Evidence against psi?

[Paul C. Anagnostopoulos]

In the latest issue of the Journal of Cognitive Neuroscience, Moulton and Kosslyn published an article titled "Using Neuroimaging to Resolve the Psi Debate."

http://jocn.mitpress.org/cgi/content/abstract/20/1/182

The abstract is not particularly enlightening. Has anyone read the article, or do you have a subscription to the journal?

Aha, here is more information on the study:

http://mindblog.dericbownds.net/2007/12/using-neuroimaging-to-resolve-psi.html

~~ Paul

[TV's Frank]

Just switch to metric.

[Gord_in_Toronto]

Just switch to metric.

So what are your thoughts on KPSM?

My thoughts on the OP.

Looks good!

. . . psi stimuli and non-psi stimuli evoked indistinguishable neuronal responses—although differences in stimulus arousal values of the same stimuli had the expected effects on patterns of brain activation. These findings are the strongest evidence yet obtained against the existence of paranormal mental phenomena.

Let's see what the believers have to say.

[El Greco]

Darn, I read the thread title and for a moment I thought II was back :D

[Paul C. Anagnostopoulos]

Let's see what the believers have to say.
This came up on another forum and one person said the study is flawed because it assumes psi is in the brain. He believes that the brain is merely a receiver for psi events. I pointed out that even if this is the case, there are brain functions involved in receiving the information that should be different from the functions involved in internal thought.

~~ Paul

[Paul C. Anagnostopoulos]

Darn, I read the thread title and for a moment I thought II was back.
No, but he does hang out at the other forum I mentioned above. He's the one who posted the article there.

~~ Paul

[Rodney]

This came up on another forum and one person said the study is flawed because it assumes psi is in the brain. He believes that the brain is merely a receiver for psi events. I pointed out that even if this is the case, there are brain functions involved in receiving the information that should be different from the functions involved in internal thought.
That may be, but if the overall results in this study were indistinguishable from chance, why would anyone expect the MRIs in the study to be distinguishable from MRIs of similar mental activities where no one claimed anything paranormal was taking place?

[Paul C. Anagnostopoulos]

That may be, but if the overall results in this study were indistinguishable from chance, why would anyone expect the MRIs in the study to be distinguishable from MRIs of similar mental activities where no one claimed anything paranormal was taking place?
I agree that if there were no statistically significant psi results, then the study isn't particularly interesting. Of course, one might argue that even attempting psi communication should result in interesting brain function, since one must be engaging "psi information transfer modules" in the brain. Either that or psi is purely passive.

~~ Paul

[davidsmith73]

I agree that if there were no statistically significant psi results, then the study isn't particularly interesting. Of course, one might argue that even attempting psi communication should result in interesting brain function, since one must be engaging "psi information transfer modules" in the brain. Either that or psi is purely passive.

~~ Paul

I read the full paper. I thought it was interesting in that a mainstream journal had published such a study. Brain imaging is potentially a powerful tool for studying psi effects but I have some criticisms of the paper.

Firstly, I agree with you and Rodney that this study isn't particularly helpful because they did not find any significant behavioural data. But what baffled me is that the authors make no issue of this in their conclusions and discussion. Effectively, they are making conclusions from fMRI data when their behavioural data shows that the participants were not doing what they were supposed to be doing, ie, producing psi. This kind of criticism can be levelled at some conventional imaging research too. Sometimes, imaging papers just don't report their behavioural data which is quite bad practice. We really need to know how effectively the participants were performing the task in order to make conclusions about the brain data. We are, after all, trying to explain behavioural data in terms of what the brain is doing. One of the reasons that they got no behavioural psi might have simply been due to the environment of the scanner. They are very noisy things and no effort was made to reproduce a ganzfeld type state for the receivers. Indeed, it would be very hard to do in an fMRI experiment I imagine. So, no psi no brain data.

