Detroit’s Michigan Avenue Murders: John Eric Armstrong and Three Systems That Failed

Before Michigan Avenue

John Eric Armstrong was born November 23, 1973, in New Bern, North Carolina. His childhood was defined by instability. He later claimed that his biological father was abusive and sexually assaulted him as a child. His younger brother Michael died of sudden infant death syndrome in 1978, at two months old. Armstrong attempted to take his own life in the aftermath of that loss. His father left shortly after. He did not receive any formal mental health treatment until 1989, a full year after an incident in which he locked himself in a school bathroom because a female classmate was pressuring him into a sexual encounter.

He joined the U.S. Navy in 1992 at eighteen years old. He served as a petty officer third class aboard the USS Nimitz aircraft carrier for approximately seven years. His shipmates described him as moody, but he accumulated no disciplinary record. He was soft-spoken, boyish-faced, and largely unremarkable to the people around him. His fellow sailors called him Opie. He was honorably discharged in 1999.

After his discharge, Armstrong settled in Dearborn Heights, Michigan, a suburb approximately ten miles west of Detroit. He worked at a Target store, then found employment refueling aircraft at Detroit Metropolitan Airport. He was married. He had two young children. He drove a black 1998 Jeep Wrangler with a “Baby Doll” front license plate. He enrolled at Schoolcraft College in Livonia. He attended church. He was, by every visible measure, a young family man making a quiet life after the Navy.

What he told investigators after his arrest was this: every sex worker he had ever paid for sex, he had either killed or tried to kill. He said this was a rule he had set for himself. He said it had begun before he joined the Navy.

The Confirmed Victims

Five women are confirmed by criminal conviction to have been killed by John Eric Armstrong in the Detroit area between December 1999 and April 2000. Each of them is named here because naming them is the minimum the record requires.

Monica Johnson was the first. She was found beaten and unconscious on December 3, 1999, and died the same day. The manner of her death, at the hands of Armstrong, would not be confirmed until he confessed months later. In December 1999, she was another case in a city with many cases.

Wendy Jordan was thirty-nine years old. Her partially clothed body was found on January 2, 2000, in the Rouge River in Dearborn Heights. She had been strangled and thrown from a bridge. She had prior arrest records for soliciting in Detroit. Those records are what most public accounts mention about her. She also had a life that those records do not describe, which is true of everyone and worth stating plainly.

Robbin Brown was twenty years old. Her body was found April 10, 2000, in a railroad yard in southwest Detroit. She had been strangled. She was identified through fingerprints. She had prior arrest records.

Rose Marie Felt was thirty-two years old and a Detroit resident. Her body was found in the same railroad yard on April 10, in a more advanced state of decomposition than Brown’s, indicating she had been there longer. Ligature markings confirmed strangulation. She was also identified through fingerprints.

Kelly Jean Hood was thirty-four years old. Her body was found in the same railroad yard on April 10, the most decomposed of the three, indicating she was likely the first to die there. She had a son. His name is Kyle Cazares. He was nine years old when his grandmother told him his mother was gone. Her younger sister later said Kelly always had a heart of gold, no matter how things looked from the outside. Kyle said his mother had been discarded like trash. He was not wrong about what had happened to her. He was right that she deserved better than what she got.

Armstrong positioned the bodies of Brown, Felt, and Hood deliberately. He stated in his confession that he arranged them in sexually provocative poses so that he could return after the killings to engage in post-mortem sexual activity. He told investigators he was confident the bodies would remain where he placed them. He had calculated the timeline and acted accordingly. This is documented in court records and in the accounts of the investigators who took his confession. It is included here because it is part of what happened to these women, and what happened to them is the record.

The First Contact: January 2, 2000

On the evening of January 2, 2000, Dearborn Heights police received a call about a body in the Rouge River. The man who made the call was already at the scene when officers arrived. He explained that he had been driving when he felt nauseated, pulled over near a bridge, walked to the railing, and noticed the body in the water below while attempting to vomit. He gave his name as John Eric Armstrong.

The detective who interviewed him found his story plausible, if oddly delivered. Armstrong was insistent that he was not responsible for what had happened, and he pointed out repeatedly, without being prompted, that he had been the one to call for help. Investigators noted that this behavior was not unusual for someone trying to appear cooperative, but it was also not unusual for an innocent witness. Armstrong was released.

