Pigasus Awards & Sheldrake

[CFLarsen]

(actually, unknown callers were included in the experiment)

Including unknown, known, animals or the speaking clock. It doesn't matter as long as the correct controls are in place.

Hello, there.

Third time: Are you going to pay for the two experiments you wanted done (post #20)?

 

[davidsmith73]

Aah! Assumption. That great leveller

Just to be clear, I don't think you can make definitive judgements about the effectiveness of your controls from the results alone. You have to provide yourself with a mechanism as to why your controls are ineffective. Once you have remedied this you continue with your experiment. In our telephone telepathy case, if we get postive results with known callers and null results with unknown callers and both are satisfactorily controlled for collusion, we shouldn't conclude that collusion is at work. That would be contradicting your intial assessment of the effectiveness of your controls and would be reveal confirmation bias on the part of the experimenter. The more scientific thing to do would be to accept the results and make conclusions accordingly.

 

[Paul C. Anagnostopoulos]

Controlling for all possible leaks through the known senses is only possible if you can enumerate all possible ways in which information can be passed through the known senses. Does anyone actually believe we can do this?

This is one reason why a null hypothesis like "information wasn't passed through the known senses" is so unsatisfying and why the null hypothesis is actually "the p value will be greater than .05."

How do other sciences combat this problem? At some point in their history, some fundamental experiment becomes repeatable, so they can begin to derive hypotheses from theories about how the phenomenon actually works. Such a hypothesis pinpoints a small aspect of the theory to determine whether it is, in fact, a correct piece of the puzzle. Since the basic phenomenon has become repeatable, it's no longer a question of whether the experiment will work at all, but a question of how changing one aspect of it affects the results.

Parapsychology is still in data collection mode.

~~ Paul

 

[davidsmith73]

Controlling for all possible leaks through the known senses is only possible if you can enumerate all possible ways in which information can be passed through the known senses. Does anyone actually believe we can do this?

This is one reason why a null hypothesis like "information wasn't passed through the known senses" is so unsatisfying and why the null hypothesis is actually "the p value will be greater than .05."

How do other sciences combat this problem? At some point in their history, some fundamental experiment becomes repeatable, so they can begin to derive hypotheses from theories about how the phenomenon actually works. Such a hypothesis pinpoints a small aspect of the theory to determine whether it is, in fact, a correct piece of the puzzle. Since the basic phenomenon has become repeatable, it's no longer a question of whether the experiment will work at all, but a question of how changing one aspect of it affects the results.

Parapsychology is still in data collection mode.

~~ Paul

I agree with most of what you've said. With regards to experiments that can achieve a level of repeatability that should attract mainstream science, I believe the presentiment experiments may already be there, if not now then only a few more years.

With reagrds to eliminating all possible leaks, if it can't be done in principle then at some point we have to stop and say "these positive results are not due to some leak we haven't thought of yet". If you don't do that then how are the experiments supposed to pass the repeatability test?

Even if a repeatable psi experiment were demonstrated, would the sceptical community continue to try to look for potential flaws or instead accept the phenomena as real?

 

[Zep]

You can prevent collusion with the correct experimental procedure.

Of course you can. And the method includes mixing up known and unknown callers, so the recipient does not know which type are calling. Which is precisely what I indicated initially.

If you've controlled for collusion, then you can't turn to it as an explanation if trials with known callers produce results and unknown callers do not. If you did, then you are admitting that you actually didn't control for collusion at all! That's not how science works.

See above: The operative condition is indeed successful control. Then you would indeed be right in your assertion.

Incidentally, did Sheldrake do this?? Hmmm? (A: No, if you are still wondering.)

 

[calebprime]

I agree with most of what you've said. With regards to experiments that can achieve a level of repeatability that should attract mainstream science, I believe the presentiment experiments may already be there, if not now then only a few more years.

can you explain to one who is not too knowledgeable what these experiments are, what the results are, and why you think they are repeatable?

Since the basic phenomenon has become repeatable, it's no longer a question of whether the experiment will work at all, but a question of how changing one aspect of it affects the results.

Parapsychology is still in data collection mode.

~~ Paul

My question is basically the same: Are there any repeatable experiments that show that there is anything like telepathy?