Secondly, and this is what I suspect, their experimental design might have confounded their analysis. What they did was to combine a "telepathy", "clairvoyance" and "precognition" design into every trial. Telepathy was provided by the sender viewing the target, clairvoyance by the sender's computer monitor displaying the target and precognition by immediate feedback of the correct target to the receiver after they made a binary choice between the target and decoy. However, the receiver's task was primarily geared towards precognition. In other words, their attention was very probably devoted to predicting which picture was going to be displayed as feedback because that was the task they were actually performing. They could well have just forgotten about or ignored the telepathy and clairvoyance aspect of the design. This may be cruical if psi performance in the laboratory depends, to a certain extent, on the participants strategy.

So if participants were primarily engaging in a precognitive strategy, would the authors' analysis necessarily be looking in the right place? The answer to this question is really - we don't know. They looked at the difference in brain activation during the period when the participant viewed the decoy and target pictures (before they received feedback on the correct one). So they were looking for different activation during viewing of the target. But why should we assume that this is the period when psi information expresses itself? In fact, their whole premise for the study is that differential activation occurs during the viewing of the target picture. This may not be the case at all. It may the case that viewing an external stimulus inhibits any brain activation caused by psi by drawing attention away from internal imagery! There was a whole load of other trial periods when differential brain activation could have been expressed.

IMO, what they should have done is a presentiment analysis, especially in light of the fact that no overt psi was found in the behavioural data. During the period when the participants were making a binary choice as to which picture was the real target, they should have subtracted the brain activation during a correct choice from an incorrect one (but before the real target was displayed as feedback). According to the null hypothesis, there should have been no difference, and this contrast would have nicely controlled for expectation effects since expectation would have been equal, on average, for correct and incorrect choices according to the NH.

[Henners]

since one must be engaging "psi information transfer modules" in the brain.

Isn't is also possible that the psi signals are relayed to the brain by receptors elsewhere in the body - up the rectum, say.

[Paul C. Anagnostopoulos]

Davidsmith, do you have a source for the paper online?


Firstly, I agree with you and Rodney that this study isn't particularly helpful because they did not find any significant behavioural data. But what baffled me is that the authors make no issue of this in their conclusions and discussion.
How odd. How did they note that they didn't find any significant behavioural data?


Secondly, and this is what I suspect, their experimental design might have confounded their analysis. What they did was to combine a "telepathy", "clairvoyance" and "precognition" design into every trial. Telepathy was provided by the sender viewing the target, clairvoyance by the sender's computer monitor displaying the target and precognition by immediate feedback of the correct target to the receiver after they made a binary choice between the target and decoy. However, the receiver's task was primarily geared towards precognition.
I'm not sure this is so bad. After all, every psi experiment is a test of at least two types of psi, right? Is there any way to eliminate clairvoyance from a psi study?


IMO, what they should have done is a presentiment analysis, especially in light of the fact that no overt psi was found in the behavioural data. During the period when the participants were making a binary choice as to which picture was the real target, they should have subtracted the brain activation during a correct choice from an incorrect one (but before the real target was displayed as feedback). According to the null hypothesis, there should have been no difference, and this contrast would have nicely controlled for expectation effects since expectation would have been equal, on average, for correct and incorrect choices according to the NH.
That sounds better. Of course, we could still argue that the entirety of the psi work was done two weeks before the experiments, via precognitive clairvoyance.


Isn't is also possible that the psi signals are relayed to the brain by receptors elsewhere in the body - up the rectum, say.
And then imagine how much information you could get if you could clairvoyantly analyze the receptors directly!

~~ Paul

[Rodney]

Effectively, they are making conclusions from fMRI data when their behavioural data shows that the participants were not doing what they were supposed to be doing, ie, producing psi.
And yet, the authors seem to think that their "findings are the strongest evidence yet obtained against the existence of paranormal mental phenomena." That conclusion speaks volumes about the authors' motives.

[davidsmith73]

How odd. How did they note that they didn't find any significant behavioural data?

Just a couple of lines on the overall hit rate and no significant individual performances.


