The Black Widower:
Six Wives, Four Dead,
and the Record That
Finally Held
Thomas Randolph was convicted twice for the murder of his sixth wife and the hitman he hired to kill her. Four of his six wives are dead. Two survived. The record of what happened in between is a case study in what courts allow, what they exclude, and what happens when a pattern goes uncharged for thirty years because the documentation exists in different jurisdictions and nobody assembles it.
Thomas Randolph’s defense at both of his murder trials was simple. He is a romantic. He loves women. He has been unlucky. Six marriages, four dead wives, and whatever else happened in between is the cost of a man who keeps falling in love in a world that keeps taking the women he loves away from him.
The record disagrees.
The record shows a man who married on the day his divorce was finalized. Who collected life insurance payouts after each death. Who tried to hire a police officer to kill one wife, pleaded guilty to witness tampering in that case, and walked on the murder charge. Who shot a gun in the direction of a third wife and called it an accident. Who was in the hospital room of a fourth wife when she died after heart surgery, and whose family had her cremated within twenty-four hours. Who hired a handyman to stage a home invasion and kill his sixth wife, then shot the handyman to close the loose end, then called 911 to report the intruder he had “heroically” killed.
That is not a romantic. That is a pattern. The record documents it across three states and thirty years. The question this case raises is not whether Randolph is guilty. Two juries answered that. The question is how the pattern survived as long as it did, what evidentiary rules allowed it to keep surviving, and what a proper records review of a case like this actually looks like.
The Randolph case is a master class in how prior-bad-act evidence rules work, how they fail, and what happens when prosecutors push too hard and hand a defendant a reversal on appeal. The conviction stood the second time. But the road from 2008 to 2023 should not have been that long, and the reason it was is documented in the record.
What the Record Shows, Wife by Wife
Six marriages. This is the foundation of everything. Each marriage produced a life insurance policy. Each policy named Randolph as beneficiary. The deaths, the circumstances, and the outcomes are documented in police reports, court transcripts, coroner’s records, and insurance files spread across Utah and Nevada. Here is what the record shows.
Married Randolph at eighteen. He was twenty. The marriage lasted seven years and produced two children. Thomas testified at both of Randolph’s murder trials that he was controlling, manipulative, and psychologically abusive. She remembered the day their divorce was finalized: he married his next wife the same afternoon. She participated in the 2024 Investigation Discovery documentary with her face obscured, still unwilling to be fully identified. The fact that she survived and eventually testified is part of how investigators built the behavioral picture they needed.
Randolph married Gault on the same day his divorce from Kathryn Thomas was finalized in 1983. Three years later, she died from a gunshot wound to the head in their Utah home. Randolph told police she had killed herself. The death was ruled a suicide. He collected a $250,000 life insurance payout. Gault’s family and friends disputed the suicide finding, and investigators later developed a theory that Randolph had arranged her death. He was charged with her murder. He also hired a hitman to kill her before her death and, separately, tried to hire a police officer to kill her after her death to prevent her from testifying in some capacity, according to prosecutors. He pleaded guilty to witness tampering in connection with that latter conduct. At trial for the murder charge itself, he was acquitted in 1989. The acquittal would later come back to haunt prosecutors in the Nevada case in a specific way that is worth understanding.
Allmon testified at Randolph’s first murder trial that she believes he tried to kill her. She described an incident in which Randolph was cleaning his gun and fired it in her direction, leaving a bullet hole in their kitchen wall. He said it was an accident. She did not believe him. She left. Like Thomas, she later testified for the prosecution. Her survival and willingness to testify is one of the few things that worked as it should have in this case. She recognized the pattern early enough to get out. Most people in a case like this do not get that chance.
Francis had a pre-existing heart condition and underwent surgery. She died in the hospital afterward, in what was officially recorded as complications from the procedure. Randolph was present in the room before her death. Her family believed, and has continued to believe publicly, that he was responsible. The cremation occurred within twenty-four hours of her death, which eliminated any possibility of an independent forensic examination later. No charges were filed. The speed of the cremation, the life insurance payout, and the circumstances of the death were referenced in the Causse investigation but were not usable as prior-bad-act evidence at either trial under the Nevada Supreme Court’s eventual ruling.
Stapleton died of cancer. This is the wife whose death generates the least documentary controversy. The circumstances are not comparable to the others, and the pattern-of-conduct evidence in the Causse trial did not lean heavily on her. She is named in the record because she was married to Randolph and died while they were together. That alone is worth noting, but there is no documented theory of foul play in her case that has been publicly advanced by investigators or her family.
