Ganzfeld Experiments: Suggestions please.

I would decline to say either way. There is no objective means to declare the trial a success and you don’t know the probability that the trial would be “successful” by chance. As I mentioned earlier (and see below), these problems might be easier for you to recognise if the subject says “I saw a picture of an animal”.

It is easier to understand if the subject says, "I saw a picture of an animal", but then your example goes to prove my point. Your choice of example demonstrates that you are able to subjectively understand the difference between "I saw a picture of an animal" and "I saw a picture of a bunny with eyeglasses" in terms of identifying whether the subject perceived the target ("eh, maybe" for the former, "holy cow!" for the latter).

Consider these scenarios:

Scenario A:
Parnia is running the AWARE study. Ten people who have some auditory/visual memories from the period of time of their cardiac arrest have been interviewed. One person leaves his body and clearly sees a "bunny with eyeglasses" picture from his OBE perspective. Parnia checks which target was in the room and it was the bunny with eyeglasses.

Result A:
Linda and the rest of the non-proponents say "holy cow!" But for the sake of a little more rigour, Linda gives several different sets of 5 different "bunny with eyeglasses" pictures to a blinded intern and asks her to get the subject to identify whether the picture he saw is among them. He thinks it may be the fourth picture in the first set of 5, but when he gets to the second set of 5, the subject confidently picks out the actual target. Champagne is popped.

Result B:
EF gives the picture set from which the target was drawn, for each of the ten subjects, to the intern. She takes them to four of her friends. They sit down with the pictures and the transcripts and pick a picture for each subject. Two are correct, including the "bunny with eyeglasses" target. They give the results to Parnia who is frustrated by his lack of success.

Scenario B:
Parnia is running the AWARE study. Ten people who have some auditory/visual memories have been interviewed. One says they saw a picture of "an animal of some sort". Parnia checks the target and finds it was a "bunny with eyeglasses".

Result A:
Linda gives several sets of 5 different animal pictures to the intern and gets her to ask the subject if the picture he saw is among them. The subject says that maybe it's the elephant in the first set and maybe it's the bunny with eyeglasses when he looks at the second set, or maybe the camel when he looks at the third set. The result is considered highly equivocal.

Result B:
EF gives the picture set from which the target was drawn, for each of the ten subjects, to the intern. She takes them to four of her friends. They sit down with the pictures and the transcripts and pick a picture for each subject. They pick "bunny with eyeglasses" because it's the only animal in the set. This time they get 5 correct and Parnia excitedly writes it up as promising.

So the truth of the situation is that one person spotted the target and the rest didn't. Which scenario and result accurately identified when this amazing result occurred and when it did not? Which scenario and results created results which were false - falsely negative and falsely positive?

I’ve already suggested that a forced choice procedure be used. However, you would need to randomly select the target from the same set of images used in the judging phase. That would give you an objective means to establish whether the trial is a success and you can work out the probability of observing a success by chance. You should also dispense with drawing conclusions based solely on subjective judgements as you were doing previously.

It doesn't give us a means of establishing whether a trial is a success because the subject saw the target, though. All it does is give us the opportunity to identify the target based on judging preferences and chance, when previously, it was almost impossible to identify the target unless the subject had actually seen the target. By turning it into a four-picture, forced-choice situation, you have manufactured a way for the target to be identified which has nothing to do with perceiving the target. And the effect of chance and judging preference is vastly greater than any real effect - so much so that the incredible number of false-positives you are generating will completely engulf the few true-positives you may be lucky enough to obtain. This means that you won't be able to reach the "statistical significance" you need to show that there is an effect. It also means that even if you do manage some "statistical significance", you won't have any idea which one or two, of your dozens or hundreds of "correct" results, actually involved perception of the target.

The way Parnia has it set up now, there is essentially no way to generate a "hit" unless the subject perceives the target. That is, it is essentially impossible to generate a false hit. The way you want to set it up, a "hit" can very easily be generated, so that essentially every hit which you generate will be false.

Linda
 
It would seem less compelling to me but I would decline to say whether the subject perceived the target or not for the reasons I mentioned previously.

Exactly. Even though you make out like this is a subjective opinion which will vary widely, in reality, pretty much everybody but the stupid or stubborn will find "I saw an animal" less compelling than "I saw a bunny with eyeglasses", and would be reluctant to say the subject saw the target, in the former, and fairly confident in the latter.

Validity means whether your test actually measures what it is purported to measure. If you test whether most people recognize that "I saw a picture of an animal" is less compelling than "I saw a picture of a bunny with eyeglasses", you would have good validity with respect to measuring how subjective impressions vary and the degree to which they concur. You would have poor validity with respect to measuring whether the subject perceived the target.

