JREF discusses 182. Andrew Paquette Brings Statistical Rigor to Psi Experiences

Andrew Paquette

Administrator
Or to be more exact, they discuss the paper the episode also talks about. The link to their pages is here. As I wrote on the old forum, my account was not approved over there, making it impossible to respond. For that reason, I responded on Skeptiko, which at least one person at JREF noticed, though without replying to any of the answers I gave. I have tried registering again today. We'll see if the account is approved this time. In the meanwhile, here are some quotes from their discussion, almost all of which seem more straightforward than anything that came up on the Skeptiko forum in the form of criticism. I do think that every point is incorrect, but appreciate the clear descriptions of their complaints.

Here is a link to the paper.

  1. Would'nt it be more likely that since we demonstrably dream about things that happen day to day, that some of these daily events would in various ways mirror events that occur normally in the course of events (aircraft crash pretty regularly....) and that we are merely conflating them with precognition by cherry-picking the data?
  2. I would like to know if he had precognition dreams how he will explain this to me:

    1. We can manipulate dreams: http://sciencefocus.com/blog/scienti...te-dreams-rats
    2. We can record dreams: http://www.mnn.com/green-tech/resear...-and-play-them
    3. Dreams are only for storing our memories that we acquired through the day:http://www.sciencedaily.com/releases...0225132249.htm
    4. Even animals have dreams.
    5. You can loose your ability to dream thanks to brain damage..
  3. I read his paper. Some of his "anecdotes" are after the fact.. Like Arouet wrote in the mind-energy forum.
  4. In order to determine whether dreams predict future events more accurately than would be expected by chance it is first necessary to determine what that chance accuracy rate is. It's not clear to me from the abstract whether Paqart has done this,
  5. I also found most of the anecdotal evidence and it is right. He is using people who he knows. Therefore it is a weak study..
  6. I agree with Pixel42's assessment. As so often, the veridicality of the experiences described in the anecdotes depends on a subjective judgement by the experiencer.
  7. Both papers clearly use the same data (his dream journals), so my assessment applies to both. Without some way of establishing the expected chance accuracy it is not possible to establish that the actual accuracy exceeds it. Every time this subject has come up I have wracked my brain to come up with a suitable protocol, without success. Early in the paper Parqat seems to acknowledge this difficulty and claims to have come up with a strategy that addresses it. Maybe I'm missing something, but I can't see how.
  8. The two issues I raised (the need to establish a baseline of expected chance accuracy to which actual accuracy can be compared, and define in advance for each potentially precognitive dream precisely what would need to occur in reality for the dream to be considered precognitive and then count both the hits and the misses) should be addressed in the paper. But I can't make head or tail of about half of it, so as I said I may be missing something.
  9. Also the author of the paper is not a a statistician. He is a writer and artists:

    http://en.wikipedia.org/wiki/Andrew_Paquette

    Even when he claims in the mind-energy forum that he was looking into statistics to create his own methodology. I know it doesn't have a impact on the paper but I just wanted to post what I found out about him and why I think that its flawed..
  10. If I had dreams which I suspected might be precognitive I would do the following for, ideally, several years:

    1. For every dream I had which might be precognitive (i.e. that didn't include flying or the other weird stuff that only happens in dreams) I would write it down in as much detail as I could remember.

    2. I would then assign a score to possible corresponding future events. For example if I dreamt that someone I knew would be involved in a car accident and escape with minor injuries, although the car ended up with its top caved in, I might assign points as follows:

    1 point if the person in the dream had an accident

    1 point if someone else I knew had a car accident

    2 points if the person in the dream had a car accident

    3 points if the person in the dream had a car accident and the car's damage matched the description I'd written down

    3 points if the person in the dream had a car accident and their injuries matched the description I'd written down

    4 points if the person in the dream had a car accident, and both the car's damage and their injuries matched the description I'd written down

    0 points if nothing that corresponded to the dream had occurred by the end of the study period

    3. At the end of the study period I would tot up the number of points I'd awarded and divide it by the maximum possible number of points to calculate the percentage accuracy.

    4. I would then compare the actual accuracy with the accuracy that would be expected by chance to see if it exceeded it by a statistically significant margin, which I would have set as the success criteria before beginning the study.

    Point 4 is, of course, the problem. What percentage accuracy would be expected by chance? Clearly the answer isn't zero: coincidences happen all the time, and our pattern seeking brains tend to vastly underestimate the frequency with which they occur. I can see no way of reliably estimating the chance accuracy in order to have a baseline with which to compare my final score.


    Workable protocols which have been discussed on this board before have only been possible if the possibly precognitive dreams have been fairly specific. For example if someone claimed to dream about earthquakes before they happened then you could get a seismologist to make predictions based on the best available information and see if the dreamer's accuracy was consistently better. Likewise if the dreamer regularly saw, say, the front page of tomorrow's paper you could come up with a protocol where you got them to select from a bunch of photos the one that most resembled the one they saw and see if they picked out the right one significantly more often than would be expected by chance. But when you have no idea what your precognitive dreams are going to be until you've not only dreamt them but something which resembles them has happened ... how on earth do you test that?

