I wrote (27Feb2007):

**********
(1) C. C. Spencer, a self-proclaimed eye-witness to John Henry's death, wrote (ca 1927) to Guy Johnson that he had died on September 20, 1882. His year has to be wrong - the only year the railroad through Dunnavant, AL, was under construction in September was 1887, the year of John Henry's death given by Glendora Cannon Cummings, who claimed (ca 1927) that her had uncle witnessed John Henry's death. Harvey Hicks (Evington, VA, ca 1930) gave Louis Chappell a version of "John Henry" containing the line, "John Henry died on a Tuesday." September 20, 1887, *was* a Tuesday.

For a particular date, the probability matching the day of the week with a random guess is 1/7, 14%. Thus, the probability that the match that is found is not accidental is 86%.

If it is not accidental, what scenario, other than that it is truth, could account for the agreement. The only such scenario I can think of is that an untrue story including the date and day of the week made the rounds. It is hard for me to imagine how an untrue story could have originated. It is even harder for me to imagine why an untrue story would include such a detail.

My conclusion: From this, it is 86% probable that the historic John Henry died on Tuesday, September 20, 1887.
**********

Sorry. I've retracted this bit of statistical thinking in several places and I thought I'd done it here already but I just looked and discovered that I hadn't.

It is correct that

IF a choice of day of the week is random
THEN the probability of agreement with a date is 1/7 (14%).

It is *not* correct that

IF a day of the week and date agree
THEN the probability that neither is a random choice is 6/7 (86%).

The second "IF...THEN..." does not follow the the first. I erred in thinking that it did. I take it back. I'm sorry.

There *is* a probability treatment of this situation that can lead to the conclusion that agreement between day and date implies that day and date are likely true, with a probability over 80% or so, but it depends on some assumptions and it is more complex that the simple reasoning I tried to use. I thank Nathan Rose for pointing out this approach in another forum.

I still think that the probability is zero that three independent informants could come up with "Dabney"/"Dabner" as the name of John Henry's boss and there not be a true historic background. Add to this the independent fact that Captain Fred Y. Dabney *was* in charge of the construction of the C & W, the line to which two of those informants pointed, and we have a very strong case, I believe.