How so? Aren't we in agreement that we are reasonably confident that the subject perceived the target when they say "I saw a picture of a bunny with eyeglasses" and not particularly confident when they say "I saw a picture of an animal"?

And doesn't that make your suggested forced-choice procedure ridiculously invalid, considering that pretty much every 'hit' you get is not an example of whether the subject perceived the target?

To further illustrate this point, consider a test for whether people agree that a person standing in one corner of the Ames room looks taller than a person standing in the other corner. You might devise a valid measure of subjective judgement based on simply asking the subjects to estimate how tall they think each person is. No doubt, all subjects would agree that one person looks taller than the other. However, that would not be a valid test of how tall each person actually is.

Agreed, but how is this highly unrealistic and manipulative example relevant? The only reason we think that people have perceptions from an out of body perspective is because they supposedly accurately report on objects from an out-of-body perspective. Which means that it makes sense to focus on those reports, not on largely unrelated forced-choice exercises.

No, I was talking about a forced choice procedure where we create a target set containing four images and one of these images gets randomly chosen as the target. The target is then placed on the shelf. Thus, there is a 0.25 probability of the target being chosen by the blind judge from this set. The judge’s preference for any particular image in the target set doesn’t influence this probability regardless of whether the target is selected before or after the subject attempts to perceive it. I described the precognition example to illustrate this point as it is easier to understand.

But you can't possibly think that a judge selecting a target by guess is an example of a subject perceiving a target?

The forced choice procedure can answer the question “can the subject perceive the target” because the alternative hypothesis “the subject perceives the target” predicts that the target will be chosen more often than chance during judging. You don’t have to specify the relative contribution of signal and noise to each hit. You hypothesise that the presence of a signal will result in a hit rate significantly greater than chance.

But you can already easily identify the signal without surrounding it by all that noise!

What you are trying to decide is whether the subject saying “animal” represents an instance of perceiving the target. Of course, it is possible for someone to suppose that the subject perceived the target but only acquired (or remembered) enough information to be able to make a basic level categorisation (“animal”).

First of all, I'm trying to get you to confirm that "I saw a picture of a bunny with eyeglasses" is sufficient to say that the subject perceived the target. Then I am trying to get you to confirm that "I saw a bunny with eyeglasses" is specific, especially compared to "I saw an animal". Then we can move on to whether or not these less specific descriptions could still represent perception of the target, And what we need to do is treat that the way we treat eye-witness identifications. We already have lots of research on how to tease out valid and reliable identifications from vague descriptions, without also manufacturing false identifications. So when we have a less specific identification, a positive identification using eye-witness procedures may still be compelling to non-proponents when a correct guess by an independent judge definitely wouldn't be.

http://web.williams.edu/Psychology/Faculty/Kassin/files/ET.whitepaper.pdf

It is also possible for someone to suppose that the subject did not perceive the target and said “animal” for non-psi reasons. According to the protocol you have described so far, which option is chosen would be down to subjective opinion, if indeed these options are recognised at all.

Excuse me? The option can only go unrecognized under your protocol. The advantage to my protocol is that it is easy to recognize when the description is less specific. And while you call the opinion "subjective", there won't actually be any variation or subjectivity in that opinion. Everybody can easily recognize that "animal" is not specific to "bunny".

This means that the protocol has poor validity with respect to answering the question “can the subject perceive the target”.

First, it won't have poor validity, since the distinction between "bunny" and "animal" will probably have a hundred percent accuracy with respect to "perceived the target" vs. "may not have perceived the target". Second, your procedure will have almost zero accuracy with respect to those two options. More importantly, it won't distinguish between "bunny" and "animal" when the judge correctly picks out the bunny picture only because "bunny" was the only animal option. That is, your procedure exactly demonstrates the criticism you tried to apply to my procedure.

Yes, a forced choice procedure will give you that because the probability (under the null) of a hit is fixed and does not change according to personal opinion. With the protocol you are suggesting, what constitutes a “success” and the chance probability associated with that success is based on personal opinion.

The probability of a hit does not change based on personal opinion. Whether or not a description is specific enough to be called a hit might, but the use of eye-witness identification procedures would obviate that. Regardless, this is still way, way, way better than a procedure where a 'hit' tells you nothing about whether there was a specific description of a target.

Like I said before, if you have 1 million images and select one at random, the probability that a given image is the target will be 0.000001. However, what you are trying to assess is whether the trial can be deemed a success...

