I'll Be Gone in the Dark: One Woman's Obsessive Search for the Golden State Killer

Dad

Many have attempted to profile EAR-ONS over the years, but Michelle wanted to go one step further and dive deep into the locations of the rapes to see whether geographic profiling could lead to his identity. Among the pieces she left behind were her musings about EAR-ONS’s geography:

My feeling is that the two most important locations are Rancho Cordova and Irvine.

The first and third rapes were only yards apart in Rancho Cordova. He walked away in an unhurried fashion from the third attack without his pants on, suggesting he lived close by.

He murdered Manuela Witthuhn on February 6, 1981, in Irvine. Five years later he murdered Janelle Cruz. Manuela and Janelle lived in the same subdivision, just two miles apart.

Interestingly, Manuela’s answering machine tape was stolen in the attack. Was the suspect’s voice on the tape? If so, was he worried it was recognizable as someone in the neighborhood?



A document Michelle created in August 2014, entitled “Geo-Chapter,” has her rethinking the map after more than three solid years of nonstop research. When you open it, there is just one line: “Carmichael seems like central clearing, like a buffer zone.”



FINDING THE KILLER WITH GEO-PROFILING

While his most fundamental characteristics—his name and his face—are unknown, it can be said with reasonable certainty that the East Area Rapist was, among approximately seven hundred thousand other humans, a resident of Sacramento County in the mid-to-late 1970s.

The EAR’s connection to the many other places in which he struck—Stockton, Modesto, Davis, the East Bay—is less clear.

The East Area Rapist was a highly prolific offender in Sacramento, exhibiting the familiarity and ubiquity of someone who was undoubtedly local. With places like Stockton, Modesto, and Davis, where he struck two or three times apiece, one questions what connection he had to these cities, if any. Perhaps he had family there or had business there. Maybe he was just passing through. Maybe he flung a dart at a map.

But you’d be hard-pressed to find an investigator who doesn’t believe the EAR lived or at least worked in Sacramento.

If we accept that the EAR was living in Sacramento from 1976 through 1978 or 1979, which is nearly certain, and then lived in Southern California during the first half of the 1980s, which is highly probable, then the haystack gets considerably smaller. Devise a list of people who lived in both areas during those time periods, and the suspect pool shrinks from nearly a million to maybe ten thousand.

It would be ideal if the process were as simple as, say, applying filters to a product search on Amazon. With a few clicks, one could filter by gender (male), birth year (1940–1960), race (white), height (5′ 7″ to 5′ 11″), places lived (Carmichael AND Irvine; or Rancho Cordova AND in the 92620 zip code; or Citrus Heights, Goleta, AND Dana Point), and maybe occupation for good measure (real estate agent, construction worker, painter, landscaper, landscape architect, nurse, pharmacist, hospital orderly, cop, security guard, OR serviceman—all of which are among the many occupations that various investigators and armchair sleuths have posited the EAR may have had). Just set all these search parameters and voilà! You’d be left with a manageable yet all-inclusive list of potential suspects.

But it’s not that easy. The names have to come from somewhere, and there’s no central database of, well, people. It must be either composited or built. And creating such a list is indeed one of the projects Michelle felt most optimistic about.

He may have come from Visalia. Or perhaps Goleta was his hometown. He may have lived in the 92620 zip code of Irvine. He may have gone to Cordova High School. His name may appear in both the 1977 Sacramento phone book and the 1983 Orange County phone book. We didn’t need access to restricted information or an official suspect list to uncover some potential suspects who might otherwise have flown under the radar. All the necessary information and tools that could be used to process it were already available in the form of online public records aggregators, vital records, property records, yearbooks, and yellowed phone directories from the 1970s and ’80s (many of which have fortunately been digitized).

In the year prior to Michelle’s death, Paul had begun creating master resident lists for Sacramento and Orange Counties for the relevant time periods, which combined names from sources such as Ancestry.com’s marriage and divorce records, the appropriate county’s registry of deeds (which entailed using a Web scraper), alumni lists, and old crisscross directories and telephone books.?

Michelle then connected with a computer programmer in Canada who offered to volunteer his help in whatever way he could. Per Paul’s specifications, the programmer built a cross-referencing utility that processes multiple lists and finds matching lines of text. With that application, Paul could begin feeding it two or more lists and then analyze its match results—now numbering over forty thousand.

Once the list of matches had been generated, Paul would go through it and weed out the false positives (far more likely with common names like John Smith) by using public records aggregators. Paul would then collect as much information as possible on each match until he was satisfied that neither he nor any of his male relatives was viable. The names of those he was unable to rule out would be added to a master list of potential suspects.

IN CASES OF SERIAL BURGLARY, RAPE, OR MURDER, THE SUSPECT lists often swell to several thousand names and beyond. Difficulty in managing a list of this size enforces the need to devise a prioritization system, whereby suspect ranks are determined by factors such as prior criminal offenses and police contacts, availability for all crimes in the series, physical characteristics, and—if a geographic profile has been done—the suspect’s work and home addresses.

Geographic profiling is a specialized criminal investigative technique—perhaps more useful and scientific than behavioral profiling, which is arguably closer to an art than a science— whereby the key locations in a linked crime series are analyzed for the purpose of determining the likely anchor points (home, work, etc.) of a serial offender. This allows one to focus on isolated bubbles within a much broader suspect pool.

Although the general technique has informally been around for a while—you see investigators employing it to find a kidnapper in Akira Kurosawa’s High and Low (1963)—the methodology of geographic profiling didn’t even have a name until the late 1980s, about ten years after the phrase “serial killer” first entered the popular lexicon. Given that it wasn’t yet an established investigative procedure, awareness of geo-profiling could not have been a factor motivating the EAR—a lover of misdirection—to misdirect geographically by commuting great distances to faraway neighborhoods in Southern California. Moreover, his Southern California crimes were not generally recognized as EAR crimes (and he specifically seemed to want to avoid this recognition, which is likely one reason he began killing his victims—to eliminate witnesses) until DNA evidence established them as such. The logical conclusion, per the principle of Occam’s razor, is that the EAR was living in Southern California during the period in which he was offending there.

That said, while we would not advocate completely eliminating someone merely because a Southern California residence cannot be established, it would take some damn compelling reason to muster any interest in such a suspect.

Michelle McNamara's books