Sharon Causse met Randolph through an online dating site in 2005. They married quickly. By May 2008, she was dead. The official story from Randolph was a home invasion: they returned home, she went inside first, he found her shot by an intruder, he shot and killed the intruder in self-defense. The intruder was Michael Miller, a handyman Randolph had employed for odd jobs at the house. Detectives from the Las Vegas Metro Police Department thought the scene was wrong from the first interview. A video walkthrough Randolph recorded for investigators showed a man who discussed his dead wife’s body in clinical, unemotional terms. Miller’s presence at the house was explained by a bag of jewelry lying near him, suggesting a robbery. Investigators eventually concluded Randolph hired Miller to kill Causse, staged the home invasion, and then shot Miller to eliminate the only witness. Randolph collected nearly $400,000 in life insurance. He was charged with two counts of murder.
Two Trials, One Reversal, and What the Prior-Bad-Act Fight Was Actually About
The case did not go to trial until 2017, nine years after the murders. This is not unusual for a complex investigation, but nine years is a long time to build a case and the delay shaped what prosecutors tried to do at trial. The case they built was not just about Sharon Causse and Michael Miller. It was about Thomas Randolph as a man with a documented history of violence toward wives, failed murder plots, and insurance payouts. That history is what the prior-bad-act fight was about.
At the first trial in 2017, prosecutors introduced evidence about the Becky Gault case, including Randolph’s witness tampering plea and the circumstances of her death. They argued this evidence established motive, intent, and a pattern of conduct. The jury convicted on both counts. Randolph was sentenced to death.
Randolph appealed. In 2021, the Nevada Supreme Court agreed with him on one specific point.
The court found that the district court had improperly admitted prior-bad-act evidence from the Utah case involving Becky Gault, determining that the prejudicial effect of the evidence outweighed its probative value under the applicable evidentiary rules. The entire conviction was reversed and the case remanded for a new trial.
This is the part of the case that criminal defense attorneys and evidence scholars find genuinely interesting. Nevada Rule of Evidence 404(b) tracks the same framework as Federal Rule of Evidence 404(b): prior bad acts are inadmissible to show propensity, but may be admissible to show motive, intent, knowledge, identity, plan, or absence of mistake. The question is always the same: does the probative value of the prior conduct on a legitimate purpose outweigh its prejudicial effect on the jury?
In 2017, the trial court said yes. In 2021, the Nevada Supreme Court said no. The court found that the Utah evidence, specifically the facts of the Gault case and Randolph’s witness tampering conviction, was more likely to lead a jury to convict based on character than it was to establish the specific modus operandi or identity issues the prosecution claimed it addressed.
So the case came back. The retrial began in 2023. This time, prosecutors were constrained: much of the Utah evidence that had anchored their prior-bad-act theory was now off the table. They had to build the case around what the 2008 crime scene, the detective work, and the surviving witnesses could establish on their own. They did. A second jury convicted. Randolph was sentenced to sixty years.
The 2021 reversal does not mean Randolph was innocent. It means prosecutors overreached in their first trial and gave the Nevada Supreme Court a legitimate basis to send the case back. That is a costly mistake. It cost six additional years, a second trial, the ongoing expense of the appellate process, and the psychological toll on every surviving family member who had to go through it again. When you have a strong case, you build it cleanly. If the prior-bad-act evidence is genuinely necessary, you lay the foundation properly and you make the record of your balancing analysis airtight. They did not do that in 2017, and the court told them so.
What a Pattern-of-Conduct Case Actually Requires
The Randolph case is an extreme example of a category that comes up more often than people realize: a case where the individual incident is provable but the full picture requires stitching together conduct across years, jurisdictions, and unrelated proceedings. The more evidence you have, the harder it becomes to use it all.
This is not an accident of the rules. It is a feature. Prior-bad-act evidence is restricted because the alternative, letting a jury convict based on who a person is rather than what they did on a specific occasion, is a more serious threat to fair adjudication than the risk of letting a guilty person go. That principle is correct. The application to Randolph is what generates tension, because the pattern here is not a vague propensity. It is a documented, specific, repeating methodology: acquire wife, take out life insurance, arrange or benefit from death, collect payout. The jury that convicted him in 2017 saw that pattern and found it persuasive. The Nevada Supreme Court said the jury’s persuasion was the problem.