An autopsy the following day confirmed the victim as Wendy Jordan, thirty-nine, previously arrested for soliciting in Detroit. The cause of death was strangulation. A rape kit was performed. In 2000, rape kit processing at the Michigan State Police laboratory took several months. That timeline was not an anomaly. It was the standard pace. Investigators documented the case and waited for results.

Armstrong drove home to Dearborn Heights in his Jeep Wrangler. His wife and children were there.

The Policy That Extended the Count

At some point in the weeks following Jordan’s death, preliminary DNA results came back linking Armstrong to the murder. Fiber evidence from Jordan’s clothing also matched the interior of his Jeep. Dearborn Heights detectives had what they believed was sufficient cause to detain him. They went to the Wayne County Prosecutor’s Office and requested a homicide arrest warrant.

They were turned away.

The Wayne County Prosecutor’s Office maintained a standing policy: no arrest warrant would be issued for a homicide based on preliminary forensic results. The final laboratory report from the Michigan State Police had to be in hand before a warrant could issue. The policy existed for a defensible reason. Preliminary results carry more risk of error than finalized analysis. Warrants issued on incomplete evidence can be legally compromised and can expose prosecutions to reversal. The policy was designed to protect the integrity of the case.

The consequence of that policy, in the documented sequence of this case, is that three more women died.

The Wayne County Prosecutor’s Office was not the only institution that touched this case in the interval between the preliminary evidence and the arrest. But it is the institution where a decision was made that can be documented, named, and evaluated. The policy had a name. It had a rationale. And in this case, it had a consequence.

Whether three specific deaths can be causally attributed to that policy in a strict legal sense is a separate question, one with real complexity. The documented record does not require a simplified answer to still be disturbing. Armstrong had preliminary DNA linking him to a murder. He had committed that murder with a method he had used before and would use again. The policy that kept him free was operating exactly as designed.

The Investigation Closes In

The discovery of three bodies in the railroad yard on April 10 moved the investigation quickly. The Detroit Police Department’s Violent Crimes Task Force connected the railroad yard deaths to the earlier Jordan murder through DNA evidence. One person had committed all of the known killings. The signature was consistent: sex workers, strangulation, the Michigan Avenue corridor.

Devon Marcus had already given investigators a name. Armstrong was identified, located, and arrested on April 12, 2000. As officers placed him in a patrol car, a woman approached the scene and identified him as the man who had recently attacked her. That woman was a second surviving victim, in addition to Marcus.

In the interrogation room, a detective from the Wayne County Sheriff’s Office sat across from Armstrong and told him, with deliberate specificity, that biological material Armstrong had left on the detective during the interview was now in the investigators’ possession, and that investigators had considerably more of Armstrong’s DNA from the crime scenes. Armstrong, twenty-six years old, broke. He confessed to the killings of all five women. He described picking each victim up on Michigan Avenue, engaging in a sex act, and strangling them with his hands or with the victims’ own clothing. He described the compulsion as something he could not stop.

He was formally charged on April 14, 2000, with five counts of first-degree murder and three counts of assault with intent to commit great bodily harm.

In March 2001, Armstrong was convicted of first-degree murder for the killing of Wendy Jordan and sentenced to life in prison without the possibility of parole. He was subsequently convicted of first-degree murder for the killing of Kelly Jean Hood, receiving a second concurrent life sentence. He pleaded guilty to three counts of second-degree murder for the killings of Robbin Brown, Rose Marie Felt, and Monica Johnson. He was resentenced in August 2003 to concurrent terms of 31 to 70 years on those three counts, to run concurrently with the two life sentences.

The Visibility Deficit

The Armstrong case did not create the conditions that allowed him to operate. It exposed them.

All five confirmed victims were women with prior arrest records for soliciting. Monica Johnson died in December 1999. Wendy Jordan died in early January 2000. The bodies of Brown, Felt, and Hood were found in April 2000. The interval between the first confirmed death and the railroad yard discovery was approximately four months. During that time, Armstrong was operating openly on Michigan Avenue, a corridor where sex workers were known to gather, in a distinctive Jeep Wrangler with a personalized license plate.

Investigators who reflected on the case years later were direct about the dynamic. Sex workers are frequently targeted by killers, in part, because the investigative response to a missing sex worker is structurally different from the response to a missing person without that history. Families are assumed to be accustomed to unexplained absences. Disappearances are logged without immediate investigation. Connecting cases into a pattern requires someone to be actively looking for the pattern, which requires resources allocated toward a population that institutions do not consistently treat as a priority.