    Paqart's paper talks about identifying an anchor in each dream and does some statistics on their probability. I'm not qualified to assess the validity of the statistics, but my more fundamental problem is that I just don't see how these anchors address the difficulty I'm describing.
  11. My opinion hasn't changed. He starts out to find evidence to support his conclusion (at the outset he says that around 10% of recorded dreams are of "OBE, precognition, after-death communication, past life memories, psi training, healing, auras, telepathy, psychokinesis, and less easily defined categories of a spiritual or religious nature").
  12. His verification example involves telling someone he dreamed about them, reading them the dream he'd written, and asking them if any of it fitted what they were doing recently. A double layer of interpretation with priming.
  13. I didn't see where he analysed hits vs misses, nor was it clear how he decided a dream was psi and worthy of verification at all. How do you tell which of the paranormal types on the list above a dream is? suppose it's a mixture of types?
  14. I skipped the pages of messing with statistics because they were irrelevant to the gaping holes elsewhere in his approach.
  15. There were other problems with the presentation - such as describing veridical details he dreamed about but didn't write down; a standard ploy to bolster dubious evidence.
  16. Just a quick reminder that the end of the study period must be clearly defined at the beginning of the study period. Obviously, you understand that but sometimes the folks doing the research do not see the importance of that step.
  17. (continued next post)
 
  1. If I am in a seven-year study and on the first night, I dream that there will be a plane crash on the Eastern Seaboard, then do I earn points for any crash over the next seven years?

    17. I am also having some problems with this. Some of the events he predicted as precognition were 11 years in the future not recent ones and he lived in that city to be precise. It was about the WTC:

    18. This is a problem. He lived in Manhattan. However it is a long stretch and according to the text in the pdf. He awaited it in a near future not a distant one because he moved away out of fear. Also after reading it again. He added information which were not in the original dream - My long dream of May 17, 1990, from this group referenced the 1989 dream, inspiring me to type
    an account of the earlier dream for my records at that time. - He added one dream after he had the second dream into the journal. This is a error in my point of view because I hardly believe that he remembered a dream and that his memories are bullet proof and he can remember everything clearly after a year.

    19. Yes, but the longer the study goes on the more likely it is that such a prediction would be true by chance. The number of hits required to meet the success criteria would be set to be significantly greater than that chance success rate. Which we currently have no way of reliably estimating.

    20. Take all the dream predictions and see if they match up to reality when randomized. So, for example, a dream that predicted a car crash for dreamer X next week happens to be paired with dreamer Y four weeks from now.

    Mix up all the dreams and dreamers and then see what random dreams predict.

    If this procedure would not work, it should make us suspect that dreams are too predictive - like horoscopes that seem to apply to everyone at any time. If the dreams cannot be randomized in this way it means they are not specific enough. In other words, a predictive dream that cannot be wrong, cannot be right either.

    21. And if I understand your suggestion correctly, then we have to add the possibility that the folks on the psychic-side of the issue might then assert backwards causality and say that the universe knew that dreamer X's dream was going to be paired with dreamer Y and therefore the universe gave dreamer X a dream that does count as a hit when moved to dreamer Y's column. I.e. dreamer X "predicted" dreamer Y's car crash. The universe being just that damned clever.

    22. I have those dreams too, but there is always the temptation to piece together some kind of narrative or extract some meaning from fragments of them. This is a problem with dream interpretation; there will be dreams that are only a bit weird and could be interpreted in various ways...

    23. The problem with this is that the dreamer generally knows at least some context of the lives of the people he/she is dreaming about. So, for example, dreams about a horrific car crash might be expected if the dreamer was previously scared by the driving of the person being dreamed about (the 'dreamed'), or knows that the dreamed has a fast car, or drives badly, or has points on their licence, etc.

    I think this contextual knowledge probably plays a bigger part in apparently predictive dreams than the dreamers think.

    24. What's more troubling to me is how to measure significance. If I have a dream in which I awake in bed, and son of a gun, I do indeed awake in bed, that would be predictive. So there has to be a "surprisingly so" element.

    25. Pondering on marplots' suggestion the possibility occurs to me of simply pairing each dreamer with someone similar who would be given the dreamer's descriptions and scoring instructions and asked to score events in their own lives which happened to correspond, thereby acting as the control. There are obvious problems with this as well, of course, but they might be surmountable with care.

    26. A wiki entry he himself wrote BTW.

These are all the questions, criticisms, and suggestions I could find in the thread as of today. I intend to address them during my next PhD break in a couple hours. Until then, perhaps someone else would like to take a stab at either increasing the list of these critical comments or answering them?

I will answer the last item now, because no one else would know this. That wikipedia article really annoys me. Years ago, a friend told me about Wikipedia for the first time. He suggested I write an article with a bio of myself so that I could advertise my website. About two days later, I saw it had been flagged because it was me that wrote it. Remember, we're talking about 2005 or so. I hadn't realized there was a rule discouraging that kind of thing, but after finding out, I deleted the page. A week later, it was back. I tried deleting it several more times, but it kept coming back. Then I discovered the talk pages and discovered they thought it was a legit article because of my involvement in film, TV, video games, and comic books, so they kept on restoring the page. I posted a disclaimer on the talk page that I had not intended to violate Wikipedia rules and would have preferred the page removed or completely rewritten by someone else. I don't know if that disclaimer is still there.
 
