Waco: What the Inquiries Could and Could Not Establish

AI-generated editorial illustration for The Truth Files. The building and landscape are symbolic; this is not a photograph of Mount Carmel or a historical record.

At Mount Carmel near Waco, Texas, a federal firearms raid on 28 February 1993 turned into a gunfight. Four ATF agents and six Branch Davidians died that day, according to the later Danforth inquiry. A 51-day siege followed. On 19 April, the FBI drove armoured vehicles into the building to deliver CS gas. Fires consumed it. At least 76 more Branch Davidians died, including children. These are different events, with different evidence and different decisions to investigate. ^{[7] [9] [15]}

The enduring question is often compressed into one accusation: did the government cause Waco? That cannot be answered responsibly with a single yes or no. The record can distinguish who likely lit the fire from whether the final operation was wise, and from whether officials told the truth about the devices used. Those are four separate questions. ^{[4] [6] [7] [8]}

The file at a glance

• 28 February: valid warrants did not make the raid plan sound. Treasury faulted a plan dependent on surprise after commanders had been warned that Koresh expected them. Who fired first remains historically contested. ^{[1] [6] [10]}
• 19 April: the original fire team concluded that fires were deliberately set at separate points inside. Danforth upheld internal ignition and rejected government ignition. A serious alternative hypothesis disputes the fire reconstruction. ^{[4] [7] [14]}
• Judgement: Danforth investigated fire spread and concluded federal action did not materially contribute to it or to the deaths. He did not decide whether the gas plan was wise. The original fire team recorded that breaches and wind contributed to spread and could not establish how many people could escape. ^{[4] [5] [7]}
• Disclosure: three pyrotechnic CS rounds were fired at a separate construction pit hours before the fire. Their use was obscured in the public record for years. Establishing that failure does not establish that those rounds ignited the building. ^{[7] [8] [17]}
• The dead: Danforth counted at least 82 Branch Davidian deaths across both dates, excluding two fetuses; four ATF agents died on 28 February. Exact counts vary with date, age threshold and identification. ^{[7] [9] [15]}

Central question: what do the surviving public inquiries actually establish about the decisions and deaths at Mount Carmel, and what remains outside their evidence or their authority? The answer matters because two opposing simplifications have shaped public memory: that a flawed raid proves federal agents set the fire, and that a finding of internal ignition vindicates every federal decision. Neither inference follows.

This investigation asks who the community was; what warrants authorised and whether the raid design held up; what can be said about the first gunfire; why the siege ended with CS gas; where the fatal fires began and how they spread; what the pyrotechnic disclosure changed; what the courts decided; and how many people died under each counting rule. Each question is answered in the named section and graded in the Evidence Ledger. The cutoff is 25 October 2026. This is an inquiry into a United States event and its US federal and Texas evidence record, not a general theory of religious movements or a claim that later radicalisation elsewhere proves anything about the fire. ^{[1] [4] [6] [7] [9] [10]}

The question needs more than a chronology. A fire investigator can identify an origin while leaving the appropriateness of a gas plan outside the report. An appeal can decide what instruction a jury should receive without settling every witness conflict for historians. A disclosure inquiry can uncover false or missing statements without proving the suppressed item caused the deaths. The map below keeps those forms of evidence in their own lanes. ^{[4] [7] [10]}

The dates show where a choice or a new finding changed the case. They do not imply that the next event was inevitable.

Mount Carmel was a religious community, not merely the backdrop to a police operation. The Branch Davidians grew out of a Seventh-day Adventist offshoot. Their leader, Vernon Howell, who took the name David Koresh, read the Bible's Seven Seals as a prophecy he was uniquely placed to explain. The community included men, women and children who lived, worked and worshipped at the property. Its beliefs mattered to how members understood a siege and how negotiators' words were heard. ^{[13] [16]}

Religion scholars warned federal officials against treating a strongly apocalyptic group as a simple hostage case. Nancy Ammerman, who was asked to review the confrontation after the fact, argued that officials lacked useful knowledge of the group's religious world and that the advice they received was not integrated well enough into planning. Her report is expert interpretation, not an eyewitness account of the first gunshot or a forensic fire study. ^[13]

Koresh's conduct and the community's beliefs must also be kept separate from the rights and choices of individual residents. The presence of children, reported abuse and weapons concerns were serious reasons for investigation. They do not make every person inside a combatant, or prove that every adult wanted to die. ^{[1] [5] [9] [14]}