In other words, is there something worth checking out here because there's a real phenomenon?

 

[davidsmith73]

Of course you can. And the method includes mixing up known and unknown callers, so the recipient does not know which type are calling. Which is precisely what I indicated initially.

Simply mixing up unknown callers with known does not prevent collusion at all. If we did what you suggest, but did not prevent the known callers from signalling to the recipient with a text to a concealed mobile phone set to vibrate, then we would end up with positive results for known callers and negative results for unknown callers. The solution is to prevent concealed mobile phones, not to simply mix known and unknown callers.

See above: The operative condition is indeed successful control. Then you would indeed be right in your assertion.

see my response.

Incidentally, did Sheldrake do this?? Hmmm? (A: No, if you are still wondering.)

Correct.

But lets be clear that I'm arguing the case for a properly controlled experiment with only known callers. What you are suggesting does not address collusion because simply mixing up unknown callers does nothing to prevent it when known callers are undertaking their trials.

 

[Zep]

Simply mixing up unknown callers with known does not prevent collusion at all. If we did what you suggest, but did not prevent the known callers from signalling to the recipient with a text to a concealed mobile phone set to vibrate, then we would end up with positive results for known callers and negative results for unknown callers. The solution is to prevent concealed mobile phones, not to simply mix known and unknown callers.

You have mixed up two unrelated issues - deliberate cheating, and unwitting collusion. You have described a method of deliberately circumventing ALL reasonable control - that's cheating, pure and simple. I was describing a method to eliminate unwitting collusion as a confounder - removing a possible variable in the testing that could confound the results.

You do realise, I hope, that the whole idea of experimental design is to ensure only ONE variable is under test at any one time? If you have two or more, the results will not be considered reliable. Do you not agree?

But lets be clear that I'm arguing the case for a properly controlled experiment with only known callers. What you are suggesting does not address collusion because simply mixing up unknown callers does nothing to prevent it when known callers are undertaking their trials.

What you are suggesting is that Sheldrake introduce or allow one or more confounding factors to exist in the testing. Prior knowledge of who is calling is an obvious confounding factor in this testing design.

 

[davidsmith73]

You have mixed up two unrelated issues - deliberate cheating, and unwitting collusion. You have described a method of deliberately circumventing ALL reasonable control - that's cheating, pure and simple. I was describing a method to eliminate unwitting collusion as a confounder - removing a possible variable in the testing that could confound the results.

Ok, so I didn't realise you mean't "unwitting collusion" (which is a bit of an oxymoron by the way. Collusion, by definition, is deliberate).

So what kind of "unwitting collusion" are you talking about and why can't we do an experiment with only known callers that controls for it?

What you are suggesting is that Sheldrake introduce or allow one or more confounding factors to exist in the testing. Prior knowledge of who is calling is an obvious confounding factor in this testing design.

You've completely lost me. Firstly, I'm not talking about Sheldrake. I'm talking about our hypothetical experiment. Secondly, I have no idea how you can say that doing an experiment with only known callers is "introducing a confounding factor". Please explain.

 

[Paul C. Anagnostopoulos]

I agree with most of what you've said. With regards to experiments that can achieve a level of repeatability that should attract mainstream science, I believe the presentiment experiments may already be there, if not now then only a few more years.

Then it's theory time!

With reagrds to eliminating all possible leaks, if it can't be done in principle then at some point we have to stop and say "these positive results are not due to some leak we haven't thought of yet". If you don't do that then how are the experiments supposed to pass the repeatability test?

Certainly they don't have to be 100% repeatable. The problem with psi experiments is that they are all about information transfer through the senses. This makes the whole sensory leak issue paramount. So also is the statistical analysis, because the result, as of now, can only be seen through statistics.

Even if a repeatable psi experiment were demonstrated, would the sceptical community continue to try to look for potential flaws or instead accept the phenomena as real?

I'd continue to look for flaws, just in case there is some information transfer mechanism, based on the known senses, that we haven't discovered yet. In the meantime, as the theory-based experiments are more common and successful, it becomes harder to scoff at parapsychology.

I would push researchers to develop theories and test them, even if the theories are really shakey. For example, let's try to find out whether psi is based on electromagnetic energy transfer. Faraday cages should do the trick. That would seem like a fundamental mechanistic question worth answering.