Last edited:
Or to be more exact, they discuss the paper the episode also talks about. The link to their pages is here. As I wrote on the old forum, my account was not approved over there, making it impossible to respond. For that reason, I responded on Skeptiko, which at least one person at JREF noticed, though without replying to any of the answers I gave. I have tried registering again today. We'll see if the account is approved this time. In the meanwhile, here are some quotes from their discussion, almost all of which seem more straightforward than anything that came up on the Skeptiko forum in the form of criticism. I do think that every point is incorrect, but appreciate the clear descriptions of their complaints.

Here is a link to the paper.

The link takes me to my google drive,
 
I'll have to reply to this tomorrow. I spent too much time replying to something in the Stevenson thread and now it is too late to do this.

AP
 
LOL! Number 2 cracks me up.

1. We can manipulate dreams: http://sciencefocus.com/blog/scienti...te-dreams-rats
2. We can record dreams: http://www.mnn.com/green-tech/resear...-and-play-them
3. Dreams are only for storing our memories that we acquired through the day:http://www.sciencedaily.com/releases...0225132249.htm
4. Even animals have dreams.
5. You can loose your ability to dream thanks to brain damage..

Yet he fails to explain why any of those being true precludes precongition in dreams. Maybe number 3, but I think that is highly exaggerated.
 
Would'nt it be more likely that since we demonstrably dream about things that happen day to day, that some of these daily events would in various ways mirror events that occur normally in the course of events (aircraft crash pretty regularly....) and that we are merely conflating them with precognition by cherry-picking the data?
This doesn't address the actual data described in the paper. There is no "cherry-picking" because:
No items are left out. It is a continuous sequence of dream items, almost 600, regardless of apparent veridicality.
The anchors link the dreams to specific people within a certain narrow time frame for OBE examples. This limits considerably the potential of regular events coinciding with the line items, particularly when they are unique or highly unusual.
Precognition examples are more difficult to tie down because they are only recognized as resembling a dream after they have been recognized. OBEs are recognized as OBEs with a specific individual at a specific time frame in advance. To counter this, the probability tables I used factor in the number of days between dream and the correlative event to multiply the probability it is chance, as well as multiplying the probability by the size of the potential affected population. These adverse variables should adequately compensate for this criticism.
I would like to know if he had precognition dreams how he will explain this to me:

  1. 1. We can manipulate dreams: http://sciencefocus.com/blog/scienti...te-dreams-rats
Without commenting on the quality of the study, it can be said that even in humans dream content can to an extent be manipulated, though I wouldn't use the term "controlled." Anyone who has found themselves looking for toilets in a dream, then awakened to find an urgent need to find a toilet should know that drinking fluids before bed can affect dream content. This is no more relevant to the psi value of a dream than Tiger Woods' ex-wife's dating habits is to his ability to play professional golf. It may affect his score, but not the fact that he plays at a very high level regardless.
  1. 2.
I seriously doubt this is true. This claim is a bit extravagant considering the paper used as justification. Even if it were possible, which I really doubt, how would it be relevant to psi occurring in dreams? There is a similar claim that an electronic device implanted in the brain can restore vision to the blind. What I saw of this did not match the claim. Rather than creating an image, it stimulates the sensation of unresolved light.
  1. 3.
    Dreams are only for storing our memories that we acquired through the day:http://www.sciencedaily.com/releases...0225132249.htm
The day residue theory amounts to speculation. In my own journals I have looked for examples of it and note occasions when it looks possible or probable. It does occur, but is so rare that I do not consider this seriously as an explanation for dreams in general, let alone psi dreams that involve content previously unknown.
  1. 4.
    Even animals have dreams.
What is the point of this? So what? Does that mean we don't? Or that our dreams can't be psi because a dog's dreams can't be psi? Why can't an animal have psi dreams? This makes no sense as a criticism.
  1. 5.
    You can loose your ability to dream thanks to brain damage..
At best one could claim that a person loses the ability to remember a dream and that there is no measurable REM activity, not that they don't dream. There isn't enough known about consciousness to justify this claim.
I read his paper. Some of his "anecdotes" are after the fact.. Like Arouet wrote in the mind-energy forum.
There is one "after-the-fact" record of a dream mentioned, but that dream still preceded the event it is connected to by a period of years. Therefore, it is before-the-fact of the event it described, though the dream was written about a year after the dream. This characterization is inaccurate for quantity ("some" as opposed to "one") and meaning ("after-the-fact" where "fact" refers to the dream, not the future event.).
In order to determine whether dreams predict future events more accurately than would be expected by chance it is first necessary to determine what that chance accuracy rate is. It's not clear to me from the abstract whether Paqart has done this,
In my paper I explain why I think it is totally impossible to calculate something like this. There are simply too many variables. What I do instead is to use probability values that are extremely adverse to the psi hypothesis and clearly wrong because they are so adverse. For instance, I am talking highly unlikely events, such as a friend of mine talking with someone who has had a tree branch fall on and destroy two cars in the same week in two separate incidents. This means that when I call the friend to verify it, this has to connect to him within 24 hours of the event. He verifies it for that morning (about simultaneous as far as we can tell). Despite this, I give the events a probability of .25 happening every day. Cars don't get smashed this way every day, and my friend doesn't have conversations every four days with someone who has had this happen. Despite this, I give it this adverse probability and the resulting p-value still comes out against chance because of the number of connected hits.
I also found most of the anecdotal evidence and it is right. He is using people who he knows. Therefore it is a weak study..
I can't help who I dream about, but this really isn't relevant unless you are assuming fraud of some kind. It isn't like I know anyone well enough to predict that the one and only time bee sting therapy for paralysis is important to them, it will occur on the same night I dream about it for the first and only time out of a 24 year period.
I agree with Pixel42's assessment. As so often, the veridicality of the experiences described in the anecdotes depends on a subjective judgement by the experiencer.
Can't agree with this. We aren't talking about things where there is a wide variation between description and event. This claim is akin to saying that "a blue hat" is not "a navy blue hat."
Both papers clearly use the same data (his dream journals), so my assessment applies to both. Without some way of establishing the expected chance accuracy it is not possible to establish that the actual accuracy exceeds it. Every time this subject has come up I have wracked my brain to come up with a suitable protocol, without success. Early in the paper Parqat seems to acknowledge this difficulty and claims to have come up with a strategy that addresses it. Maybe I'm missing something, but I can't see how.
Apparently you have missed the whole point of the article, which is to use adverse scoring to overcome the barrier you mention. Also, see the item about adverse probability scoring above for more detail.
AP
 