ATF investigated allegations of illegal firearms and explosives and obtained search and arrest warrants. Treasury's review accepted the warrant basis but sharply criticised the tactical choice. Its investigators said alternatives, including an arrest away from Mount Carmel, were dismissed too quickly. The planned dynamic entry depended heavily on surprise. On the morning of the raid, undercover agent Robert Rodriguez reported that Koresh knew agents were coming. Commanders proceeded. The existence of a lawful warrant therefore answers a different question from whether this was the safest way to execute it. ^[1]

When ATF trucks reached the building, a firefight began. Treasury reconstructed the first shots as coming from inside. In the criminal case United States v. Branch, a Fifth Circuit majority held that the trial evidence did not support the defendants' proposed self-defence instruction and treated Davidian first fire as the only reasonable inference on that record. Judge Schwarzer's dissent said conflicting testimony and witness credibility should have been left to the jury. A House majority later called Davidian first fire likely but said its limited testimony could not settle the point conclusively. ^{[1] [6] [10]}

Assessment: the first-shot question remains contested as an independent historical reconstruction. The appellate ruling decided a jury-instruction issue in a particular prosecution. It does not erase the House's caution or establish that any named convicted defendant fired the opening shot. The planning failure matters even if one accepts the Treasury sequence: the operation went ahead after the warning that its central tactical advantage had been lost. ^{[1] [6] [10]}

The FBI took over after the failed raid. Negotiators made contact and people left Mount Carmel, while Koresh gave religious reasons for delaying surrender. Tactical pressure and negotiation ran alongside each other. The Justice Department's 1993 factual report describes the FBI's choices; Edward Dennis's separate evaluation judged the handling of the standoff, relying largely on that factual record. They are different kinds of review, not two independent witness pools. ^{[2] [5]}

Dennis recorded that negotiators objected to some pressure tactics immediately after people had come out, fearing the tactics damaged trust. He nevertheless gave priority to tactical safety and judged the final plan reasonable. In the FBI's account, an indefinite cordon was dangerous and impractical: officials thought the community had enough supplies for a year, agents were fatigued, and a gun battle, explosion or other disaster could occur while they waited. Reno received a detailed briefing on 12 April, sought more information about CS effects on children and pregnant women, heard further advice on 14 April, and approved the plan on 17 April. The proposed gas insertion was staged over as much as 48 hours, not presented to her as a rapid demolition. Dennis concluded that all reasonable alternatives had been considered. Those are his findings and the reported decision sequence, not proof that waiting would fail. ^[5]

The House majority drew the opposite judgement: it called the 19 April choice premature, argued that the possible continuation of talks and Koresh's work on the Seven Seals had been discounted, and said the claimed urgency was overstated. It also examined conflict between tactical commanders and negotiators. The House dissent challenged the majority's assignment of fault. Ammerman's review adds a narrower process criticism: religious expertise was available, but not integrated adequately into decision making. These sources do not show that a peaceful exit was certain. They do show why the timing and method of the final operation remain a legitimate dispute even after the fire origin is assessed. ^{[5] [6] [13]}

The conflict was not abstract. Dennis records that electricity was cut on 12 March after two residents left, and vehicles were moved on 21 March after seven more exits, despite negotiators' objections. He says no one left after 23 March and he could not be confident that gentler pressure would have brought more out. The House majority pointed to the gas plan's contingency: if residents fired, agents could rapidly insert much more gas than under the slower plan Reno had heard. It argued that commanders expected this response and that Reno had not adequately weighed it. Dennis judged her briefings and the alternatives adequate; the House majority did not. ^{[5] [6]}

Assessment: the record supports criticism of coordination and risk communication, and it does not demonstrate that 19 April was the only viable day to act. It also does not prove that waiting would have brought everyone out safely. The danger of weapons, fatigue and a long cordon was real in the FBI's decision record; the House's evidence about a still possible religious route to surrender and the likely contingency made the proportionality judgement genuinely contested. Danforth's later finding on fire origin cannot decide it. ^{[5] [6] [7] [13]}

On 19 April, armoured vehicles broke openings in the building and delivered CS gas over several hours. The FBI plan envisaged pressure to make residents leave, but the gas operation and the breaches altered the building before the fires. They therefore belong in any account of consequences even if agents did not intentionally set it alight. ^{[3] [4] [5]}

