These documents did not disappear; they were simply excluded from the official narrative that coalesced in the following months. The counting process had not merely produced numbers; it had produced a selection of which numbers would count.

By mid-May, the various counting institutions had begun to converge on a common framework that would govern subsequent reckoning. The Committee of Fifty’s final report of May 15, which would be widely circulated as the authoritative account, adopted the formulation that the earthquake caused the fire, but the fire caused the destruction. This elegant causal chain solved multiple problems simultaneously. It acknowledged the earthquake sufficiently to explain why the fire had been uncontrollable, broken water mains, collapsed fire stations, the death of Chief Sullivan, while attributing the actual damage to flame. It provided a basis for insurance recovery that the industry could accept. It offered a narrative of civic resilience: San Francisco had not been destroyed by geological forces beyond human influence, but by a fire that could have been fought more effectively, and would be prevented in the future through better preparation.

The convergence on this framework did not resolve the underlying tensions. It displaced them into other arenas: the litigation that would occupy courts for the next decade, the political struggles over reconstruction policy, the historiographical debates that continue to the present. The first official counts of the ruins had established that quantification was not neutral, that the categories of disaster were themselves stakes in a larger conflict. The numbers that circulated in the last week of April 1906 were not simply measurements of what had happened. They were interventions in what would happen next.

The stark, contradictory totals published by different authorities established that the battle to define the disaster’s scale and cause had been joined. A clerk in the Relief Committee offices, copying figures from interview forms into a master ledger on the evening of April 30, made a marginal notation that captured the emerging awareness: Many cases do not fit the categories. He did not specify which cases, or what categories failed to contain them.

The observation stood alone, without elaboration, in a record that would otherwise be read only for its aggregate totals. It marked a recognition that would not be officially acknowledged: that the counting systems designed to manage the disaster were themselves producing a new form of loss, as experiences that failed to match the available categories were rendered invisible in the very process of being recorded. The clerk finished his entry, turned the page, and continued his work. The totals accumulated.