~~ Paul

 

[davidsmith73]

can you explain to one who is not too knowledgeable what these experiments are, what the results are, and why you think they are repeatable?

Basically, take two sets of pictures. One is a set of calm pictures such as landscapes, calm seas etc. The other set is emotional pictures like pornography or extremely violent car crash images. Present the pictures in a randomised order to someone and continually measure some aspect of that persons physiological response to the pictures, from a few seconds before picture presentation untill a few seconds after. Typically, skin conductance is measured because it gives a well characterised differential response to calm vs emtional stimuli, but experiments have been done with heart rate and fMRI. Even when the pictures are adequately randomised, the physiological measure just before picture presentation seems "predictive" of which picture will appear in the near future. Various normal explanations have been put forward most notably being the "gamblers fallacy". That's where it gets complicated, but follow these links if you're interested.


One of Radin's
http://www.boundaryinstitute.org/articles/presentiment99.pdf

Bierman's. Also, go to his publication page for more
http://m0134.fmg.uva.nl/publications/2002/fmri.presentiment.pa2002.doc

http://m0134.fmg.uva.nl/publications/2002/expectationbias_PA2002.pdf

May and Spottiswoode. Experimental design addresses the expectation bias problems associated with previous designs
http://www.lfr.org/LFR/csl/library/SCpsr.pdf

http://www.lfr.org/LFR/csl/library/MPZjacm.pdf

 

[Zep]

Ok, so I didn't realise you mean't "unwitting collusion" (which is a bit of an oxymoron by the way. Collusion, by definition, is deliberate).

So what kind of "unwitting collusion" are you talking about and why can't we do an experiment with only known callers that controls for it?




You've completely lost me. Firstly, I'm not talking about Sheldrake. I'm talking about our hypothetical experiment. Secondly, I have no idea how you can say that doing an experiment with only known callers is "introducing a confounding factor". Please explain.

The answer to both parts is that it turns it into a guessing game, NOT a test of "I wonder who is calling me?". And while you may think that is a fair test (it is, in a pure sense), it is NOT what the original research was investigating.

For example, should the claimant say: I can tell who is phoning me before I pick up and I'm right about 25% of the time, if the callers were random people out of the phone book then that's a highly significant result.

But if they made the same claim, but the callers were limited to four people they knew well, it's an insignificant "chance" result, and one that most people would get over a long series of tests by simply guessing. And with a small sample to work from, it's not at all inconceivable to get whole strings of "correct" guesses in a row in tests.

That's a fairly obvious difference, don't you think?

 

[Beth]

You do realise, I hope, that the whole idea of experimental design is to ensure only ONE variable is under test at any one time? If you have two or more, the results will not be considered reliable. Do you not agree?

No. This is a misconception about how to design experiments. The entire area of DOE - Design of Experiments - is all about varying multiple factors in a controlled way and making sense out of the results. Varying only one factor at a time doesn't take interactions between factors into account. If you try to run experiments at all levels of all variable in order to assess interactions, the number of experiments goes up exponentially with the number of factors and levels of each factor.

When I was learning DOE, varying one one factor at a time was referred to as 'fishing with one line'. Varying multiple factors in a controlled manner was like fishing with more than one line in the water.

 

[Jeff Corey]

No. This is a misconception about how to design experiments. The entire area of DOE - Design of Experiments - is all about varying multiple factors in a controlled way and making sense out of the results. Varying only one factor at a time doesn't take interactions between factors into account. If you try to run experiments at all levels of all variable in order to assess interactions, the number of experiments goes up exponentially with the number of factors and levels of each factor...

Are you saying this?
In one factorial experiment, the number of combinations of levels of factors increases as the number of factors and levels do, i. e., 2x2, 2x3,3x3,etc. The point of running a factorial experiment is to not have to run multiple experiments and to be able to assess interactions.

 

[davidsmith73]

The answer to both parts is that it turns it into a guessing game, NOT a test of "I wonder who is calling me?". And while you may think that is a fair test (it is, in a pure sense), it is NOT what the original research was investigating.