The original July 1993 fire-scene report, reproduced by the Justice Department, identified three separated areas of origin seen on infrared imagery between 12:07:41 and 12:09:45. Its team used scene examination, imagery and laboratory findings of petroleum products and concluded the fires were deliberately started by people inside. Danforth later added court-authorised recordings from inside the building, including conversations about fuel before flames appeared. His report also attributes to surviving Davidians Graeme Craddock an account of seeing fuel poured and to Clive Doyle an account that fuel was placed in designated locations; investigators cited agent observations and further fire analysis. The recordings required enhancement and Danforth's expert did not claim certain speaker identification for all voices. These are distinct evidence types, though the survivor statements reach us here through the inquiry rather than direct interview transcripts. The Justice Department's 1993 report and Dennis's evaluation also describe the original arson team's work; repeating that account does not make three independent fire investigations. ^{[2] [4] [5] [7]}

There is material counterargument. Sociologist Stuart Wright challenged the confidence with which an internally planned collective death has been inferred. Citing fire expert Richard Sherrow's 1996 declaration, he argued that a vehicle striking a fuel lantern could possibly cause an accidental ignition and that separate infrared heat signatures need not necessarily mean separate starting fires. The full filed Sherrow declaration and underlying imagery were not available to this review; Wright's account is a substantial challenge, but a physical possibility is not proof that it happened. The later House inquiry recorded both the alternative and the government's objection to its evidential basis. ^{[8] [14]}

The original fire team's own qualifications deserve equal prominence. It said the FBI's wall breaches, combined with strong winds, contributed to spread. It could not exclude some spilled fuel from the breaches. It could not determine how many people could have escaped, and considered several possibilities, including people prevented from leaving inside or a blocked route. Danforth explicitly investigated spread: he acknowledged that openings may have accelerated it in some areas, but concluded the government did not materially contribute to spread and that the openings and local acceleration did not contribute to deaths. He said the openings were intended to deliver gas and create exits, and some people used them to escape. The two reports thus make different kinds of statements: one identifies a contributory physical factor and uncertainty; the later inquiry rejects material contribution and death causation on its review. ^{[4] [7]}

Assessment: deliberate ignition within the building is the best-supported conclusion of the accessible forensic inquiries. The alternative remains a serious reason to describe the conclusion as an inquiry finding, rather than absolute knowledge of every ignition mechanism. Neither position proves that all residents chose death or could leave. ^{[4] [7] [14]}

The 1993 public account did not clearly disclose everything fired by the FBI that morning. In 1999, evidence emerged that an agent had fired three pyrotechnic CS rounds at a concrete construction pit away from the living quarters at about 8:08 a.m. Attorney General Janet Reno ordered a new investigation. Danforth found the rounds landed roughly 75 feet from the quarters, around four hours before the building burned, and did not start the fatal fires. This finding must not be inflated into the claim that no pyrotechnics were used. ^{[7] [17]}

The missing disclosure was real. Danforth described inaccurate statements and failures to pass material to prosecutors and investigators. He found that some prosecution leaders knew of the rounds in 1993 and failed a legal disclosure duty, while rejecting the charge that Reno knowingly concealed them. A later House majority and minority also accepted problems with the earlier reviews but differed over individual responsibility and the degree of concealment. The House report was adopted on 19 October 2000, before Danforth's 8 November final report, though ordered printed in December. It should not be read as a response to all of his final findings. ^{[7] [8]}

Assessment: the use and prolonged failure of clear disclosure are documented. The fire-causation allegation about those three rounds was tested and rejected by Danforth. The exact intent and responsibility behind every omission remain disputed. The disclosure failure damaged trust for good reason; it cannot by itself answer how the fire began. ^{[7] [8]}

Flashes on airborne infrared footage became another allegation: that agents fired into the building during the final operation. Danforth examined the infrared recording with tests and expert analysis, and compared it with ballistics, witness accounts, other video and pathology. He found no government gunfire into the building that day. The later House inquiry reported an early analyst's contrary gunfire interpretation, but said he died before producing a scientific report; a second analyst did not identify the flashes as gunfire. The underlying recordings and a full adverse technical report were not independently inspected for this article. ^{[7] [8]}

Assessment: no substantiated 19 April government gunfire finding emerged from these inquiries. That conclusion concerns the final day. It says nothing about which side fired first in the separate 28 February gunfight. ^{[1] [7] [8]}

These investigations did not ask identical questions. Reading them as a single collective verdict creates false certainty. The table is a map of authority and evidence, not a vote count.

