Five Element Sequences

by Zoltan Dienes and Mike Flanagan

Procedure

Before the experiment proper, tori was trained on the seven points until he could locate them quickly and accurately on different people on left and right sides without pressing on other points. Also, a procedure was determined for randomly allocating subjects to the four counterbalancing conditions (these conditions were: destructive sequence first-left hand side first, destructive second-left first, destructive second-left first, destructive second-left second). That is, a random number generator was used to generate a list of seven random permutations of the numbers 1 to 4. As each subject was picked from the class, he was assigned the next number in the list, assuring that all the counterbalancing conditions were used equally often and, crucially, still otherwise in random order.

As the class proceeded, tori took each subject aside privately, stuck a label on them determining the subject's condition, took the pain rating, and then indicated for them to rejoin the class. On classes in which the experiment was ongoing, no kyusho moves were practiced.
Some points on the logic of the experimental design

Just some points on experimental design for those without a scientific training. You may say: but different people are differently sensitive to pressure points. This is true, but as each uke is compared to himself on the destructive and creative sequences, a non-responder will give zero pain both times, everyone else should give some pain each time and we can see if there are any consistent differences between the conditions. You might say that people's sensitivity may vary with time (depending e.g. on their adrenalin release at that point in the session) and also that tori's effectiveness may vary systematically or randomly throughout the session, creating differences between the destructive and creative conditions. These factors whilst present do not undermine the logic of the experiment because of the random assignment of subjects to conditions. This means that there won't be any *systematic* difference between the destructive and creative sequences. Accounting for random differences between the conditions is exactly what the signficance testing statistics are for; indeed, without such statistics we wouldn't be able to conclude anything, any mean differences could just be random.

Results

This section will be difficult to follow completely for those without previous exposure to inferential statistics, even with the brief explanation given for each term. However, a comprehensible summary of the results is given in the Discussion. For the table below, we have tried to be thorough so that people can get a feel for various aspects of the data. In order that the wood can be seen for the trees, though, the important points will be highlighted after the table.
In the table below the mean pain ratings are listed according to the target element, and according to the following contrasts: whether the creative or destructive sequence was followed, whether it was the left or right side of the body, and whether it was the first time that uke was tested on that point or the second time (20+ minutes later). Standard deviations are given in parentheses (standard deviation is a measure of variability between different subjects: roughly 2/3 of people lie within one standard deviation of the mean). For each contrast, the mean difference is reported and the 95% confidence limit on the difference. To appreciate this concept, bear in the mind the distinction between a sample ( the set of people that we actually tested at a particular time) and the population (the set of all people and times we could have sampled from): We are really interested in the population, not in the random vagaries of a particular sample. The 95% confidence interval says (loosely) that we can be 95% sure that the true population value of the mean difference lies between the quoted lower limit (first number in parentheses) and the quoted upper limit (second number in parentheses). (This is not the technical definition but it captures how confidence intervals are best thought about.) If zero is included in the interval then we have no evidence that there is any population difference (i.e. a t-test would not be significant at the .05 level). The limits tell us that whatever population difference may exist, we are 95% sure that they don't lie outside the stated limits; e.g. to take the difference between the creative and destructive sequences on all the data, we have no evidence that the sequences caused different set-ups of the target points (zero contained within the interval), and whatever difference as there may be, we are sure it is not more than 0.5 of a pain rating (i.e. whatever effect as there may be is very small). That gives a measure of how sensitive our experiment is.

Because the data for some elements were not entirely normally distributed, Wilcoxon p's are also reported. This test does not require normally distributed data. If the quoted number is less than .05, there is evidence of a difference. It can be seen that the results produced by the Wilcoxon's are entirely consistent with those produced by the confidence intervals based on t-tests.