I'm not sure this is so bad. After all, every psi experiment is a test of at least two types of psi, right? Is there any way to eliminate clairvoyance from a psi study?

There's not really any way to eliminate clairvoyance as far as I can tell. As soon as either the analyst, experimenter or receiver views the correct target this will introduce the concept of clairvoyance and since there is no way to tell if psi has happened without looking at the correct target then the concept of clairvoyance is inevitable. Like you say below, even if the target is only revealed in the future, it's still a type of precognitive clairvoyance. Perhaps this means that there is no real difference between these types of psi. The only difference is in the way we conceptualise the circumstances by which the information is obtained. In other words, maybe we can theorise a way for all three types to fall under one process. But my point was that the participants were most likely encouraged to use a strategy that they thought was precognitive, due to the nature of the task. This, together with the lack of behavioural psi suggests that the differential brain activity might have actually occured during the binary choice, which the authors didn't analyse. However, it appears they have brain activity data during this period so maybe they could perform a suitable post hoc test on which to base a further study. But I think their minds have been made up...


That sounds better. Of course, we could still argue that the entirety of the psi work was done two weeks before the experiments, via precognitive clairvoyance.

Indeed! Such is the nature of doing experiments on things we know very little about. But we have to make some assumptions about the limits of this process in order to get a practical experimental design. As the paper stands, I think their assumptions about when differential activity should happen were perhaps too stringent.

[davidsmith73]

And yet, the authors seem to think that their "findings are the strongest evidence yet obtained against the existence of paranormal mental phenomena." That conclusion speaks volumes about the authors' motives.

I don't like to speculate on motivations but I do wonder if the authors would have made conclusions on fMRI data in the absence of behavioural performance if the subject matter was a conventional psychological process.

The only hint of this issue is in the introduction when they say "although the unexpected presence of a behavioural effect mandates a neural correlate, the expected absence of a behavioural effect does not necessarily imply null neuroimaging results". But they fail to address the possibility that the absence of a behavioural effect might actually imply null neuroimaging results! Curious...

[Paul C. Anagnostopoulos]

I have sent an email to Stephen Kosslyn asking him this very question.

~~ Paul

[Rodney]

I don't like to speculate on motivations but I do wonder if the authors would have made conclusions on fMRI data in the absence of behavioural performance if the subject matter was a conventional psychological process.

The only hint of this issue is in the introduction when they say "although the unexpected presence of a behavioural effect mandates a neural correlate, the expected absence of a behavioural effect does not necessarily imply null neuroimaging results". But they fail to address the possibility that the absence of a behavioural effect might actually imply null neuroimaging results! Curious...
It seems to me that an appropriate analogy would be if someone were investigating the claim that "HeadOn" cures headaches. The investigator rounds up a few subjects with headaches and administers HeadOn to them. A few minutes later, the subjects are asked whether their headaches are gone, and all answer no. However, the investigator insists on taking MRIs of each subject's brain, which reveal images similar to the brains of individuals suffering from headaches. The investigator then concludes that "the findings are the strongest evidence yet obtained against the belief that HeadOn cures headaches." Would that study be published in the Journal of Cognitive Neuroscience?

[Paul C. Anagnostopoulos]

Your analogy is flawed, Rodney. If you administer a drug that is supposed to cure something and it doesn't, that is good evidence that the drug doesn't cure.

That said, I don't see how the study in question is any more powerful than any other psi study with negative findings. It does establish a link between absence of subjective findings and absence of interesting brain activity, which might prompt others to keep trying to see what they get for fMRI results when there is positive subjective finding.

~~ Paul

[Rodney]

Your analogy is flawed, Rodney. If you administer a drug that is supposed to cure something and it doesn't, that is good evidence that the drug doesn't cure.
The point is that many individuals claim that HeadOn has cured their headaches, and it is those test subjects -- not subjects who state that their headaches have not been cured -- who should be given MRIs. Now, perhaps for completeness, all test subjects could be given MRIs, but the investigator cannot answer the question under consideration if none of the subjects state that their headaches were cured.

