Here are the results of my basic analysis. I got too it a little sooner than this weekend it turned out. Any comments are appreciated.
----
Introduction
Are psychic mediums really communicating with the spirits when they state letters/names, or are they thinking probabilistically and simply choosing the letters/names that are common?
The medium in question is John Edward (JE). JE has a television show called Crossing Over, and makes appearances on Larry King Live. On each of these shows he does readings, and claims to get messages from the spirits. Skeptics say that JE is performing cold reading, a collection of techniques used (and in use) by magicians to simulate getting messages from the beyond. Which of these explanations fits the facts? Through a basic analysis, this paper will show that JE used the high frequency letters
less than what was expected.
We can only speculate without actually collecting and analyzing data. Many skeptics and believers alike have dissected JE's transcripts in order to reveal what was going, or not going, on. These approaches, while instructive, do not lend themselves to numerical analyses. A few members (skeptics and believers) on the James Randi Education Foundation (JREF) bulletin board thought it would be instructive to see if they could analyze the letters/names JE gives in a numerical manner, and see what conclusions could be drawn. This paper is one skeptic's attempt.
The transcripts of three Larry King Live shows in which JE appeared were obtained from the JREF thread:
http://www.randi.org/vbulletin/showthread.php?s=&threadid=24032. Specifically, these transcripts were:
http://www.cnn.com/TRANSCRIPTS/0109/10/lkl.00.html,
http://www.cnn.com/TRANSCRIPTS/0209/06/lkl.00.html, and
http://www.cnn.com/TRANSCRIPTS/0302/28/lkl.00.html.
Methods
This analysis focused on the letters/names clearly referring to humans, that occured only where JE was doing readings over the phone. The instances where a letter/name guess occured on a Crossing Over clip were not included in the counts.
This paper reports the results of a chi-square goodness of fit analysis on JE's usage of high-frequency letters. The “high-frequency” letters were defined by this author as the letters where
(see footnote 1) 80*frequency for that letter according to the 1990 combined adjusted male and female Census Bureau data is greater than 5. Using that definition, the following 6 letters were determined to be high-frequency: A, C, D, J, M, R.
The full frequencies can be obtained at:
http://www.randi.org/vbulletin showthread.php?s=&threadid=24602&perpage=40&pagenumber=10.
The counts of A, C, D, J, M, and R were obtained from the three transcripts. Occurances of a string of names were counted as a singular occurance of that letter. Hypothetically, 'like a John, Jake, or Jack', if listed at the same place in the reading by JE, would be counted as 1 guess of J, not as 3 guesses. It seems reasonable that JE focuses more on the sound of the name, as he has stated himself numerous times, although whether this is the case is still up for debate. In this example, JE would be talking about a name that starts with a J. This author agrees that a more complicated analysis would treat name guesses and letter guesses separately, and/or somehow penalize the medium by weighting multiple name guesses less.
The table of the counts, per transcript and the total counts, is shown below:
(9/10/01, 9/6/02, 3/28/03,
Observed total counts)
A: (0, 1, 0,
1)
C: (2, 0, 2,
4)
D: (0, 1, 0,
1)
J: (4, 2, 2,
8)
M: (0, 1, 1,
2)
R: (1, 1, 1,
3)
Expected letter frequencies (from the 1990 combined adjusted male and female Census Bureau data):
A: .0648
C: .0721
D: .0742
J: .1336
M: .1003
R: .080
Expected counts = 80 guesses*1990 combined adjusted male and female Census Bureau data frequencies =
A:
5.18
C:
5.768
D:
5.936
J:
10.68
M:
8.030
R:
6.40
The test statistic for a chi-square goodness of fit test is: SUM[(Observed-Expected)^2/Expected]. This test statistic would be compared to a chi-square distribution on k-1 degrees of freedom, where k is the number of distinct groups. Readers interested in specific details can do a Google search for 'chi-square goodness of fit test' for specific calculation details.
Conclusion
We reject the hypothesis that JE uses letter frequencies according to the 1990 combined adjusted male and female Census Bureau data (p-value from a chi-square goodness of fit test on 5 degrees of freedom = .0102). According to the observed counts, JE used the high frequency letters
less than what was expected.
Discussion
There are many potential problems with the analysis presented in this paper.
First, it is not clear to the author if defining the high-frequency letters in the ad-hoc manner which was done was appropriate. Second, one assumption for a chi-square analysis is that the sample values are independent. The independence assumption was violated several times when JE would guess more than one letter/name while he was reading the same person.
It would be reasonable to perform a follow-up analysis with more than 3 transcripts. The author feels that this study is limited by "pseudo replication". The sample size, "n", is really 3, not 80. Having more guesses per transcript doesn't increase our number of transcripts.
It would also be interesting to use a more current Census Bureau list and see how that impacts the analysis.
It is also important to note that many different people have different methods of counting the letter/name occurances. Clearly the analysis is highly sensitive to the counting methodology. It would be wise for the skeptical and believer communities to develop and agree upon a standardized counting method.
Ideally, the author believes that it would be more productive from a scientific viewpoint to study JE in a lab setting as opposed to studying the transcripts from a television show, but believes one can still learn something informative by analyzing the transcripts from JE's television appearances.
---
(1) 80 guesses was estimated by the author as an eyeball average of JREF posters' counts of the total number of guesses.
.......neo
I can understand you not wanting to make the trip for JE, but to pass up the chance to have a drink with us????? 
.......neo