This is not an accusation specific to the Detroit or Dearborn Heights investigators who worked the Armstrong cases. Many of them worked the cases diligently and closed them as quickly as the available forensic infrastructure allowed. It is a description of the structural reality those investigators were operating within. The rape kit processing timeline alone, several months from submission to result in 2000, was an infrastructure failure that preceded and outlasted the Armstrong investigation. Michigan has since moved to address rape kit backlog through legislative action and policy reform. The Armstrong case is among the documented reasons why that reform became necessary.

The structural condition is this: when victims are drawn from populations that institutions have deprioritized, killers who target those populations operate in a wider margin. Armstrong knew where to find women who would not be quickly missed. That knowledge was not an accident. It was the same knowledge the system had, pointed in the opposite direction.

The Navy Years: A Confession Without a Corroboration Infrastructure

After confessing to the five Detroit murders, Armstrong told investigators that his killing had begun earlier. He said the first victim had been a woman in North Carolina in 1991, before he joined the Navy. He then described eleven additional murders committed during his years of service aboard the USS Nimitz.

He named cities. Seattle: three victims. Honolulu: two. Hong Kong: two. Virginia and North Carolina: one each. Thailand: one. Singapore: one. He claimed he had followed a consistent pattern: at each port where the Nimitz docked, he had paid for sex and then killed the woman. He described it not as an escalation but as a continuation. The Detroit killings, in his account, were not the beginning. They were the end of a seven-year sequence.

Investigators forwarded the information to law enforcement in each named jurisdiction. What they found was nearly nothing.

Police in Hawaii searched their records and found no cases matching Armstrong’s description. Law enforcement in Thailand similarly found no matches. Israeli investigators examined a potential link to a murder in Haifa, but confirmed that Armstrong had not been stationed in Israel at the time the killing occurred. The Haifa link was discounted.

The Seattle claims were undermined by Armstrong himself. Between his first and second accounts, the description of one Seattle victim changed in a fundamental way. What he initially described as a transgender woman became, in his revised account, a man who had approached him seeking money. The shift in description was significant enough that investigators concluded the Seattle claims could not be credited. If the Seattle account was fabricated or conflated, the reliability of the other international confessions became correspondingly uncertain.

Armstrong could not be conclusively linked to a single international murder. No charges were filed. The confessions produced no additional convictions.

Whether Armstrong killed eleven people across multiple countries, or fewer, or none outside Detroit, is a question the available record cannot resolve. What the record does resolve is the absence of any architecture that could have answered it. A confession spanning a decade of port-city killings in seven countries was handed to investigators who had no common database, no coordinated protocol, and no institutional mandate that crossed jurisdictional lines. The confession sat in each city’s file and went nowhere.

This is not a failure specific to any single investigator or agency. It is a structural gap that the Armstrong case made visible and that has not been specifically closed in the years since.

What the Record Confirms and What It Does Not

Five women are confirmed dead at the hands of John Eric Armstrong: Monica Johnson, Wendy Jordan, Robbin Brown, Rose Marie Felt, and Kelly Jean Hood. Those five deaths are established by criminal conviction. Everything beyond them is documented but unresolved.

The Wayne County Prosecutor’s warrant policy has not been a sustained subject of public policy reform discussion since 2000. Whether that policy remains in its original form is a question for the current Wayne County Prosecutor’s Office. The population it is most likely to affect, women engaged in sex work with prior arrest records, remains the same population it affected in 2000.

The rape kit backlog that delayed DNA results in this case has been the subject of Michigan legislative action, including reforms that accelerated statewide testing and established tracking requirements. The international coordination gap that prevented verification of Armstrong’s port-city confessions has not been directly addressed by any policy mechanism that this record can identify.

Three systems intersected in the Armstrong case: the prosecutorial warrant standard, the investigative priority structure for sex worker disappearances, and the absence of cross-jurisdictional military port-city homicide coordination. None of them created Armstrong. All of them shaped what he was able to do, for how long, and at what cost to the documented count of the dead.

Armstrong is fifty-two years old. He is housed at a medium-security facility. He will not be released. The five women he was convicted of killing are still dead, and the thirteen he confessed to killing elsewhere are still unaccounted for in any official record, in any country, in any database that connects the cases. That gap is the record too.