That said, I don't see how the study in question is any more powerful than any other psi study with negative findings.
I don't even know that their psi study was done properly. Perhaps, since davidsmith73 has read the full study, he can enlighten us as to their methodology, including the number of study participants.

It does establish a link between absence of subjective findings and absence of interesting brain activity, which might prompt others to keep trying to see what they get for fMRI results when there is positive subjective finding.
Fine, but the absence of interesting brain activity is what would be expected in the absence of above chance results. So I see no basis for the study's conclusions.

[Paul C. Anagnostopoulos]

... but the investigator cannot answer the question under consideration if none of the subjects state that their headaches were cured.
That seems reasonable, yes.


Fine, but the absence of interesting brain activity is what would be expected in the absence of above chance results. So I see no basis for the study's conclusions.
Nor do I.

~~ Paul

[fls]

It seems to me that an appropriate analogy would be if someone were investigating the claim that "HeadOn" cures headaches. The investigator rounds up a few subjects with headaches and administers HeadOn to them. A few minutes later, the subjects are asked whether their headaches are gone, and all answer no. However, the investigator insists on taking MRIs of each subject's brain, which reveal images similar to the brains of individuals suffering from headaches. The investigator then concludes that "the findings are the strongest evidence yet obtained against the belief that HeadOn cures headaches." Would that study be published in the Journal of Cognitive Neuroscience?

Your analogy is flawed. Just like all correct guesses count towards the total when claiming evidence for psi, all headache recoveries would count towards claiming evidence for HeadOn. And like psi, it counts as proof if the number recovered (or the number of correct guesses) exceeds that which you'd otherwise expect in the absence of HeadOn (psi). A more appropriate analogy would be if all subjects are administered HeadOn and a few recover. Scans comparing those who recovered with those who did not show no difference. The investigator concludes that HeadOn provides no objective change, but believers point out that we'd expect a few people to recover due to the placebo effect. So the scans were of people who had recovered due to the placebo effect and not the effect of HeadOn, and this doesn't prove that HeadOn has no objective effect.

Linda

[fls]

Somehow, none of the correct guesses in this experiment count as examples of psi. Yet in any other psi experiment, all correct guesses contribute to the total when determining whether or not the expected number is exceeded. I'd be interested in how you differentiate a correct guess due to psi from a correct guess due to chance.

Linda

[Paul C. Anagnostopoulos]

Somehow, none of the correct guesses in this experiment count as examples of psi. Yet in any other psi experiment, all correct guesses contribute to the total when determining whether or not the expected number is exceeded. I'd be interested in how you differentiate a correct guess due to psi from a correct guess due to chance.
Well now, ain't that an interesting point. The psi proponent shoud be willing to grant that at least some of the correct guesses were due to psi, unless he can show a reason why the protocol suppressed psi completely. If some correct guesses were due to psi, but not overall a statistically significant number, should that have shown up in the fMRI analysis?

In other psi experiments, all correct guesses are counted because, of course, you're trying to determine if the number of guesses exceeds chance by a statistically significant amount. I don't think there is any assumption, however, that all the hits were due to psi. Imagine if you could tell which were which!

~~ Paul

[davidsmith73]

I don't even know that their psi study was done properly. Perhaps, since davidsmith73 has read the full study, he can enlighten us as to their methodology, including the number of study participants.


I can't find flaws in their methodology, as it was described. It had all the controls in place. They had 19 pairs of participants, and to their credit they recruited emotionally close pairs and included some twin pairs. They didn't select for creativity though which would have been even better. They had about 3500 trials overall.