There is a version of this argument that applies to a lot of cases that never generate this level of appellate scrutiny. The discipline is the same regardless of scale. If you are building a pattern-of-conduct case, here is what the record needs to show:
The Life Insurance Structure as Documentary Evidence
The financial pattern in the Randolph case is as important as the physical evidence, and it is documented in a way that physical evidence often is not. Life insurance policies are paper. Payouts are paper. Beneficiary designations are paper. The timing of policy procurement relative to marriage dates, health changes, and deaths is reconstructable from public and commercial records.
In Randolph’s case, investigators estimated the total life insurance proceeds from all of his wives’ deaths at over a million dollars. That figure came from assembling policy documents across multiple deaths in multiple states over multiple decades. The $250,000 payout from Gault’s death in 1986. The nearly $400,000 from Causse’s death in 2008. The policy on Francis, whose death produced a payout that has not been publicly specified but whose existence was documented. The pattern of policy procurement is the kind of evidence that does not require a theory. It is just math. Each marriage produced a policy. Each policy named him as beneficiary. Each death produced a check.
Life insurance policy documentation can be obtained through civil discovery, public records requests in connection with estate proceedings, and subpoena in criminal cases. The procurement timeline, the beneficiary designation, and the payout records are all documentable. In a case involving multiple deaths where financial motive is part of the theory, the insurance record is the backbone. It is not circumstantial in the way that behavioral evidence is. It is transactional. Someone applied for the policy on a specific date, named a specific beneficiary, paid the premiums, and collected the payout after the death. That chain of transactions is reconstructable, and it is devastating when the pattern holds across multiple deaths and multiple insurers over decades.
The documentary challenge in the Randolph case is that this insurance evidence, while assembled by investigators, could not be fully deployed at the retrial in the same way it was at the first trial precisely because the prior-death evidence was constrained. You can introduce the Causse policy and the $400,000 payout as direct evidence of motive in the Causse case. But if you want to show that Causse was part of a pattern of insurance-motivated killings, you need the prior policies and payouts, and getting those in front of the jury requires clearing the prior-bad-act hurdle. When that hurdle gets higher, some of the documentary evidence becomes collateral damage.
05 · The SceneWhat Detective O’Kelly Saw That Randolph Did Not Expect
Homicide Detective Dean O’Kelly responded to the scene on May 8, 2008. His account is part of the trial record. He thought something was wrong before he had any evidence that it was. The 911 call. The body position. The bag of jewelry beside Miller’s body, which Randolph would have had to place there to support the home-invasion narrative. The relationship between Randolph and Miller, which became clear as soon as investigators identified Miller and started asking neighbors and contacts about his connection to the house. The walkthrough video, which Randolph agreed to record and which prosecutors played at trial. In it, Randolph discusses the crime scene in the affectless tone of someone who has processed the event rather than someone in shock.
The scene, in other words, told a story. But the story the scene told required interpretation. Detective O’Kelly’s interpretation was that the scene was staged. That interpretation required him to work backward from the physical evidence to reconstruct what had actually happened, which is a different exercise from working forward from an obvious crime scene. Staged crime scenes are harder to document, harder to prove, and easier to argue against at trial, because the defense’s entire theory is that the prosecution’s interpretation is wrong and the most natural reading of the physical evidence is the innocent one.
In this case, the prosecution’s reading won. Twice. But it took fifteen years from the date of the murders to the final sentencing in the retrial.
Staged crime scenes are a documentary challenge because the staging is itself evidence. The placement of the jewelry bag next to Miller’s body was not accidental. The decision to shoot Miller after he had already killed Causse, which is what prosecutors argued happened, required Randolph to manage two different narratives simultaneously: the story he told 911, and the story the physical evidence was actually telling. When those two stories conflict, the conflict is in the record. Every statement Randolph made, every interview, every walkthrough video, became a document in which his version of events was being compared to the physical evidence version. That comparison is how staged-scene cases get made. Build the timeline from the physical evidence first. Then put every statement on top of it and look for where they diverge.
How Prosecutors Won Without the Prior-Bad-Act Evidence
The 2023 retrial is the more interesting trial from a forensic standpoint, because it had to work with less. The Nevada Supreme Court’s 2021 ruling constrained what the prosecution could present. The Utah history, including the circumstances of Gault’s death and the witness tampering conviction, was off or severely limited. Prosecutors had to build a case around what the 2008 crime scene and the investigation of the Causse murder itself could establish.
