But did Mrs. Clements actually testify against her husband? We don’t know. The Atlanta Constitution reported on February 26, 1909, at the start of the trial, that “Clements’ aged father, his wife and seven small children sat around him. In his lap he held his little son.” This suggests that Mrs. Clements was supporting her husband at the time of the trial, and also, it is unlikely that a wife could testify against her husband in 1909. But it does not appear that any reporter actually attended more than a few minutes of the trial, and we do not know what evidence was produced that led to Clements’s conviction.

  The authors accept in a general way that Clements was probably guilty of these murders, and also, if Clements wasn’t, then it could have been Edmondson himself. However, we will point out the following:

  1. Clements never confessed to the crime.

  2. The fact that Clements (in the story told by his wife) left the house and returned the next day to set fire to the house and barn strongly suggests that the Edmondson house was sufficiently isolated that the crime was not discovered for some time, which means that we don’t really know (other than by Mrs. Clements’s story) when the crimes occurred. For all we know, the crimes could have occurred in the middle of the night, as most of “our” crimes did, or they could have begun, as several of the other crimes did, with a late-evening attack on the man of the family in the barn, followed by an attack on the remainder of the family in the house.

  3. No account of the murders says what the murder weapon(s) was or were, on which basis we cannot and do not assume that it was an axe, but also cannot assume that it was not.

  4. It does appear that there were young females among the victims.

  5. As noted above, Woodland Mills has every possible characteristic consistent with the established patterns, other than the fact that the railroad didn’t go there, and

  6. What would you suppose is the chief industry in a settlement called Woodland Mills?

  Thus, based on what we know, Clements could possibly have been an innocent man convicted of a crime committed by the Crazy Tramp, as several other men almost certainly were.

  The Atlanta Constitution on March 11, 1909, “reported” that “on a lonely island in the Tennessee River, several miles below Decatur, there lives a strange and mysterious white man with a negro family, according to stories told by Tennessee River steamboat men.

  By some this strange man is thought to be Tom Edmondson, who was supposed to have been murdered and his body cremated with other members of his family at Woodland Mills, this county, on November 25, 1908. Those who have seen this man and who have read descriptions of Tom Edmondson say the descriptions of the two men are one and the same.

  This story, which goes on to throw the conviction of Robert Clements into doubt, was almost certainly planted in the newspaper by Clements’s attorney in an effort to help his appeal. Note the spectacularly vague sourcing of the information about the man on the island, and the rank implausibility of establishing identity by comparing descriptions—not to mention the implausibility of the story itself.

  * * *

  Looking at the three suspicious crimes of 1908, there is no convincing evidence that The Man from the Train was involved in any of the three. The murder of the Hart family is a low-information event. The murder of the Gerrells may have been committed by two men, who may have been using clubs, rather than axes, and the murder of the Edmondsons is more probably than not unrelated.

  The Man from the Train was probably in prison in 1907, and we cannot say with real confidence that he was not still in prison, or still inactive for some other reason, in 1908. However, the fact that there are three crimes in 1908 that are more or less consistent with the pattern suggests the possibility that he was active. If he was in prison in 1908, there could well be one “incidental similarity” case that looks like it might be his, but it is not likely that there would be three.

  At this point there are 121 murders that have been discussed in this book, including eleven in this section:

  SECTION IV

  Where

  Who

  No.

  Date

  Year

  Frazier, Georgia

  Hart family

  2

  March 4

  1908

  Watauga, Texas

  Gerrell family

  3

  April 12

  1908

  Woodland Mills, Alabama

  Edmondson family

  6

  November 25

  1908

  For the sake of clarity, the authors do not believe that all of these 121 murders were committed by the same man. We believe that a substantial number of the murders were committed by the same man, and that any of the murders on this list might possibly have been his work.

  SECTION V

  CHAPTER XXXIV

  Conversation with the Reader

  We have circled back now to the point at which we began. The next murders in the series are the murders in Hurley, Virginia, which were the first story in the book. I have more to say about that crime and those that followed it, but that begins in the next chapter.

  Let’s assume that some of you are resisting the notion that this was in fact a series of related crimes, rather than a series of unrelated crimes. I would suppose that many of you were skeptical about this thesis when you began the book, and I would hope that most of you are not as skeptical now that you have heard most of the facts. Still, the time has come to address the nagging concerns of the unconverted. Let me make the argument that I think a skeptical reader might make, were she here to speak for herself.

  What real evidence is there, that reader might argue, that these murders were all committed by the same man?