The same distinction applies to the courts. At the criminal trial, the Davidian defendants were acquitted of conspiracy to murder the agents; five were convicted of aiding and abetting voluntary manslaughter of federal agents, and firearms convictions followed. The Fifth Circuit in Branch affirmed convictions in part, rejected a proposed self-defence instruction on the trial record, and vacated some firearms sentences for further findings. The Supreme Court's later Castillo decision held that the machinegun provision was an offence element for a jury, not merely a sentencing factor; it did not decide the first shot or the fire. In the civil case brought by survivors and families, the plaintiffs took nothing. The Andrade appeal affirmed that outcome mainly on recusal and procedure, without relitigating every substantive liability finding. Acquittals, convictions and a failed civil claim cannot be combined into an all-purpose judicial verdict on the event. ^{[10] [11] [12]}

Numbers move when a source changes the date, age rule or stage of identification. The Justice Department's early 1993 report counted 75 people recovered after the April fire, including 25 children under 15, and referred to five February Davidian deaths. The later Danforth report said at least 82 Branch Davidians died across both dates, six in February and at least 76 in April, excluding two fetuses. Four ATF agents also died on 28 February. Its use of at least matters: burning and commingling prevent a perfectly exact final total. ^{[2] [7] [15]}

Pathologist Michael Graham's review for Danforth counted recovered remains as 54 adults, 28 children aged 18 or under, and two fetuses. Six adult remains were assigned to February, leaving 48 adults and 28 children among the recovered April dead. The 28 and the Justice Department's earlier 25 are not competing answers to the same age question: one includes people through 18, the other counted under 15. Graham also noted possible additional remains and identification limits. His report is reproduced on a third-party site; a final county case-by-case register was not obtained for this article. ^{[2] [7] [9]}

No number should turn the dead into an undifferentiated crowd. The available pathology and inquiry findings do not establish each person's mechanism of death, knowledge of the fire, ability to escape or consent. Children could not meaningfully consent to such danger. ^{[4] [9] [14]}

This file was researched to 25 October 2026. We read substantial relevant portions of the Treasury report; the official Justice Department HTML report and the original fire-scene appendix; Dennis's evaluation; the 1996 and 2000 House majority and minority reports; Danforth's final report; Graham's pathology review; the cited appellate decisions; Nancy Ammerman's report; and Stuart Wright's full 2009 challenge. The central search intents were checked against accessible ranked articles before the manuscript was written. Search volume was not verified. We did not certify every page of these long reports as read. ^{[1] [2] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]}

Sources share evidence. The 1993 Justice Department report, Dennis and the fire-scene team are linked; Danforth added expert work but also reviewed that earlier record. A count of agreeing reports is not a count of independent observations. House majority and minority views are attributed separately. Scholarly accounts are interpretation, even when they cite primary material. ^{[2] [4] [5] [6] [7] [8] [14]}

We could not inspect the original infrared footage, Sherrow's complete filed declaration, the full trial transcript, the complete civil record, or a final county death register. Those limits prevent a fresh forensic verdict on each flame, shot or death. They do not prevent a bounded assessment of what the named inquiries found and what each was empowered to decide. Medical or chemical claims about CS exposure have been left out because the record reviewed here is not sufficient for a specialist toxicology conclusion. ^{[4] [7] [9] [14]}

The cover is an AI-generated cinematic editorial illustration of a generic Texas landscape and inquiry files. It is not a photograph of Mount Carmel or historical evidence. No image of a dead person or a fabricated evidence scene is used.

Waco was not one failure with one evidential answer. Treasury found a raid plan that continued after surprise was lost. The first shot remains contested as an independent historical question. The original fire investigators and Danforth found powerful evidence of deliberate ignition inside Mount Carmel, while an adverse expert hypothesis and the inaccessible underlying materials limit absolute certainty. The original fire report says federal breaches and wind contributed to spread and could not establish how many people could escape. Danforth assessed spread and found no material government contribution or resulting deaths from those openings; his finding must be read alongside the earlier report's physical observations. ^{[1] [4] [6] [7] [10] [14]}

The later revelation of three pyrotechnic rounds exposed serious failures of disclosure. Danforth rejected those rounds as the cause of the fire, but his inquiry did not decide whether the gas operation was sound. A forensic exoneration on a narrow question must not be sold as a vindication of every tactical decision. Nor does criticism of the operation prove that agents set the fire. ^{[5] [6] [7] [8]}

Verdict: the best-supported fire-origin finding is deliberate ignition within the building; the choice and consequences of the final operation remain open to serious criticism; the nondisclosure is established; and no evidence reviewed here justifies treating every resident's death as a voluntary act. This is a judgement about the accessible record, not a claim to have solved every contested moment. The materials most likely to sharpen it are the original infrared recordings, complete adverse expert filing, full witness testimony and case-by-case death records. ^{[4] [7] [9] [14]}