They won on that record. The second jury convicted on both counts. The sentence was sixty years, down from death, because prosecutors chose not to seek the death penalty in the retrial. That decision allowed Randolph, in his allocution before sentencing, to claim it would allow him to “fast track” his appeal. He said he did not kill Sharon. Judge Tierra Jones sentenced him to two consecutive life sentences plus additional time for gun charges. He will die in prison.
His own children testified against him at the retrial. That is the piece that does not appear in the prior-bad-act analysis or the evidentiary arguments. His children watched what he did. They testified to what they knew. The family record, which is its own category of documentation, was part of what built the prosecution’s case at the retrial when the institutional record from Utah was no longer fully available.
“He’s never going to take responsibility. It’s always going to be someone else’s fault.”
What the Record Cannot Close
Thomas Randolph was charged with two murders. He was convicted of two murders. The deaths of Becky Gault, Francis Randolph, and, in a different way, the circumstances around the gun incident with Gayna Allmon, remain uncharged or unresolved.
Gault’s death was ruled a suicide in 1986. Randolph was acquitted of her murder in 1989 after pleading guilty to witness tampering. That acquittal cannot be retried in Utah under double jeopardy. What new forensic examination of the record might show about her death is no longer operationally relevant to a criminal case, because the criminal case is over. It matters to her family. It matters to anyone trying to understand the full pattern. It does not create a legal path to a new charge.
Francis Randolph was cremated within twenty-four hours of her death in 2004. There is no body to examine. There are hospital records, and those records are theoretically obtainable, but they document a death in a hospital after a surgery with a pre-existing cardiac condition. Building a murder case on that record, absent physical evidence, would require testimony from hospital staff who were present, documentary evidence of specific acts in the room, and a prosecution theory that could survive a reasonable doubt standard against a defense pointing to the obvious alternative explanation: she had a bad heart, she had surgery, she died.
The criminal justice system resolved what it could resolve. That is not the same as saying it resolved everything there is to know about this case.
The Randolph case illustrates something that comes up in pattern-of-conduct investigations at every scale: the record is bounded by what was preserved, what was charged, and what was adjudicated. Francis Randolph’s cremation within twenty-four hours eliminated the physical evidence that might have made a charge possible. Gault’s death being ruled suicide in 1986, before modern forensic reconstruction techniques, set the evidentiary baseline for everything that followed. The record holds what it holds. Working with it means understanding what it can and cannot establish, and being precise about the difference between what is documented and what is probable. Those are not the same thing. The criminal justice system requires proof. Pattern analysis operates on a different standard. Know which standard applies to what you are trying to do.
Reading the Randolph Record Without a Case Number
The Randolph case is over in the sense that matters most: he is in prison. He will not have another wife. The pattern is documented, adjudicated, and closed.
What the case teaches is applicable to investigations where the pattern is still in progress, where the charges have not been filed, or where a client or a subject has a record across jurisdictions that nobody has assembled into a coherent document. That assembly work is what investigative forensics does. It is not glamorous. It is not a single revelation. It is finding the Utah records and the Nevada records and the insurance company records and the hospital records and the court transcripts and the investigator’s notes and the walkthrough video and putting them in chronological order and looking at what they show when they are next to each other.
In Randolph’s case, that work took fifteen years to produce a final conviction. Some of that time was the investigation. Some was the trial. Some was the appeal. Some was the retrial. A non-trivial portion was the fallout from the prosecution’s overreach on prior-bad-act evidence in 2017, which handed the defense a reversal and required the entire proceeding to start over. The record of the case is also the record of how it was handled. Both are instructive.
The first time a jury convicted him, they had everything. The second time, they had less. They still convicted. The core case, the 2008 crime scene, Detective O’Kelly’s investigation, Miller’s connection to Randolph, the insurance policy, the staging of the scene, was strong enough to stand on its own. That is what you build first, before you reach for the pattern evidence. Build the case that stands alone. Then add the pattern, carefully, with the documentary foundation in place and the balancing analysis documented. The 2017 conviction failed on appeal because the extra layer was added without that foundation. The 2023 conviction held because the core case was strong enough to work without it.
The record is the source. It was in this case. It is in every case. The question is whether anyone reads it carefully enough, early enough, to use it well.