  In a country the size of the United States there must be a good many families murdered each year. Many, many people lived close to the railroad track. Many of those who are murdered were no doubt murdered with axes. I am not questioning your sincerity, Bill and Rachel (the reader says politely), but how do I know, from my position, that you have not merely chosen certain crimes that accidentally happen to have these few critical characteristics in common, and then built upon that by pointing out whatever other coincidental similarities might appear?

  In response to that, let me point out, first, that it was widely recognized at the time, beginning in October 1911, that this axe murderer was at work; that is, it was recognized that a limited number of these crimes were connected. The newspaper attention that came immediately to Villisca was drawn there in large part because the moment that the crime was discovered, it was suspected that this was the latest crime committed by the unknown man who had murdered several other families.

  And second, let me point out that there are no fantastic or highly improbable elements of the story that we have told you or will tell you, stretching from 1898 to 1912. What element of this story would you find fantastic? Do you not believe that a man could be so evil as to do this? Do you not believe that that man could be clever enough to develop this simple scenario, by which he could escape detection (towns too small to have a police force, access to multiple railroad lines, strike late at night, be gone before dawn)? What, really, is improbable about it?

  What seems improbable is that the series of crimes could have gone on for so long without anyone realizing what was happening. But turn that around and look at it from the other end. If you look at the first “recognized” crimes of the 1911 to 1912 era, which were the Colorado Springs murders, is it easy to believe that this is where the series of crimes begins?

  To a modern reader it should be obvious that this is not where the series begins. A sophisticated serial murderer gets to that stage one step at a time—the same way that an athlete becomes an athlete, the same way that a writer becomes a writer, the same way that a musician becomes a musician. A serial murderer normally commits lesser crimes such as arson or molestation, graduates to murder in an isolated event, and then, more often than not, hunkers down for a
time to blend into the walls. Eventually he resumes acting out his fantasies and, with the passage of the time, becomes bolder and more active.

  The Man from the Train murdered two families in Colorado on one night in 1911, murdered another family two weeks later, and murdered another family two weeks after that. That is not a young, inexperienced offender. That is a serial murderer who has reached a mature phase.

  Newspaper writers and ordinary citizens, at that time, did attempt to make connections between the Colorado Springs murders and earlier crimes in the Portland/Washington area. They lacked sufficient information to go very far with it. There simply was no way, in 1911, to connect the dots among events stretching over a period of years and occurring in different places. There was no place where information could be gathered, sifted, and systematically reviewed to look for patterns. There was no technology to compile the information or to sort through it.

  Contemporary newspapers and others realized that a Subsection of the crimes were related, the Subsection beginning in Colorado Springs. But if you go back just weeks before Colorado Springs, to Ardenwald, Oregon, there is another little girl laid out in the middle of the room, covered with bloody fingerprints, her family murdered with an axe, with the blunt side of an axe, the axe taken from a neighbor’s yard and left in the room, her family murdered in the middle of the night. And if you go back a couple months before that, to San Antonio and the Casaways, there is another one. Why would we not believe that these crimes are part of the same series?

  Actually, the main reason that the crimes have not been connected by other writers is irrational skepticism. I’ll get back to that point, irrational skepticism, but let me proceed toward it in an orderly way. I had points one and two in response to the rational skeptic above. The third point is that, of course, we don’t know in many cases which crimes he committed. There are perhaps fourteen crimes about which we have enough information to be certain they were committed by the same man, and then there are also many other similar crimes, and we believe that The Man from the Train was responsible for some of them or many of them, but not all of them. I have about six more arguments that I need to make here, and let me try to organize them.

  * * *

  (1) Visualize a target, with bullet holes spread around the surface of the target.

  If the bullet holes are randomly distributed across the target, then there is no reason to believe that anyone was aiming at the target; it is more likely merely that the target was there, and there were a bunch of bullets flying around, and some of the bullets hit the target.

  If, on the other hand, you have a cluster of bullet holes around the center of the target, and many fewer bullet holes on the outer edges of the target, then you have to conclude that someone was actually aiming at the target.

  We have a large group of tight clusters here. If we had as many murders ten miles from a railroad track as we do a five-minute walk from a railroad track, that would be a random pattern. But when we have many, many murders within a five-minute walk of a railroad track and two or three crimes that are a few miles from a track, that’s a cluster.

  If we had as many murders committed at 4:00 a.m. as at midnight, some at 6:00 a.m., some in the afternoon, some at dusk, we could conclude that this was a random pattern. But all of the murders which are central to this story—all of the murders that you have to believe are related or you’re off the reservation—all of them were committed within ninety minutes of midnight. That’s a cluster.