Yes. The trial is a success if the subject gives a report specific to the target. Further investigation can be done on less specific descriptions using the valid and reliable techniques listed in the report I linked to above.

...and the probability that the trial is a success if the null is true.

Not really. It is near impossible to generate a "hit" if the null is true, so the main focus is identifying whether or not a trial is a success.

You seem to be happy to have these probabilities depend on a number of variable subjective factors that I have already mentioned.

Yes, because as even you have admitted, your opinion as to whether a description is specific is the same as mine and everyone else's. That is, whether or not a trial is a success is an easy call to make, doesn't vary from person to person, and identifies when a target has been perceived.

This will give you poor validity with respect to assessing whether the subject has actually perceived the target.

So when the target was a picture of a bunny with eyeglasses, "I saw a bunny with eyeglasses" is not a success? "I didn't see anything" is not a failure? And "I saw an animal" is not a "maybe"? And you would expect to find almost no agreement if you asked for the opinion of everyone on this forum?

I agree that the way to go would be a forced choice procedure (in which case you could dispense with the previous subjective evaluation about whether the response is “specific” or “barely specific” because it leads to the problems I’ve mentioned before). However, for a valid forced choice procedure, you would need to randomly select the target from the same set that is used for the blind judging. I’ve already suggested that this be done.

No, this is exactly the wrong way to go about it, for the reasons outlined on page 23 of the article I linked under "Structure of Lineup or Photospread".

Linda
 
Pretend Parnia used your method in the AWARE study. He asks everyone whether they saw a target - each says no. He then has them make their forced choice and gets a hit rate of 30%.

What do we conclude about whether anyone saw the target?

The same thing they conclude when 30% of subjects report hits in a ganzfeld study.

Must be a trick question.
 
Pretend Parnia used your method in the AWARE study. He asks everyone whether they saw a target - each says no. He then has them make their forced choice and gets a hit rate of 30%.

What do we conclude about whether anyone saw the target?

Presumably, the first research question they are asking is whether the subject can explicitly perceive a image placed on a shelf during an NDE/OBE. If nobody reported an experience of perceiving an image on a shelf, they should regard the study as null in this respect (because they didn't manage to measure any responses).

If they followed this up and gave a forced choice test, presumably they would be testing some sort of implicit processing of the target. If the 30% hit rate is significant, they might conclude that this is evidence that the subject processed information about the target implicitly through some unknown means.
 
I'm not keen on reintroducing subjectivity in to some of these tests; we already have old studies that used subjective judging and they were derided for using subjective judging, kind of like how one of the remote viewers is claimed to have been judged as drawing a picture of a crane that turned out to later be a soviet warehouse. So we have purely objective measures now to combat the subjectivity argument, and we have skeptics who want to reintroduce the subjectivity argument. I can pretty much guarantee you that if we went back to that model, we would just walk back in to "its subjective therefore worthless" which is probably why nobody is interested in doing it.
 
I'm not keen on reintroducing subjectivity in to some of these tests; we already have old studies that used subjective judging and they were derided for using subjective judging, kind of like how one of the remote viewers is claimed to have been judged as drawing a picture of a crane that turned out to later be a soviet warehouse. So we have purely objective measures now to combat the subjectivity argument, and we have skeptics who want to reintroduce the subjectivity argument. I can pretty much guarantee you that if we went back to that model, we would just walk back in to "its subjective therefore worthless" which is probably why nobody is interested in doing it.
I don't know where you're getting the idea that subjectivity has been derided, that there is an objective measure available in this case, and that researchers have otherwise removed the use of subjective measures.

You want to guard against expectation influencing your outcome measure (e.g. someone judges a match to be close when they know it's the target, and not as good when it is known not to be the target). Subjective measures are more easily influenced by expectation, than objective measures. But that's the point of incorporating blinding and other methods into the design. What's most important is that the measure is valid. And in many cases, the most valid measures are subjective, and there isn't a decent 'objective' substitute. Parapsychology, quite appropriately, uses many subjective measures.

With respect to remote viewing, I suspect the derision may come from performing non-blinded evaluations, not from performing subjective evaluations.

Linda
 
Hey everyone,

Just checking in to say that I've not given up on writing that analysis I mentioned in post #826, but priorities are priorities (especially now that we're getting into finals period here at my university). It'll be my policy to keep making these little posts, intermittently, until such time as I find I have the resources again. Make no mistake, this stuff is time-consuming! Perhaps more than any other post, this one will require copious reading and research; I may decide to preserve the material for a blog as well.
 
Back
Top