However, the procedure did not put the receiver in any sort of ganzfeld state. This was partly due their experimental design - receivers simply sat in the scanner and looked at two pictures presented one after the other and then decided, with the press of a key, which was the target. The actual target was then presented a few seconds after this key press. Each trial lasted about 15 sec on average and many trials were repeated one after the other. Couple this with the noise of the scanner, then ganzfeld environment it was not! It was effectively a forced choice ESP experiment in a distracting and uncomfortable environment. This may have unfortunately undone any increased chance of psi by selecting the right participants. But hey, this was a first of a kind experiment (almost) and as such we should not expect things to go smoothly, nor should the authors have drawn such strong conclusions IMO.

[davidsmith73]

Somehow, none of the correct guesses in this experiment count as examples of psi.

They were hypthesising that the receiver simply looking at the correct target (before feedback) would produce differential brain activity due to either psi-mediated implicit knowledge of the target (decreased brain activity) or increased attention towards the target (increased brain activity). This may or may not affect the choice hit rate depending on whether the psi mediated information can be accessed by the cognitive systems that are involved in the binary decision process.

[jimbob]

I'd say the eye and the ear are good arguments against psi. It is amazing that slight variations in the environment can be sensed, very reliably by a vast number of animals, and that these senses have evolved on many different occasions, yet psi is very unreliable, and hasn't despite the obvious advantages it would confer on any organism with such a paranormal sense. (Maybe, looking at stories about "fey" running in familise, one could tihink it should be inheritable, if it existed).

I have a simpler idea:

Maybe primitive cultures didn't understand different types of mental illness.

[Paul C. Anagnostopoulos]

Let's suppose that 20% of the guesses were correct due to psi. The remaining 80% should divide into 40% correct and 40% incorrect by chance. This puts the correct guesses at 60%, yet they were only at the expected 50%. So I think that if any significant number of correct guesses were psi-mediated, it would have shown up.

Have I got that right?

~~ Paul

[fls]

They were hypthesising that the receiver simply looking at the correct target (before feedback) would produce differential brain activity due to either psi-mediated implicit knowledge of the target (decreased brain activity) or increased attention towards the target (increased brain activity). This may or may not affect the choice hit rate depending on whether the psi mediated information can be accessed by the cognitive systems that are involved in the binary decision process.

So how do you conclude that the lack of an affected choice hit rate means that psi wasn't present (i.e. this wasn't a test of psi and the conclusions can be ignored)?

Not all psi experiments are done under Ganzfeld conditions.

Your criticisms demonstrate very nicely the problem with parapsychology. Psi is unconstrained. It seems there are no limits placed on when and how it acts, so any failure can be excused. And it doesn't seem to matter that the excuses for one experiment contradict those provided for another experiment. Or that when Bem combines telepathy, clairvoyance, and pre-cognition unrestrictedly in a meta-analysis and gets a 'positive' result nobody complains, but when someone else obtains a negative result, it becomes unjustifiable.

I agree that the authors over-stated their conclusions. Non-believers don't really need additional confirmation and believers will excuse the findings anyway.

Linda

[Rodney]

Your analogy is flawed. Just like all correct guesses count towards the total when claiming evidence for psi, all headache recoveries would count towards claiming evidence for HeadOn. And like psi, it counts as proof if the number recovered (or the number of correct guesses) exceeds that which you'd otherwise expect in the absence of HeadOn (psi). A more appropriate analogy would be if all subjects are administered HeadOn and a few recover. Scans comparing those who recovered with those who did not show no difference. The investigator concludes that HeadOn provides no objective change, but believers point out that we'd expect a few people to recover due to the placebo effect. So the scans were of people who had recovered due to the placebo effect and not the effect of HeadOn, and this doesn't prove that HeadOn has no objective effect.

Linda
The point I was making is that -- even with regard to something as politically correct as bashing HeadOn -- I seriously doubt if a study would be accepted by the Journal of Cognitive Neuroscience unless that study included subjects who claimed that HeadOn cured their headaches.

[Paul C. Anagnostopoulos]

So I got a response from one of the authors. I won't post it, since he may assume it was private. He said that they did not necessarily expect to get any positive subjective results, given the replicability issues with psi experiments. What they were trying to do was see if they could find evidence of psi using more sensitive measures, that is, fMRI.