  If we had murders committed in big cities and small towns and all sizes of towns in between, we could view those as random events. But, in fact, we have no crimes at all (relevant to this series) that were committed in a large city, one crime committed in what was then a small city (San Antonio), and the vast majority of our crimes committed in towns and settlements that were too small to have a police force. That’s a cluster. That’s a pattern.

  If we had murders committed with an axe, and some with a knife, and some with a gun, and some with a baseball bat, then of course we could view those as random events. If we have a series of murders committed with an axe or similar instrument, that’s a cluster.

  Except that, in approaching this problem in this way, I have tremendously understated the implausibility of these being random, unrelated events. They were not merely near railroad tracks, they were extremely near to railroad tracks, and they were predominantly near the intersections of multiple railroad tracks, and they were almost all at places where railroad trains would have to stop, and the bloodhounds in several cases tracked the scent directly to the railroad line, at which point it was lost.

  And these crimes were not merely committed with an axe, they were committed with the blunt side of the axe, and they were not merely committed with an axe, they were committed with an axe that was taken from the family’s yard or from a neighbor’s yard or, in two cases, the coal shed. And the axe was (generally) left in the room. And after fingerprints were developed, the handle of the axe was washed off in a bucket of water. You think it just happened that way?

  And I haven’t even mentioned the idiosyncratic behavior of the criminal. I haven’t even mentioned the moving of lamps, or the removal of the chimneys from the lamps, or the placing of cloth over the windows, or the covers over the heads of the victims, or the practice of locking up the house tight as he left, or the special attention to the young female victims.

  There are thirty-three elements that help identify a crime which may be connected to this series, not thirty-three distinctly different elements, but thirty-three elements with some overlap and some redundancy. I will list the thirty-three elements at the end of this chapter. But it is preposterous to suggest that there is not a cluster of events that surrounds a small target—and it is clear that this cluster of events describes events other than just The Subsection which has previously been connected by other writers. There is a target there, and there is somebody aiming at the target.

  * * *

  (2) There are nowhere near enough murders of families in the relevant years for these to be random clusters of events.

  The murder of an entire family is a relatively rare event. It was relatively rare in 1911, and it is relatively rare now. How rare?

  We tried to document every family murder in the United States in the years 1890 to 1920. I can’t tell you absolutely that we have found every single event, but we made a serious and determined effort to find every single event, and we certainly found most of them. This is the count, by years:

  1890

  7

  1891

  6

  1892

  8

  1893

  9

  1894

  6

  1895

  7

  1896

  11

  1897

  10

  1898

  2

  1899

  7

  1900

  4

  1901

  8

  1902

  14

  1903

  10

  1904

  10

  1905

  9

  1906

  9

  1907

  4

  1908

  11

  1909

  9

  1910

  10

  1911

  19

  1912

  14

  1913

  9

  1914

  6

  1915

  5

  1916

  6

  1917

  5

  1918

  1

  1919

  7

  1920

  5

  That averages exactly eight families murdered per year, 248 in thirty-one years. In the years when The Man from the Train was most active, there would be about ten families murdered per year in the United States. (The number jumped in 1911 because there wer
e at least two other persons murdering families that year, in addition to The Man from the Train—the New Orleans Axeman, and Clementine Barnabet. Story to come.)

  After The Man from the Train was no longer active, the number of families murdered per year dropped sharply. It could be that the number dropped sharply simply because our man was no longer active; that could be true and probably is true, but we are not making that argument.

  The argument we are making is this: that there are nowhere near enough crimes here to form random clusters of identifying characteristics. The crimes that we have described for you in this book represent a very significant portion of all the family murders that occurred in the United States in those years.

  Another thing that unites these murders is simply that they are so horrible. Almost every crime in this book was described in the newspapers as “the most horrible crime ever committed in this region” or “one of the most terrible crimes ever in this state” or by some similar phrase. We have quoted those phrases periodically throughout the book, and we did that to make a point: that these are not common events. There simply are not many crimes like this.

  Further, family murders in general—what could be called unremarkable family murders—almost always have certain characteristics in common. Unremarkable family murders usually occur in daylight or in the early-evening hours (because they usually occur in the context of a heated family dispute). Unremarkable family murders almost always are committed by a family member or by a person close to the family such as a rejected suitor, a servant, or an overfriendly neighbor, often followed by the culprit’s suicide. It is usually immediately obvious who has committed the crime. Often that person has a history of mental illness, alcoholism, or drug abuse, although in some cases the crime is committed due to greed. Unremarkable family murders are normally committed with a gun, a knife, or a sharp instrument such as a scythe.