Before California authorities turned to IGG—back when they relied only on traditional investigative methods—the Golden State Killer suspect list never shrank to fewer than several hundred names. There was no way to link the crime-scene DNA to any of the tens of thousands of men who had lived within or near the massive crime-spree field, which spanned nine California counties. Investigators would have needed an extremely lucky break, or a serious misstep by the killer, to narrow the suspect list.

And because the Golden State Killer apparently stopped his assaults in 1986, there had not been a break in some seventeen years, since the cases of the East Area Rapist and the Original Night Stalker—another name coined for the GSK—were connected in 2001.

But now, in just a few weeks, we were down to a suspect list of nine names.

It was a heady time for Paul Holes, but also a tricky one. There is something that happens when an IGG search gets closer and closer to the subject of the search—what I call the temptation to chase bad rabbits.

When you begin adding living people to a speculative family tree that you are building, you often have to turn to sources such as BeenVerified or TruthFinder for information about the person’s relatives and where they have lived. In my experience that information is generally reliable, although sometimes we have found conflated files (files for two people with similar names combined). So you have to remain alert and be thorough.

On occasion, the information from sources like BeenVerified includes references to criminal records. If you are working on identifying a suspect in a rape case and you see that the person in your tree has arrests for sexual assault, it can be tempting to focus on that individual. That temptation, I have seen, can be extremely strong. This phenomenon is sometimes called confirmation bias—trying to make the facts fit a preconceived conclusion. The temptation is to focus in on that one person without making sure that they fit all the criteria in the profile that you have developed for your suspect—criteria such as their estimated ethnic background from their DNA, their relationship to the suspect, their estimated age, and any known physical characteristics either from witnesses or from the DNA. When you succumb to confirmation bias, you also potentially are ignoring solid leads.

Or perhaps someone who is a potential suspect checks some of the boxes in the profile you created for your suspect but not all of them. That, too, can lead you to play a hunch.

But that is not what you are supposed to be doing in an IGG case. In such a search, there is no room for hunches, only identification of potential suspects based on the DNA.

Paul Holes followed a hunch when he seized on one of the names on our nine-man list as a potential suspect. Research showed that the man Holes focused on had owned property in Citrus Heights—the scene of many of the rapes. This man had also worked in the real estate business. He had a connection to Los Angeles, the scene of one of the Golden State Killer’s crimes, and he fit the age range we had established for our killer. All of that squared with the theory Paul had developed—that because so many of the Golden State Killer’s crimes occurred near new housing developments, the killer might have been connected to the housing or real estate industries. Paul focused on this man and became convinced he was the killer.

But there was a problem. While Paul’s preferred suspect did fit the profile he had assembled, he did not fit with what the DNA was telling us.

Our crime-scene DNA showed that the Golden State Killer’s bioancestry included 44.4 percent southern European heritage, which explained why we were seeing so many matches who were of Italian ancestry. When I researched Paul’s preferred suspect, however, I learned that his ancestors all traced back to the United Kingdom. His ancestry was purely English. In fact, he did not seem to have a single ancestor who had ever set foot in continental Europe.

Based on the DNA, Paul’s choice for a suspect could not be the killer.

I told Paul about this finding, but he responded that the admixture could be wrong, and he refused to allow the admixture report alone to rule out his man as a suspect. After all, he argued, his suspect checked so many other boxes in his profile of the Golden State Killer.

Paul had worked on the Golden State Killer case for a long, long time, and being this close to finding him, and especially so close to his own retirement, had to feel to him as if his man was right there, within his grasp, and all he had to do was reach out and grab him. No matter what I told him, I was not able to shake Paul off his hunch.

Even before Paul had focused on his chosen suspect, though, what turned out to be a potentially crucial piece of information had emerged.

This information was in an old newspaper article discovered by Sacramento County investigative assistant Monica Czajkowski, the member of Team Justice who was digging up documents relating to the nine men on our list. The short article she found appeared in California’s Auburn Journal on, appropriately enough, Wednesday, October 31, 1979—Halloween.

GUILTY OF SHOPLIFTING

Joseph DeAngelo, former Auburn policeman, was found guilty of misdemeanor shoplifting charges by a Sacramento County jury Friday.

He was given a $100 fine and six months probation by Sacramento Municipal Court Judge Thomas Daugherty.

DeAngelo was fired from the police force shortly after his arrest last July 2, at the Pay N’ Save store off Greenback Lane.

His Chico attorney, Maureen Whelan, Tuesday was unavailable for comment on the status of the appeal, but it will be heard Nov. 9 before the Auburn Personnel Board.