~~ Paul

[Rodney]

Somehow, none of the correct guesses in this experiment count as examples of psi. Yet in any other psi experiment, all correct guesses contribute to the total when determining whether or not the expected number is exceeded. I'd be interested in how you differentiate a correct guess due to psi from a correct guess due to chance.

Linda
One can never know with certainty about an individual guess; it's simply a matter of overall probabilities. For example, if someone is given a 100-question multiple choice test where there are four choices for each question and gets only 25 correct answers, the likelihood is that (s)he had little, if any, knowledge relevant to that test. If, on the other, (s)he gets 50 correct answers, it is extremely unlikely that (s)he had no knowledge relevant to that test.

[Rodney]

So I got a response from one of the authors. I won't post it, since he may assume it was private. He said that they did not necessarily expect to get any positive subjective results,
I think "necessarily" is superfluous. ;)

given the replicability issues with psi experiments.
I guess it would have taken too much effort to do a ganzfeld study.

What they were trying to do was see if they could find evidence of psi using more sensitive measures, that is, fMRI.
And the way to do that is with a study that is not conducive to psi?

[fls]

The point I was making is that -- even with regard to something as politically correct as bashing HeadOn -- I seriously doubt if a study would be accepted by the Journal of Cognitive Neuroscience unless that study included subjects who claimed that HeadOn cured their headaches.

That's my point. This study is analogous to a study where subjects claim that HeadOn cured their headache, not a study where headaches weren't cured. It's not like nobody made a correct guess.

Linda

[fls]

One can never know with certainty about an individual guess; it's simply a matter of overall probabilities. For example, if someone is given a 100-question multiple choice test where there are four choices for each question and gets only 25 correct answers, the likelihood is that (s)he had little, if any, knowledge relevant to that test. If, on the other, (s)he gets 50 correct answers, it is extremely unlikely that (s)he had no knowledge relevant to that test.

That's the point, though. Out of that fifty, which questions are right because she guessed and which are right because she had relevant knowledge? What if she knew the answer to 20 of the questions and guessed wrong on all but 5 of the rest. Does that mean she really didn't know the answers to those 20 questions?

Linda

[Paul C. Anagnostopoulos]

And the way to do that is with a study that is not conducive to psi?
They did many things to make it conducive to psi. You can't do a classic Ganzfeld in the presence of an MRI machine, at least not until the parapsychology community deigns to agree on a methodology that allows it.

~~ Paul

[Paul C. Anagnostopoulos]

That's the point, though. Out of that fifty, which questions are right because she guessed and which are right because she had relevant knowledge? What if she knew the answer to 20 of the questions and guessed wrong on all but 5 of the rest. Does that mean she really didn't know the answers to those 20 questions?
That could only be the case if you postulate some sort of negative psi, too. Otherwise you have to assume that she would get 15 of the remaining 30 correct and 15 wrong, for an overall score of 35 out of 50.

~~ Paul

[davidsmith73]

So how do you conclude that the lack of an affected choice hit rate means that psi wasn't present (i.e. this wasn't a test of psi and the conclusions can be ignored)?

I wouldn't conclude that necessarily. My complaint was that the authors don't seem to consider this a possible reason why no differential brain activity was found. It's a possible reason among others, and I agree with your point that psi results suffer from somewhat unconstrained interpretations.

But the main problem I have with their analysis is that they did not do enough of it! Even the actual contrast they made between brain activity viewing the decoy minus viewing the target was not reported in a conventional manner. Results of contrasts in fMRI usually show the actual statistical map of the subtraction. Strangely, this paper did not. It only showed us the group data for brain activity during viewing of the decoy and viewing of the target, separately. Presumably, the authors expect the readers to contrast these two activation maps by eye. It does look like a null result to me but its a bit unsatisfying to have to do that.