  1. The two issues I raised (the need to establish a baseline of expected chance accuracy to which actual accuracy can be compared, and define in advance for each potentially precognitive dream precisely what would need to occur in reality for the dream to be considered precognitive and then count both the hits and the misses) should be addressed in the paper. But I can't make head or tail of about half of it, so as I said I may be missing something.
Yes, you are missing something. If you take the time to read the methodology section of the paper, you will be better informed.
  1. Also the author of the paper is not a a statistician. He is a writer and artists:

    http://en.wikipedia.org/wiki/Andrew_Paquette

    Even when he claims in the mind-energy forum that he was looking into statistics to create his own methodology. I know it doesn't have a impact on the paper but I just wanted to post what I found out about him and why I think that its flawed..
If what I write is correct, it really doesn't matter if I am an artist, statistician, or circus elephant. In this case, I had the help of someone from the University of Edinburgh on the statistics questions, and it also went through peer review. I had more questions about the statistics than anything else, and had to rewrite that section (and redo the calculations) a couple of times before the blinded review committee accepted the paper.
  1. If I had dreams which I suspected might be precognitive I would do the following for, ideally, several years:

    1. For every dream I had which might be precognitive (i.e. that didn't include flying or the other weird stuff that only happens in dreams) I would write it down in as much detail as I could remember.
To save the trouble of thinking about such questions, I write down every dream in as much detail as I can remember and then study them later. After all, how am I to know in advance what a dream will prove to be?
2. I would then assign a score to possible corresponding future events. For example if I dreamt that someone I knew would be involved in a car accident and escape with minor injuries, although the car ended up with its top caved in, I might assign points as follows:
This is done in my paper, but as a quick comparison reveals, you aren't compounding the probability of these things happening together as you should.
1 point if the person in the dream had an accident

1 point if someone else I knew had a car accident

2 points if the person in the dream had a car accident

3 points if the person in the dream had a car accident and the car's damage matched the description I'd written down

3 points if the person in the dream had a car accident and their injuries matched the description I'd written down

4 points if the person in the dream had a car accident, and both the car's damage and their injuries matched the description I'd written down

0 points if nothing that corresponded to the dream had occurred by the end of the study period

3. At the end of the study period I would tot up the number of points I'd awarded and divide it by the maximum possible number of points to calculate the percentage accuracy.
This makes sense if you don't mind using arbitrary values and ignore the combined effect of multiple line items matching the same anchor. If you do not want to ignore the compounding effect, you'll get something along the lines of what I wrote in my paper.
4. I would then compare the actual accuracy with the accuracy that would be expected by chance to see if it exceeded it by a statistically significant margin, which I would have set as the success criteria before beginning the study.