	Water	Wood	Earth	Fire	Metal	All
Creative	6.1 (3.7)	5.5 (4.0)	7.0 (2.1)	6.5 (2.3)	5.2 (1.9)	6.1 (1.6)
Destructive	5.6 (3.2)	5.4 (2.8)	7.0 (2.2)	6.3 (2.0)	5.5 (1.8)	6.0 (1.3)
Difference	0.5 (-0.4, 1.3)	0.1 (-0.9, 1.1)	0.1 (-0.6, 0.7)	0.2 (-0.6, 1.0)	-0.4 (-1.1, 0.4)	0.1 (-0.3, 0.5)
Wilcoxon p	0.31	0.82	0.64	0.69	0.39	0.50
Left	6.0 (3.1)	5.5 (3.0)	6.7 (1.9)	6.4 (2.0)	5.6 (2.0)	6.0 (1.3)
Right	5.8 (3.9)	5.5 (3.8)	7.3 (2.4)	6.4 (2.2)	5.1 (1.7)	6.0 (1.7)
Difference	0.3 (-0.6, 1.1)	0.0 (-1.0, 1.0)	-0.6 (-1.2, 0.0)	0.0 (-0.8, 0.8)	0.5 (-0.3, 1.3)	0.0 (-0.5, 0.5)
Wilcoxon p	0.67	0.53	0.10	0.97	0.14	0.93
First	5.3 (3.0)	5.3 (3.9)	6.8 (2.0)	6.0 (1.9)	5.1 (1.7)	5.7 (1.4)
Second	6.5 (3.8)	5.6 (2.8)	7.2 (2.4)	6.7 (2.3)	5.6 (2.0)	6.3 (1.5)
Difference	-1.3 (-2.0, -0.5)	-0.3 (-1.3, 0.7)	-0.4 (-1.0, 0.2)	-0.7 (-1.5, 0.0)	-0.6 (-1.3, 0.2)	-0.6 (-1.1, -0.2)
Wilcoxon p	0.0027	0.13	0.25	0.041	0.16	0.005
Average	5.9 (3.3)	5.5 (3.2)	7.0 (2.0)	6.4 (1.9)	5.4 (1.6)	6.0 (2.5)

Considering all the data, the lowest rating given by anyone was 1 (and the highest was 20). That is, there were no nonresponders in the sense of people who felt no pain. The average amount of pain was 6 (on a scale where 10 means "so much pain one wants to sit down"). If pressing activates points at all, then points should have been activated in this experiment.

Now lets consider various questions addressed by the results:

Does the destructive cycle activate points more than the creative cycle?

For none of the elements was there a significant difference between the creative and destructive sequences. Pooling all data together, there was still no difference. In all the data, we can be 95% sure that whatever the true population advantage of the destructive sequences over the creative sequences, it is not more than 0.3 on the pain rating scale. This is a tiny amount, a change in pain of 0.3/6 or 5 percent. Put another way, our data rules out the destructive rather than creative cycle resulting in a more than 5% increase in pain.

Are the left and right sides of the body equally sensitive?

For none of the elements was there a significant difference between the left and right hand sides of the body, and there was still no significant difference when all the data were pooled together.

Are people more sensitive on the second time they are tested?

Our data bear on another claim sometimes made by people inspired by TCM. It was probably George Dillman who introduced the notion that one should not train on both sides of the body in the same session, because stimulating one side of the body activates the other side. In our experiment, the same target point was pressed on opposite sides of the body in the same session. Consistent with the TCM claims, in the data as a whole, people gave higher pain ratings on the second test (6.3) than the first (5.7). The difference was significant in only some of the elements taken individually, but an analysis (Friedman and analysis of variance) indicated that the difference between first and second testing did not vary significantly across the different elements (p > 0.3) (finding an effect significant in one condition and not in another does not indicate that the conditions differ; that has to be specifically tested). That is, as far as we can tell, the difference could be treated as a general one across the elements.

One straight forward explanation of this second time of testing effect could be that tori simply got better at pressing the points over the course of the session. However, this explanation isn't plausible when we consider how pain ratings varied across successive subjects within the first lot of testing or within the second lot of testing: A regression of these pain ratings on subject order (1 to 28) showed that there was not a positive slope, if anything there was a negative slope. That is, there is no evidence that tori was getting better with practice during the session, or even just pressing harder with time.

A note on pooling across all data for the statistically minded

Within each element 28 different people were tested. Across elements, some of the people were the same and some were different depending on who turned up for each session. That is, the data could not simply be treated as a set of 140 independent observations. Thus, when the data were pooled, subjects were matched according to the order in which they were tested for each element. This ensured that the number of observations treated as independent by the analyses did not exceed the number of different subjects tested. It may be noted that when all the observations were treated as independent, the analyses produced virtually identical significance values and confidence limits as the matched subjects analyses.

Page 1 - Page 2 - Page 3 - Page 4

Home ¦ Articles ¦ Calendar ¦ Forum ¦ Journal ¦ Chat ¦ Photos ¦ Clubs ¦ Arcade