Your criticisms demonstrate very nicely the problem with parapsychology. Psi is unconstrained. It seems there are no limits placed on when and how it acts, so any failure can be excused. And it doesn't seem to matter that the excuses for one experiment contradict those provided for another experiment.

I agree about the lack of constraints, but that can be remedied by forming hypotheses and testing them. With regards to this experiment we shouldn't really say that the results contradict other studies because its almost the first of its kind looking at high spatial resolution brain activity during a psi task. The only other experiment I know of is Dick Bierman's presentiment experiments using fMRI which got positive results. Again, I just wish they would have looked a bit more at their data.

Prior to this experiment we simply didn't know where or when anomalous brain activity happens during a forced choice experiment. Now we can be a little more confident that it doesn't happen during viewing of the picture set before the choice, but there's still the possibility that it may happen there if an anomalous hit rate is also observed. Its a shame that the authors really only focus on a contrast associated with perception of the picture stimuli rather than the action of deciding which picture to choose. In this experiment, perhaps they might have found differential brain activity during the key press after viewing both pictures, ie, on avergae more activity for correct key presses. But of course, if this was found we would have to conlcude that it was not enough activity to affect the final behaviour. But that is more or less what their hypothesis was anyway, the difference being they were looking at a different period associated with perception for recognition and attention rather than for action.

[Henners]

Isn't looking inside the brain for the purpose of hoping to see something that cannot be demonstrated to exist in external reality - but you happen to have a tool for looking inside the brain, so why not use it - a little like seeing everything look like a nail because you happen to have a hammer, or a like the drunk looking for his keys under a street light because everywhere else is too dark.

[fls]

That could only be the case if you postulate some sort of negative psi, too. Otherwise you have to assume that she would get 15 of the remaining 30 correct and 15 wrong, for an overall score of 35 out of 50.

~~ Paul

I was referring to the scenario that Rodney proposed (100 questions with 4 choices each). It's not meant to be a rigorous example, but it doesn't require negative psi to get more wrong than expected if you take into consideration chance and the patterns/bias that are often present in tests.

Linda

[davidsmith73]

I'd be interested in how you differentiate a correct guess due to psi from a correct guess due to chance.

It seems to me that this is a problem with measuring a variable from a system that has a substantial stochastic element to it. It's very difficult to separate signal from noise on an individual measurement basis.

For example, you apply UV rays to some DNA and observe an increased mutation rate. How do you know which mutations were due to chance (or non-muatgenic factors) and which were due to UV?

It's an interesting empirical question, but I'm not sure its helpful.

[fls]

It seems to me that this is a problem with measuring a variable from a system that has a substantial stochastic element to it. It's very difficult to separate signal from noise on an individual measurement basis.

For example, you apply UV rays to some DNA and observe an increased mutation rate. How do you know which mutations were due to chance (or non-muatgenic factors) and which were due to UV?

It's an interesting empirical question, but I'm not sure its helpful.

It was meant to be rhetorical. The lack of empirical answers is the distinguishing feature of parapsychology. Else it would just be 'psychology' or 'physiology'.

Linda

[Rodney]

That's the point, though. Out of that fifty, which questions are right because she guessed and which are right because she had relevant knowledge? What if she knew the answer to 20 of the questions and guessed wrong on all but 5 of the rest. Does that mean she really didn't know the answers to those 20 questions?
Linda
I understand your point, but it's the same issue with any type of statistical analysis. In your example, while it's certainly possible to know the answers to some questions on a 100-question test and still get only 25 correct (with a probability of a correct guess being .25), the odds of knowing the answers to 20 questions and guessing wrong on at least 75 of the remaining 80 questions is about 1 in 80,000.

[Paul C. Anagnostopoulos]

I was referring to the scenario that Rodney proposed (100 questions with 4 choices each). It's not meant to be a rigorous example, but it doesn't require negative psi to get more wrong than expected if you take into consideration chance and the patterns/bias that are often present in tests.
Chance should be washed out by large numbers of trials. Bias is certainly an ongoing concern in psi experiments.

~~ Paul