Point 4 is, of course, the problem. What percentage accuracy would be expected by chance? Clearly the answer isn't zero: coincidences happen all the time, and our pattern seeking brains tend to vastly underestimate the frequency with which they occur. I can see no way of reliably estimating the chance accuracy in order to have a baseline with which to compare my final score.
This is impossible to know accurately, as you suggest. You assume too much in your comment about "pattern-matching brain" however. One thing about my journals that may be uniquely useful in studies of this kind is that I know exactly how many times certain people show up in the journal, or events, places, colors, etc. This means I can calculate probability on that. I am preparing a new study that does this, but it won't be ready for probably six months.
Workable protocols which have been discussed on this board before have only been possible if the possibly precognitive dreams have been fairly specific. For example if someone claimed to dream about earthquakes before they happened then you could get a seismologist to make predictions based on the best available information and see if the dreamer's accuracy was consistently better. Likewise if the dreamer regularly saw, say, the front page of tomorrow's paper you could come up with a protocol where you got them to select from a bunch of photos the one that most resembled the one they saw and see if they picked out the right one significantly more often than would be expected by chance. But when you have no idea what your precognitive dreams are going to be until you've not only dreamt them but something which resembles them has happened ... how on earth do you test that?
Keep in mind that by focusing on precognition, you are leaving out the many veridical OBE examples. These are much easier to pin down and are much less flexible in the way they are scored than precognition examples, which are comparatively rarer.
Paqart's paper talks about identifying an anchor in each dream and does some statistics on their probability. I'm not qualified to assess the validity of the statistics, but my more fundamental problem is that I just don't see how these anchors address the difficulty I'm describing.
Take another look at the paper then, maybe that will help.
  1. My opinion hasn't changed. He starts out to find evidence to support his conclusion (at the outset he says that around 10% of recorded dreams are of "OBE, precognition, after-death communication, past life memories, psi training, healing, auras, telepathy, psychokinesis, and less easily defined categories of a spiritual or religious nature").
Well, no, not really. As I wrote, I started the journals to prove to someone that I was NOT having precognitive dreams. However, I soon realized it appeared that I was. Regardless, it took some time to be convinced of this. Again, the original purpose of the journal was to prove that precognition was NOT happening. This paper is simply a record of a part of that process and a way to demonstrate how the data could be looked at through a statistical filter.
  1. His verification example involves telling someone he dreamed about them, reading them the dream he'd written, and asking them if any of it fitted what they were doing recently. A double layer of interpretation with priming.
Well, not exactly. In many, like some in the second paper some of you accidentally read, I do not tell the person I am telling them something I expect they will find familiar. High energy physicist Richard Breedon, of CERN, sent me a letter confirming this in an incident involving him, where he was quite surprised to hear me describing something he had just been doing in Tokyo while I was sleeping in New jersey. Others are told it is a dream, though "reading the dream" and "telling about the dream" are one process, not two. The exact question I liked to use most often was "does any of this sound familiar to you for a recent event in the last 24 hours?"
  1. I didn't see where he analysed hits vs misses, nor was it clear how he decided a dream was psi and worthy of verification at all. How do you tell which of the paranormal types on the list above a dream is? suppose it's a mixture of types?
I didn't "decide" something was "psi and worthy of verification". If I dreamed of someone, I told them about it whether or not I thought it could be psi. As for precognition, if I saw something that reminded me of a dream, I would then seek out the dream the scene reminded me of and was typically rewarded with many details I had forgotten that were also relevant to other things from the same day.
  1. I skipped the pages of messing with statistics because they were irrelevant to the gaping holes elsewhere in his approach.
Hard to complain about this because of the "gaping hole" in this criticism.
  1. There were other problems with the presentation - such as describing veridical details he dreamed about but didn't write down; a standard ploy to bolster dubious evidence.
This most likely refers to the mention of my very carefully worded description of remembering the sound of jet engines in one dream where there were many other veridical items. Keep in mind that although I mentioned this in the paper, I did not count it as a line item for evaluation purposes. It is a piece of information to better describe the events, but only information written down in advance was used for probability calculations. This was stated in the paper, you must have missed it.
  1. Just a quick reminder that the end of the study period must be clearly defined at the beginning of the study period. Obviously, you understand that but sometimes the folks doing the research do not see the importance of that step.
The study period, or the period reviewed, is defined in the paper. They are the first twenty sequential records of journal 3. That is pretty clear.

AP
 
  1. If I am in a seven-year study and on the first night, I dream that there will be a plane crash on the Eastern Seaboard, then do I earn points for any crash over the next seven years?
    No. If, however, there is a crash seven years later and that reminds you of the dream, you have an anchor. The anchor is not counted as veridical, so the plane crash itself doesn't count. That is because the plane crash is used to test all the rest of the content in the dream. When you go back and look at it, if the dream only mentions a plane crash, you have a veridicality score of zero. If on the other hand, there are 20 other items mentioned and nine of them match items connected to the plane crash, such as: airline, location of crash, time of day, weather, cause of crash, number of casualties, damage to ground structures, plane falls into the ocean, and unusual rescue method, then you score 9/20. That would just barely look interesting using the method I used in my study. This is just so that we compare apples to apples here.

    17. I am also having some problems with this. Some of the events he predicted as precognition were 11 years in the future not recent ones and he lived in that city to be precise. It was about the WTC:
    Quite true. Because of this, I diluted probability for all of these by multiplying the p-value by the number of days between dream and event. In most cases, the value was one so there was no effect. For this event, the value was in the thousands of days, but multiplying the p-value by that large number did not come close to reducing it to non-significance. I also multiplied the p-value by an exaggerated figure representing the population of the world (6 billion or so) and even that didn't reduce it to non-significance. It isn't as if I didn't think of these things myself, but I appreciate that you are trying here.

