Crop Circles -> Home
Eltjo Haselhoff, PhD, young Dutch physicist affiliated in the "Dutch Centre for Crop Circle Studies" wrote a study published in the professional review of biology, "Physiologia Plantarum," January 2001. He worked on the took measurements done by Dr. Levengood and Nancy Talbot's team, on the dilation of the nodes of the stems and he formulated a mathematical model from where it appears that the effects on the plants may have been caused by an electromagnetic source located low above the center of the circle. After the controversies which followed, he answered in a series of article, underneath is the first of these.
Important note: I publish this article, but it does not mean I agree with it. Please also read my article here.
This article may be copied and distributed freely, but only in unmodified form. A link to www.dcccs.org would be appreciated.
"Cerealogists who insist some circles are genuine never use double-blind testing to eliminate bias from their experiments. When they test plants for anomalies, they always know beforehand which plants came from circles," says Joe Nickell, senior research fellow at the Committee for the Scientific Investigation of Claims of the Paranormal.
And he is not the only one. The "lack of-double blind tests" is an often-heard attack against crop circle research. However, although it probably sounds good to the layman, this remark only demonstrates lack of factual knowledge about crop circle research. Or even profound ignorance. Here is why.
First of all, it should be noted that most crop circle tests have been performed in a single or double blind fashion! The claims that crop circle researchers "refuse" to perform blind tests are simply made up.
"A double-blind test is a control group test where neither the evaluator nor the subject knows which items are controls. The purpose of double-blind testing is to reduce error, self-deception and bias."
Here is an example. Suppose, a pharmaceutical company develops a new medicine against headache. In order to test its effectiveness, people with a headache (the subjects) are given the medicine, after which an evaluator asks these persons if they feel better. Performing such an experiment in this way introduces some risks.
In the first place, if people are aware that they have just taken a newly developed drug against headache, it is well possible that they will feel better only because they know they took the medicine. This is a psychologic factor, but a very real one, which may seriously bias the outcome of the test. It is much better if the subjects do not know if they took the medicine or not. Rather than selecting people with a very short memory for these sort of tests, a more efficient approach is the use of placebos or controls. Placebos, in this case, could be pills which look identical to the pills containing the drug. However, the placebos contain nothing but calcium, or anything that does not affect the physiology of the human body. All the subjects take a pill, however, none of them knows if their pill is a real one, "or a hoax" (that is a bad way of phrasing it, but it just came out of my fingers. You know what I mean!). With this approach, when the subjects are asked if their headache is better, their answer cannot be biased anymore.
However, there still may be a problem. Suppose one of the subjects is asked if the headache got less, and he or she replies "Well, I guess so..." If the evaluator already knows at this point that the subject took a placebo (hence, not the real medicine), he may be tempted (even without realizing it!) to interpret that answer as "not really, apparently!" Similarly, if the evaluator knows the person took the real medicine, he may interpret it rather like "yes, it got better". If the evaluator was involved in the development of the drug (which is often the case), he probably also hopes that the medicine works, so that would create another bias in the outcome. I guess so? That means: YES! "I knew it! It works!"
In order to prevent this second source of bias, it is better if also the evaluator does not know if the subjects took a placebo or the real thing. This is called a double-blind test. Of course, there are many variants of double-blind tests, but the principle is always the same.
No. Blind tests are typically needed when you need to determine small differences in given groups of data. That means, differences that are so small that while measuring them, even a small change in interpretation may be significant to the outcome. If you cannot explain how the use (or the lack) of a double blind test could effect your results, you may probably perform your experiment in a more straightforward manner. Obviously, you should not use blind protocols if you don't understand why! For example, if you see immediately that there is a difference of a factor of two between the quantity to be measured in two groups of data, a blind test would not add much. Secondly, blind tests are typically used in statistical examinations to investigate if an effect actually exists or not. However, if the effect is so strong that you can actually measure it quantitatively (i.e., you can recognize a certain organized behavior in the data), the whole argument about single or double blind testing becomes more or less obsolete.
An example: the graph below shows the brain weight of a collection of mammals versus their body weight. (The original data can be found here.)
From this graph, you may conclude immediately that there exists a clear, more or less linear relationship between the two measured quantities. No double blind tests have been performed, nor would that have been necessary. The data speak for themselves.
Here is another example. Suppose you perform node length measurements on a collection of stem samples. If you find only a very small difference in average node length between the collections of stems, you may have to perform a blind test. However, when crop circle tests show differences in measured data, the effect is usually not small. This can be seen in the graph below. The different yellow bars indicate the average length of growth nodes at various locations inside a crop circle.
As you can see, the shortest nodes were about 2 mm, whereas the tallest were 4.3 mm long. That is more than two times longer. This is a typical result. When you are measuring such enormous differences, and you recognize such a clear structure in your data, there is no additional value in performing the measurements in a blind manner. Those who make the claim that this is necessary (like Joe Nickell does), basically say that you have to be blindfolded before you can reliably determine that there is a size difference between a dog and a horse!
When you work with experimental systems with limited data variability and little experimental error, and you find large differences between individual measurements (as in the case of many crop circle related experiments), the data speak for themselves, and "you should not interrupt", as they say. All the crop circle related research that has been published in the scientific literature (Levengood's papers and my own) fulfilled these conditions. Blind tests (even though they were used!) have no added value in these cases, and anyone who claims the opposite is simply wrong.
Blinded measurements are only needed when you need to determine small differences.
Eltjo H. Haselhoff.
Much more information about the issues addressed here can be found in the book: The Deepening Complexity of Crop Circles.