It is what the original research was investigating. Read the original Sheldrake survey paper. I'll quote from the abstract.

"Many people claim to have thought about a particular person who then calls them on the telephone."
http://www.sheldrake.org/Articles&Papers/papers/telepathy/pdf/telephone_telcalls.pdf

Reading the rest of the survey, I think its clear that people are refering to callers whom they already have aquaintance with, rather than, say, wrong number callers. I don't think anyone who took part in that survey, if asked "have you ever known when a wrong number caller is about to call you", would answer yes.

Furthermore, how would you propose to design an experiment with 100 unknown callers and a few known callers that would not involve a "guessing game"?

BTW, I'm not saying that using 100 unknown callers is invalid but you seem to think that using only 4 known callers is somehow invalid. I can't see where you've explained this.

For example, should the claimant say: I can tell who is phoning me before I pick up and I'm right about 25% of the time, if the callers were random people out of the phone book then that's a highly significant result.

Correct, especially if the recipient can name the callers! However, such an experiment is stupendously unrealitic.

But if they made the same claim, but the callers were limited to four people they knew well, it's an insignificant "chance" result, and one that most people would get over a long series of tests by simply guessing.

In which case we would have an unsuccessful result. So where does "unwitting collusion" come into it?

And with a small sample to work from, it's not at all inconceivable to get whole strings of "correct" guesses in a row in tests.

Then work with a large sample size.

Again, what do you mean by "unwitting collusion" and where would it get into a well controlled experiment with known callers?

 

[davidsmith73]

sorry bout the messed up message. Don't seem to have permission to edit

 

[Beth]

Are you saying this?
In one factorial experiment, the number of combinations of levels of factors increases as the number of factors and levels do, i. e., 2x2, 2x3,3x3,etc. The point of running a factorial experiment is to not have to run multiple experiments and to be able to assess interactions.

The concept extends beyond simple factorial experiments, but that's the basic idea, yes.

 

[CFLarsen]

davidsmith73,

You are not ignoring my question, are you?

 

[bfinn]

Basically, take two sets of pictures. One is a set of calm pictures such as landscapes, calm seas etc. The other set is emotional pictures like pornography or extremely violent car crash images. Present the pictures in a randomised order to someone and continually measure some aspect of that persons physiological response to the pictures, from a few seconds before picture presentation untill a few seconds after. Typically, skin conductance is measured because it gives a well characterised differential response to calm vs emtional stimuli, but experiments have been done with heart rate and fMRI. Even when the pictures are adequately randomised, the physiological measure just before picture presentation seems "predictive" of which picture will appear in the near future.

Yup, I read about these experiments a while back and the results are extremely interesting & (if true) alarming. Because this would show that causation can go backwards in time - which is philosophically problematic.

And it would have other interesting consequences. You could easily turn this into a short-term future prediction machine that could predict anything. E.g. (in another interest of mine) suppose you want to predict whether a share price is going to go up or down in the next few seconds. Build a machine where the share price direction selects the next image. Then run the experiment, and your skin conductance will predict the imminent share price move (because it will predict the image which is effectively a readout of the imminent share price move). An instant money machine.

However I don't think you can get a contradiction out of this (as in the so-called 'grandfather paradox' or 'autoinfanticide' - in which you cause the past death of yourself or your grandfather, thus preventing you initiating the cause). For as long as the causation is indeterminate - i.e. skin resistance doesn't predict the image with absolute certainty - then there is an 'out'. (The simple paradox which could otherwise be produced is that if e.g. your skin resistance predicts with certainty that a violent image will come up, the machine uses this to prevent the violent image appearing, thus contradicting the prediction.)

 

[Paul C. Anagnostopoulos]

No. This is a misconception about how to design experiments. The entire area of DOE - Design of Experiments - is all about varying multiple factors in a controlled way and making sense out of the results. Varying only one factor at a time doesn't take interactions between factors into account. If you try to run experiments at all levels of all variable in order to assess interactions, the number of experiments goes up exponentially with the number of factors and levels of each factor.

I can't imagine how this could possibly work unless the basic experiment is solidly replicable and there is a good theory about how the factors interact. If you do this with a psi experiment, you're just going to be scratching your head wondering whether you've introduced a leak.

~~ Paul