    18. This is a problem. He lived in Manhattan. However it is a long stretch and according to the text in the pdf. He awaited it in a near future not a distant one because he moved away out of fear. Also after reading it again. He added information which were not in the original dream - My long dream of May 17, 1990, from this group referenced the 1989 dream, inspiring me to type
    an account of the earlier dream for my records at that time. - He added one dream after he had the second dream into the journal. This is a error in my point of view because I hardly believe that he remembered a dream and that his memories are bullet proof and he can remember everything clearly after a year.
    This is quite irrelevant for two reasons. The first is that if you remove the dream that was written later, it does not meaningfully affect the p-value. The second is that it was still written long in advance of the later event, so it is still before-the-fact. If I did make a mistake, it would still be amazing that it contained veridical information, though that dream was so strong that I talked about it rather often and it remains clear in my mind. I doubt there were any memory errors, but again, you could discount it entirely without changing the result to a significant degree.

    19. Yes, but the longer the study goes on the more likely it is that such a prediction would be true by chance. The number of hits required to meet the success criteria would be set to be significantly greater than that chance success rate. Which we currently have no way of reliably estimating.
    You are focusing on the one long range example in the study. What do you think of the examples that have effectively no turnaround time? Also, this "more likely by chance" I have to say, is ridiculous. To the present day, I have checked, there is only one building collapse in NYC that matches the details of that dream. This is starting from the founding of the city until now. Also, I checked this angle for other cities and other countries. This is the only recorded example in the history of the world that matches that trio of dreams as well as the WTC collapse. Ask yourself, what is the chance that out of the relatively tiny number of worldwide building collapses throughout history, that my dream of one in NYC does not match any of the others as well as the WTC example in NYC?

    20. Take all the dream predictions and see if they match up to reality when randomized. So, for example, a dream that predicted a car crash for dreamer X next week happens to be paired with dreamer Y four weeks from now.

    Mix up all the dreams and dreamers and then see what random dreams predict.

    If this procedure would not work, it should make us suspect that dreams are too predictive - like horoscopes that seem to apply to everyone at any time. If the dreams cannot be randomized in this way it means they are not specific enough. In other words, a predictive dream that cannot be wrong, cannot be right either.
    The example above should be sufficient to show why this is unnecessary. When you have only one example in the history of the world, you know the chances are not so common that you could expect regular matches.

    21. And if I understand your suggestion correctly, then we have to add the possibility that the folks on the psychic-side of the issue might then assert backwards causality and say that the universe knew that dreamer X's dream was going to be paired with dreamer Y and therefore the universe gave dreamer X a dream that does count as a hit when moved to dreamer Y's column. I.e. dreamer X "predicted" dreamer Y's car crash. The universe being just that damned clever.
    This doesn't make sense to me. I didn't do that in the paper though, so it is not relevant to the discussion.

    22. I have those dreams too, but there is always the temptation to piece together some kind of narrative or extract some meaning from fragments of them. This is a problem with dream interpretation; there will be dreams that are only a bit weird and could be interpreted in various ways...
    There is no interpretation going on here. There are very simple questions: "Were you talking to a guy in the last 24 hours who mentioned having two cars destroyed in the same week by falling tree branches?" The answer is yes or no. Where is the interpretation? The question isn't "Were you feeling troubled this morning but couldn't tell why, then did you feel better?" That latter question is pretty vague and useless, but the actual questions and the actual records are pretty clear.
 
23. The problem with this is that the dreamer generally knows at least some context of the lives of the people he/she is dreaming about. So, for example, dreams about a horrific car crash might be expected if the dreamer was previously scared by the driving of the person being dreamed about (the 'dreamed'), or knows that the dreamed has a fast car, or drives badly, or has points on their licence, etc.

I think this contextual knowledge probably plays a bigger part in apparently predictive dreams than the dreamers think.
This is less true than you think, in two ways. First, to take the car example of David Ryback, we're talking about someone I had met in person once and had spoken to on the telephone maybe five times. So no, I did not know his life or routine at all. Plus, he lived in a city I had only been to once (and never visited again) so I was also unfamiliar with his environment. In addition to that, the subject of the dream was unique in his life and concerned someone else. Therefore, no amount of familiarity with his life could have predicted something in someone else's life and that person was a stranger to me (and still is.)
24. What's more troubling to me is how to measure significance. If I have a dream in which I awake in bed, and son of a gun, I do indeed awake in bed, that would be predictive. So there has to be a "surprisingly so" element.
Read the examples in the paper, you might find that the waking up in bed example you give is real serious mischaracterization of the data.

25. Pondering on marplots' suggestion the possibility occurs to me of simply pairing each dreamer with someone similar who would be given the dreamer's descriptions and scoring instructions and asked to score events in their own lives which happened to correspond, thereby acting as the control. There are obvious problems with this as well, of course, but they might be surmountable with care.
This could be done and someday I may have time to do it. As a suggestion, I think it would be interesting to see the result, but again, it is not directly pertinent to this paper.
26. A wiki entry he himself wrote BTW.
It was originally written in about 2005, at the suggestion of a friend of mine. When I saw that it had been flagged a week later as having been written by someone with a close tie to the subject (myself) I realized that wasn't how Wikipedia worked, so I deleted the page. It was restored by a Wikipedia editor. I deleted it again, and again it was restored. I then found the talk feature and asked that the page either be deleted or they leave my note in the talk page that I did not want to leave the page as long as it violated some Wikipedia policy. Here is what I wrote, quoted from the Wikipedia talk page:
I see that this entry has a flag complaining that it is written by someone with a close connection with the subject, myself. This is true, but I wasn't aware that was a breach of etiquette at the time I posted it. After seeing this flag, I tried to delete the page, but a moderator restored the page, saying it was a valid page. What I'd like to do is see one of three things happen, delete the page, have it re-written by someone else, or remove the warning. At the very least, please leave this comment in the talk file until one of these things is done.

Thank you, Paqart (talk) 22:10, 13 February 2010 (UTC)paqart
 
Over at JREF they are really piling on the replies. However, despite the initial encouragement that posters should read the paper, I get the feeling that most haven't, or only skimmed it if they did. This guy "dlorde" (nice modest username) had an interesting and amusing idea:
From me:
There is no interpretation going on here. There are very simple questions: "Were you talking to a guy in the last 24 hours who mentioned having two cars destroyed in the same week by falling tree branches?" The answer is yes or no. Where is the interpretation?
Dlorde's answer:
The interpretation is of the recalled dream content, before you even write it down, before you talk to anyone.

AP
 
While not a baseline for dream veridicality, Robertson-Roy offer a baseline for mediumship veridicality in their mediumship studies, in the absence of psi. They take all the readings in which psi is absent (i.e. non-recipients are checking for line-item accuracy) and measure the proportion of subjects giving accuracy ratings of 0, 2.5, 5, ... up to 100%. Andy's results for 93 records aren't reported in such finely-grained detail, but it can be gleaned from his results that the accuracy falls into one of three categories - 0, 0+ to 67%, and >67%. The 0 to 67% is where we lack detail! but it looks like the average accuracy for that group is about 25%.

The proportion of subjects in those same categories from the Robertson-Roy study are 0.8, 0.18, and 0.02. That is, in the absence of psi, when we have 93 records we'd expect 74 records (0.8 x 93) to have 0% accuracy, 17 records to have an accuracy between 1.5 and 67%, and 2 records to have an accuracy greater than 67%. What we find from Andy's records is that 71 have 0% accuracy, 20 have an accuracy between 0+ and 67%, and 2 have an accuracy of >67%.

71 vs. 74
20 vs. 17
2 vs. 2

While Andy claims his dreams are unexpectedly accurate, they sure correspond pretty closely to what we'd expect in other studies in the absence of psi.

Linda
 
While not a baseline for dream veridicality, Robertson-Roy offer a baseline for mediumship veridicality in their mediumship studies, in the absence of psi. They take all the readings in which psi is absent (i.e. non-recipients are checking for line-item accuracy) and measure the proportion of subjects giving accuracy ratings of 0, 2.5, 5, ... up to 100%. Andy's results for 93 records aren't reported in such finely-grained detail, but it can be gleaned from his results that the accuracy falls into one of three categories - 0, 0+ to 67%, and >67%. The 0 to 67% is where we lack detail! but it looks like the average accuracy for that group is about 25%.

The proportion of subjects in those same categories from the Robertson-Roy study are 0.8, 0.18, and 0.02. That is, in the absence of psi, when we have 93 records we'd expect 74 records (0.8 x 93) to have 0% accuracy, 17 records to have an accuracy between 1.5 and 67%, and 2 records to have an accuracy greater than 67%. What we find from Andy's records is that 71 have 0% accuracy, 20 have an accuracy between 0+ and 67%, and 2 have an accuracy of >67%.

71 vs. 74
20 vs. 17
2 vs. 2

While Andy claims his dreams are unexpectedly accurate, they sure correspond pretty closely to what we'd expect in other studies in the absence of psi.

Linda

The problem with looking at it this way is that it doesn't take into account the cumulative probability that is the heart of the study. If veridical psi normally occurs in these quantities, then this test shows nothing. Personally, I think the cumulative values at the line item level are far more telling.

AP
 
What do you mean by this statement?

Linda
You are saying that by chance one would expect a certain number of correlations between dreams and physical events. If there is a causal or other non-physical connection at about the same rate, then your test would not be able to differentiate. By looking at the many connected veridical items, it is possible to distinguish from chance because chance values can be determined. What cannot be known is what the actual incidence of psi is.

AP
 
You are saying that by chance one would expect a certain number of correlations between dreams and physical events.

Not by chance - in the absence of psi.

If there is a causal or other non-physical connection at about the same rate, then your test would not be able to differentiate.

What? Aren't you the one claiming that a non-physical connection makes your dreams more accurate than could be expected in the absence of any causal or non-physical connection?

By looking at the many connected veridical items, it is possible to distinguish from chance because chance values can be determined.

Except that nobody cares about chance. Nobody thinks that our dreams are a random assortment of entirely independent elements.

Linda
 
Not by chance - in the absence of psi.
Fine, but that doesn't change anything. If real psi happens as often as you would expect correlations to occur when it is absent (approximately) then you wouldn't be able to use this to check it.

What? Aren't you the one claiming that a non-physical connection makes your dreams more accurate than could be expected in the absence of any causal or non-physical connection?

I hadn't thought of it that way, so I couldn't have made that claim. I don't have time to think about it now either though, so I'll have to get back to you on that.

Except that nobody cares about chance. Nobody thinks that our dreams are a random assortment of entirely independent elements.

Linda
Thanks for pointing this out. Again, it is like you are pretending to not realize that when I use the word "chance" I am referring to whatever factor it is that you would use to describe as "non-psi". And what is "psi"? I don't look at it as an ability or a thing, but a natural force that is always there. When we notice it, it is because it is stronger than usual or our resistance to noticing it is lower than usual. In other words, all things are permeated with psi at all times, to greater or lesser degrees.

AP
 
Fine, but that doesn't change anything. If real psi happens as often as you would expect correlations to occur when it is absent (approximately) then you wouldn't be able to use this to check it.

I hadn't thought of it that way, so I couldn't have made that claim. I don't have time to think about it now either though, so I'll have to get back to you on that.

Thanks for pointing this out. Again, it is like you are pretending to not realize that when I use the word "chance" I am referring to whatever factor it is that you would use to describe as "non-psi". And what is "psi"? I don't look at it as an ability or a thing, but a natural force that is always there. When we notice it, it is because it is stronger than usual or our resistance to noticing it is lower than usual. In other words, all things are permeated with psi at all times, to greater or lesser degrees.

AP

Sorry, I tend to use the specific meaning of words and I forget (i.e. not "pretend") that others may not be.

The idea is that information is gained, and with psi, that information is anomalous. That is, psi allows you to gain information above and beyond what is gained ordinarily. It doesn't make sense to model ordinary gain as "random sampling from a population", since nobody thinks that's how ordinary information comes to us. It doesn't matter if you meant "non-psi". You modelled it as "chance", when you should really be modelling it as "non-psi" (i.e. the way it was done in Robertson-Roy and various other research studies, with direct control groups).

It also doesn't make sense to claim that psi does not provide any information above and beyond that which is gained ordinarily. For one, this contradicts most psi claims, including yours, where people are amazed at how they knew something which otherwise seemed unknown to them (if relying on ordinary methods). And it also doesn't make sense to claim that there is something called "psi" if it doesn't do anything - that is, if you have the same amount of information when there is no psi as when there is.

"Psi", like "health", is something which is present in varying degrees and much of it is not under our control. "Non-psi" doesn't mean that this baseline psi is not present. It means the information we are working with cannot be anomalous. In the case of mediumship studies, it is readings applied to non-recipients.

If a reading about a man whose father is named "Robert", worked in a coal mine, likes to eat pistachios, and died in a car accident, is given for a man whose father is named "William", worked as a farmer, likes to eat pistachios, and died in an accidental fall, then "likes to eat pistachios" (plus whatever else is made of the connections between coal mining and farming (dirty), traditional names, and accidents) is regarded as the medium gaining anomalous information, but it may be regarded as a bit ho-hum. If later, a woman with a father named "Robert", who worked in a coal mine, who likes to eat pistachios, and who died in an accident involving a coal car happens to hear the reading, the connections aren't regarded as anomalous, but as non-anomalous or ordinary, since the reading wasn't for her. Unless you tell her that the reading was for her. Then it will be 'amazing'.

So whether a reading is amazing or ho-hum doesn't tell you whether or not the information is anomalous. Whether or not the reading was for someone tells you when to attempt to apply the 'anomalous' label. Since non-anomalous information can give you amazing or ho-hum readings/dreams, if you want to claim that anomalous information also gives you amazing readings/dreams, you don't want to show that you have the same number of amazing readings/dreams as you get from ordinary information. You want to show that you have more. But in order to do this, some sort of measure needs to be made of how many amazing readings/dreams you would get when the information isn't anomalous and make some sort of comparison to the readings/dreams you get when it could be.

You didn't make any sort of measure or comparison. All you did was guess that the number of amazing readings from non-anomalous information would be much lower than than what you got using 'anomalous' information. But your guess seems to be overly low. Even though you call it "averse" and claim that it is overly high, you don't back this up with anything but probability calculations which are invalid. If instead we look at how many amazing and ho-hum readings are produced by non-anomalous information under other circumstances (mediumship readings), we find that the number you produced was almost exactly the same. So there doesn't seem to be any room for anomalous information to have produced any extra amazing dreams.

Linda
 
The people over at JREF stopped discussing the paper seriously, though there were a number of fair questions at first. Now it's just a bunch of wandering trolls out there. It was too good to last, but for just a moment I thought they might have a couple of real skeptics. To badly paraphrase Jesus, "Why waste your time?"

AP
 
Back
Top