During the three-day trial, a store employee testified he saw DeAngelo take a can of dog repellant from the waistband of his trousers. Another employee told jurors that he pulled a hammer from DeAngelo’s trousers.

DeAngelo took the stand Thursday afternoon and denied he was trying to steal the items.

Joseph DeAngelo was one of the nine men I had identified as potential suspects. DeAngelo’s brother was also on the list.

As soon as I read the article, I emailed Paul and Steve and asked them if Joseph DeAngelo had been ruled out as a suspect for any reason. When I was younger and living in the San Francisco Bay Area, I used to ski at Palisades Tahoe, and when I did I always took Interstate 80 north from Sacramento and stopped in Auburn for coffee. I knew exactly where Auburn was, and I knew it was a straight shot down I-80 from Auburn to Sacramento, maybe a thirty- or forty-minute drive from Auburn to the scene of so many of the Golden State Killer’s crimes.

Paul and Steve told me that Joseph DeAngelo had not been ruled out.

Now I was excited. Here was a former police officer, well versed in crime scenes and evidence, whose surname indicated he likely had Italian ancestry, which fit the man we were looking for. He had also been convicted of a crime, which made him an especially tantalizing bad rabbit. And he had lived in a city where three of the rapes linked to the East Area Rapist had occurred. To me, there was no denying the arrows were pointing to Joseph DeAngelo.

Paul Holes, however, remained convinced that his Citrus Heights property owner was the best suspect. Paul managed to locate a woman who was closely related to his suspect (I am withholding the exact relationship to protect her identity). Paul’s idea was to contact this close relative and ask her to take a DNA test. When I learned this, I quickly sent an email to Paul and Steve suggesting that if we were going to DNA-test anyone, we should surreptitiously test one of the DeAngelo brothers.

“John and Joseph [DeAngelo] are second cousins to [our MyHeritage match, whose name I have not included] and first cousins twice removed from our FamilyTreeDNA match,” I wrote in my March 22, 2018, email. “They live in California and have an Italian grandparent (Samuel DeAngelo 1882–1942). Joseph was the former police officer arrested for shoplifting. Any chance [the East Area Rapist] would have used dog repellant and a hammer [the items DeAngelo stole]? Odd things for a police officer to be shoplifting.”

Paul responded that he still planned to focus on his preferred suspect. Steve shared with me that Paul believed it was unlikely that someone who was a police officer could have committed the heinous crimes attributed to the GSK. So Paul went ahead with his plan. He contacted the close relative to his suspect, and she agreed to take a DNA test. If the test showed that she matched the crime-scene DNA at the level of her close relationship to Paul’s preferred suspect (and if his suspect did not have a male sibling), there would only be one conclusion: Paul’s preferred suspect was the Golden State Killer.

FamilyTreeDNA expedited processing of the woman’s sample. Still, it took several days before we had match results. In that time I continued researching all nine suspects on my list. I did not believe Paul’s suspect was the killer, but I was still eager to see the results of the relative’s DNA test. All the members of Team Justice had access to the FamilyTreeDNA account where the results would appear, and I am sure I was not the only one of us who repeatedly checked and refreshed the site while waiting, hour after hour, for the match results to be posted. On one day, it felt like I was accessing the site every few seconds.

Finally, late on a Friday night, around 10:00 p.m., the relative’s matches were posted.

I drew a deep breath before I checked them, and then I dove in.

The woman’s DNA did not match the Golden State Killer’s crime-scene DNA at the close level of her relationship to Paul’s suspect.

Paul’s hunch had been wrong.

But—

—the woman did match the crime-scene DNA at the level of a second cousin.

This was significant. It meant that, without a doubt, we were working on the correct family line—and that our list of nine suspects, whittled down to eight with the elimination of Paul’s preferred suspect, was solid.

The Golden State Killer was one of these remaining eight men.

But there was more.

The relative also had an X chromosome match with the GSK.

The X chromosome match indicated that the Golden State Killer was related to the relative through one of his maternal lines (men have a single X chromosome that they inherit from their mother). With that information, I would be able to narrow the suspect list even further by eliminating those men who could not have inherited the X chromosome.

I had barely taken in the new information when my telephone rang. But I knew who was calling.

Steven Kramer had access to the same FamilyTreeDNA account that I did, and I guessed that, though it was late, he had seen the same results and understood we had a close match—which meant we were potentially on the verge of identifying the GSK. I knew that Kramer would probably not want to wait even one more day to begin the process of shrinking down our suspect list. And I anticipated Kramer was going to ask me to keep working into the night, rather than catch a few hours’ sleep and start fresh in the morning. Not that I would have, anyway.

“Barbara,” Steve said cheerily when I answered the call, “how about I send you an espresso from Starbucks right now?”