The ambiguity of the questions was important. When Cromer had given a similar survey to a police department that had experienced thefts from the evidence room, seventy-four of the seventy-six respondents offered similar rationales for what could have created the situation: the evidence room was sloppily run and often left unlocked. But two of the respondents offered explanations that had nothing to do with management: “Maybe someone needed money,” they said. “Maybe someone had a drug problem.” Those answers in themselves weren’t any kind of admission to guilt, but they were different enough to get Cromer’s attention.
The order of the questions was important, too. Each question was designed to prompt a different mind-set that would set up the next question. The order had been especially useful in child abuse cases. Often a parent, when asked to speculate what could have caused burning or bruising on a child, would blame external factors—a playground fall, a rough hour of recess. But when the parents were asked how they would investigate the situation themselves, the responses wouldn’t follow at all from the previous question. “I think you should talk to the family and see if anybody has a bad temper,” they might write—an investigating tactic that made no sense if the culprit had been a swing set. If the parent truly believed the child was being injured on the school playground, their proposed solution should have involved advocating for better recess supervision, or removing dangerous equipment. Their minds would not have immediately leapt to anger management issues at home.
It was almost, Cromer analogized, like a grade schooler trying to use the test in order to beat the test. They might be able to come up with plausible answers to one question, but when asked to come at the answer from a different direction, they would falter.
Cromer had all of this in mind when he passed out the surveys, first to the fire company in Bloxom and then later to the one in Parksley. Most of the responses were what he expected them to be, blaming punk kids or vindictive insurance seekers. One of the responses came back a little different. On the last page, in response to how she would investigate the arsons, the firefighter had written, “I would start with the fire service”.
But instead of a period, she’d ended the statement with a colon. “I would start with the fire service:” It was the colon that was interesting to Cromer. Had she merely run out of time? Or was she about to elaborate on her suggestion with specific names, before thinking better of it? Cromer recommended the woman be brought in for questioning. It didn’t lead anywhere. As it turned out, none of the structured questionnaires did. But this was the nature of the investigation at this time. The investigators were willing to try anything, even open themselves up to the possibility that the entire case could hinge on an inexplicable piece of punctuation.
Cromer’s other contribution to the investigation was immediately suggesting that his superiors call Isaac Van Patten.
Van Patten, like Cromer and Morris, was a profiler. Unlike those two, both career police who had become law enforcement first and then profilers, Van Patten was an academic. He had a PhD in marriage and family therapy, and his official employer was Radford University in central western Virginia, where he was a professor in the behavioral science department. But he consulted with the Virginia State Police, too, and after years of focusing on therapeutic approaches to treating offenders and on the psychological aspects of criminal profiling, Van Patten had developed an interest in a burgeoning field of study. This field of study was the other thing that set him apart from the profilers already on the case, and the reason Cromer so badly wanted him to join the effort.
Van Patten’s specialty was called geographic profiling. It was a discipline based not on psychology but on data. Geographic profilers mapped the locations of crimes, and in doing so, mapped the criminal’s mind.
The laymen’s explanation that Van Patten found most helpful, and that he frequently referred people to, came from the pilot episode of the crime-solving drama Numb3rs. In it, a detective goes to his mathematician brother, searching for a way to predict a serial rapist’s next target. The task, the brother tells him, would be nearly impossible. Think of a lawn sprinkler: even with all the mathematical models in the world, there would be too many variables—change in wind direction, mechanical glitches—to predict with absolute certainty where each droplet would fall. But it was possible to work backward and go in the other direction. With enough fallen droplets, one could develop an algorithm to trace each one back to their common point of origin. Given the droplets, you could find the sprinkler. Find the sprinkler, find the criminal. The idea was that repeat offenders, even the ones who believe they are choosing their crime scenes completely randomly, are actually subconsciously employing patterns. They orient their actions around a home base, like a residence or place of work. Van Patten, with luck, could use a computer algorithm to point toward the home base.
Because geographic profiling was based solely on location data, the type of crime wasn’t important. It worked the same for a serial rapist as for a serial burglar or serial murderer. The only ingredients necessary were addresses, the more of them the better. In late November, the Virginia State Police had arranged for Van Patten to be sent the addresses of all of the fires so far. There were about twelve at that point, more than enough for an initial profile, with the idea that additional sites could be added in the future. Analyzing these incidents, Van Patten developed a report, which he submitted to the police in December 2012.
The beginning of the report summarized the time patterns of the fires: Most of them had been clustered between 10 p.m. and midnight—which indicated, if nothing else, that the arsonist must not work a night shift. The report also looked at the topographical spread of the fires, and where each one lay within the county. So far, though the arsonist had flirted with both county borders, he hadn’t left Accomack.
These were all observations that could have been made by any armchair detective with an Excel spreadsheet. For the second part of the analysis, an algorithm was needed. Van Patten was careful as he input the addresses, input being the most likely occasion for human error to poison the data. What Van Patten’s computer screen first showed him looked simply like a messy series of overlapping ovals and circles on a map, all roughly clustered toward the center of Accomack County. This was just the raw data. Van Patten needed to interpret and streamline it so it would make sense to human eyes. When he finished doing that, the results came back a second time. This time, the computer screen showed him one neat circle, a quarter mile in diameter, beginning just north of Parksley and ending just south of Bloxom. Somewhere within this circle, Van Patten hypothesized, was the home or workplace of Accomack’s arsonist.
Van Patten included an image of that map on the memo he submitted to the Virginia State Police, with his recommendation for what should be done with the data: