Chris Carter
Ray HymanIn the book Debating Psychic Experience, professional skeptic of parapsychology, professor Ray Hyman, wrote his lastest skeptical reflections on the current state of scientific parapsychology.
I will limit this post to one of Hyman's main objections and criticisms of parapsychology (which he considers the Aquiles' heel of this discipline): The lack of replicability of the evidence for psi.
It's well known that parapsychologists have argued that meta-analyses provide valid scientific replication of psi phenomena. Hyman directly disputes this view.
In his essays titled Parapsychology’s Achilles Heel: Persistent Inconsistency, Hyman wrote:
A meta-analysis is basically an exploratory rather than a confirmatory procedure. The notion of replication (or reproduction) of an experiment in the regular sciences is a prospective (or predictive) one. A successful replication is one that achieves essentially the same result that was predicted on the basis of a previously conducted experiment. The parapsychologists who try to justify the replicability of psi results with meta-analysis are using a retrospective notion. They are arguing for successful replication if a set of already completed experiments show evidence of similar effect sizes whose combined average is significantly different from chance. Replicability implies the ability to predict successfully from the results of a meta-analysis to a new set of independent data. This is where parapsychological evidence falls woefully short. (p.44. Emphasis in blue added).
So, we can construct Hyman's argument like this:
1-Scientific replicability is essentially prospective, i.e. it implies to predict the results of previous experiments to a new set of data which confirms the previous results.
2-Parapsychologists who use meta-analyses as justification of replicability are using a retrospective (not a predictive/prospective) notion.
3-Therefore, the positive results produced by such a meta-analyses done by parapsychologists don't support the claim of scientific replicability.
Hyman's argument is logically valid, so the question is to know if the premises of such argument are true. Premise 1 is correct, so we don't have anything to comment here.
But is premise 2 true? Answering this question is essential to the debate about parapsychology, because Hyman's whole case rests on the truth of this premise 2. If Hyman's premise 2 were true, then his position would be essentially correct, and any rational person would be forced to accept that the evidence for psi lacks replicability in the scientific sense. This would seriously undermine the parapsychology's scientific status.
ONE EXAMPLE USED BY HYMAN TO SUPPORT PREMISE 2:
In order to support premise 2, Hyman mentions the following example:
Consider the parapsychological claims that the autoganzfeld experiments successfully replicated the original ganzfeld database (Bem & Honorton, 1994). At least two parapsychologists now agree with my assertion that the autoganzfeld experiments failed to replicate the original ganzfeld data base (Bierman, 2001; Hyman, 1994; Kennedy, 2001). In the original database the average effect size was derived from studies that all used static targets. The autoganzfeld experiments used both static and dynamic (action video clips) targets. Only the dynamic targets produced a significant effect. The results on the static targets were consistent with chance and differed significantly from the results on the static targets in the original data base. (p.49. Emphasis in blue added)
Please, read carefully Hyman's example above. It's absolutely essential that we understand exactly what Hyman's objection is, and interpret it in its best, strongest formulation. No straw man is allowed in such an important question.
Hyman is arguing that the results of the original database of the ganzfeld experiments WERE NOT replicated in the autoganzfeld experiments because:
1-The original ganzfeld database consisted of studies which used only STATIC targets and produce positive results.
2-The autoganzfeld database consisted of both STATIC targets and DYNAMIC targets, but only the dynamic targets produced a significant effect.
3-The results of the autoganzfeld database on the static targets were consistent with chance
4-Therefore, the autoganzfeld experiments didn't produce positive effects in the case of STATIC targes amd hence they failed to replicate the original ganzfeld database (whose positives effects were precisely obtained with the static targets).
Therefore, Hyman's concludes, the autoganzfeld experiments don't provide evidence for replication of the original ganzfeld database.
Again, if Hyman's argument is correct, then his conclusion is valid and true, and any rational person must agree with him. But is Hyman right?
Chris Carter's reply to Ray Hyman:
In his reply to Hyman's argument, specially regarding the example mentioned above, Chris Carter wrote this:
The truth of the matter seems closer to the opposite of what Hyman tells us. The original ganzfeld experiments used quasi-dynamic targets (View Master “slide” reels) in addition to completely static targets. Studies using the View Master reels produced significantly higher hit rates than did studies using single-image targets (50% versus 34%). Meta-analysis of the original data led to the prediction that dynamic targets would show greater results than static targets. (p.158. Emphasis in blue added).
Note that Carter directly denies one of Hyman's factual premises, namely, the claim that the original ganzfeld database used only static targets.
According to Chris, the truth of the matter is that the original ganzfeld database used TWO kinds of targets (not one as Hyman misleadingly says), namely:
1-Quasi-dynamic targets (View Master "slide" reels)
2-Static targets
Keeping this fact in mind is absolutely essential, because it actually destroys Hyman's example.
Why? Because according to Hyman, the original ganzfeld database used only static target and hence it implies that any future replication of this database only will can be attained if future experiments produce positive results using static targets. And given that the autoganzfeld experiments produced results consistent with chance in the case of static targets, the original database wasn't replicated.
But Carter's mention that the original database used both (static AND quasi-dynamic) kinds of targets, leave open the possibility that at least regarding the dynamics targets, the autoganzfeld experiments produced positive results consistent with the original ganzfeld database and hence results which replicate the previous findings on dynamic targets.
And this is what actually happened. As Carter mentions: (regarding the non-static, dynamic targets) "Meta-analysis of the original data led to the prediction that dynamic targets would show greater results than static targets"
So, the autoganzfeld experiments didn't replicated the ganzfeld experiments regarding static targets, but they replicated the ganzfeld experiments regarding the DYNAMIC targets. and therefore, the parapsychologist's claim that the autoganzfeld experiments offer evidence of replication is demostrably true (contrary to Hyman's misleading position).
This point has not been understood by some readers of this book (even by readers sympathetic to the evidence for psi). For example, in Michael Prescott's blog, leading Brazilian psi researcher Vitor Moura posted the following comment regarding Carter's reply to Hyman: "But this clearly was not Hyman's criticism. What he said was that "the results on the static targets were consistent with chance and differed significantly from the results on the static targets in the original data base"
Dynamic versus static targets. The success of [these studies] raises the question of whether dynamic targets are, in general, more effective than static targets. This possibility was also suggested by earlier meta-analysis, which revealed that studies using multiple-image targets (View Master stereoscopic slide reels) obtained significantly higher hit rates than did studies using singleimage targets. By adding motion and sound, the video clips might be thought of as high-tech versions of the ViewMaster reels. The 10 autoganzfeld studies that randomly sampled from both dynamic and static target pools yielded 164 sessions with dynamic targets and 165 sessions with static targets. As predicted, sessions using dynamic targets yielded significantly more hits that did sessions using static targets (37 percent vs. 27 percent, p < .04). (p. 12)
As Hyman observed, “replicability implies the ability to predict successfully from the results of a meta-analysis to a new set of independent data.” And because of these results, virtually all ganzfeld studies ever since have used only dynamic targets. (p.158)
In conclusion, I agree with Carter and other parapsychologists that this evidence for psi has been scientifically replicated and therefore Hyman's objection that the evidence for psi is not replicable is demostrably fase.
I will limit this post to one of Hyman's main objections and criticisms of parapsychology (which he considers the Aquiles' heel of this discipline): The lack of replicability of the evidence for psi.
It's well known that parapsychologists have argued that meta-analyses provide valid scientific replication of psi phenomena. Hyman directly disputes this view.
In his essays titled Parapsychology’s Achilles Heel: Persistent Inconsistency, Hyman wrote:
A meta-analysis is basically an exploratory rather than a confirmatory procedure. The notion of replication (or reproduction) of an experiment in the regular sciences is a prospective (or predictive) one. A successful replication is one that achieves essentially the same result that was predicted on the basis of a previously conducted experiment. The parapsychologists who try to justify the replicability of psi results with meta-analysis are using a retrospective notion. They are arguing for successful replication if a set of already completed experiments show evidence of similar effect sizes whose combined average is significantly different from chance. Replicability implies the ability to predict successfully from the results of a meta-analysis to a new set of independent data. This is where parapsychological evidence falls woefully short. (p.44. Emphasis in blue added).
So, we can construct Hyman's argument like this:
1-Scientific replicability is essentially prospective, i.e. it implies to predict the results of previous experiments to a new set of data which confirms the previous results.
2-Parapsychologists who use meta-analyses as justification of replicability are using a retrospective (not a predictive/prospective) notion.
3-Therefore, the positive results produced by such a meta-analyses done by parapsychologists don't support the claim of scientific replicability.
Hyman's argument is logically valid, so the question is to know if the premises of such argument are true. Premise 1 is correct, so we don't have anything to comment here.
But is premise 2 true? Answering this question is essential to the debate about parapsychology, because Hyman's whole case rests on the truth of this premise 2. If Hyman's premise 2 were true, then his position would be essentially correct, and any rational person would be forced to accept that the evidence for psi lacks replicability in the scientific sense. This would seriously undermine the parapsychology's scientific status.
ONE EXAMPLE USED BY HYMAN TO SUPPORT PREMISE 2:
In order to support premise 2, Hyman mentions the following example:
Consider the parapsychological claims that the autoganzfeld experiments successfully replicated the original ganzfeld database (Bem & Honorton, 1994). At least two parapsychologists now agree with my assertion that the autoganzfeld experiments failed to replicate the original ganzfeld data base (Bierman, 2001; Hyman, 1994; Kennedy, 2001). In the original database the average effect size was derived from studies that all used static targets. The autoganzfeld experiments used both static and dynamic (action video clips) targets. Only the dynamic targets produced a significant effect. The results on the static targets were consistent with chance and differed significantly from the results on the static targets in the original data base. (p.49. Emphasis in blue added)
Please, read carefully Hyman's example above. It's absolutely essential that we understand exactly what Hyman's objection is, and interpret it in its best, strongest formulation. No straw man is allowed in such an important question.
Hyman is arguing that the results of the original database of the ganzfeld experiments WERE NOT replicated in the autoganzfeld experiments because:
1-The original ganzfeld database consisted of studies which used only STATIC targets and produce positive results.
2-The autoganzfeld database consisted of both STATIC targets and DYNAMIC targets, but only the dynamic targets produced a significant effect.
3-The results of the autoganzfeld database on the static targets were consistent with chance
4-Therefore, the autoganzfeld experiments didn't produce positive effects in the case of STATIC targes amd hence they failed to replicate the original ganzfeld database (whose positives effects were precisely obtained with the static targets).
Therefore, Hyman's concludes, the autoganzfeld experiments don't provide evidence for replication of the original ganzfeld database.
Again, if Hyman's argument is correct, then his conclusion is valid and true, and any rational person must agree with him. But is Hyman right?
Chris Carter's reply to Ray Hyman:
In his reply to Hyman's argument, specially regarding the example mentioned above, Chris Carter wrote this:
The truth of the matter seems closer to the opposite of what Hyman tells us. The original ganzfeld experiments used quasi-dynamic targets (View Master “slide” reels) in addition to completely static targets. Studies using the View Master reels produced significantly higher hit rates than did studies using single-image targets (50% versus 34%). Meta-analysis of the original data led to the prediction that dynamic targets would show greater results than static targets. (p.158. Emphasis in blue added).
Note that Carter directly denies one of Hyman's factual premises, namely, the claim that the original ganzfeld database used only static targets.
According to Chris, the truth of the matter is that the original ganzfeld database used TWO kinds of targets (not one as Hyman misleadingly says), namely:
1-Quasi-dynamic targets (View Master "slide" reels)
2-Static targets
Keeping this fact in mind is absolutely essential, because it actually destroys Hyman's example.
Why? Because according to Hyman, the original ganzfeld database used only static target and hence it implies that any future replication of this database only will can be attained if future experiments produce positive results using static targets. And given that the autoganzfeld experiments produced results consistent with chance in the case of static targets, the original database wasn't replicated.
But Carter's mention that the original database used both (static AND quasi-dynamic) kinds of targets, leave open the possibility that at least regarding the dynamics targets, the autoganzfeld experiments produced positive results consistent with the original ganzfeld database and hence results which replicate the previous findings on dynamic targets.
And this is what actually happened. As Carter mentions: (regarding the non-static, dynamic targets) "Meta-analysis of the original data led to the prediction that dynamic targets would show greater results than static targets"
So, the autoganzfeld experiments didn't replicated the ganzfeld experiments regarding static targets, but they replicated the ganzfeld experiments regarding the DYNAMIC targets. and therefore, the parapsychologist's claim that the autoganzfeld experiments offer evidence of replication is demostrably true (contrary to Hyman's misleading position).
This point has not been understood by some readers of this book (even by readers sympathetic to the evidence for psi). For example, in Michael Prescott's blog, leading Brazilian psi researcher Vitor Moura posted the following comment regarding Carter's reply to Hyman: "But this clearly was not Hyman's criticism. What he said was that "the results on the static targets were consistent with chance and differed significantly from the results on the static targets in the original data base"
Carter clearly missed the point here. Carter wrote about the dynamic targets, which Hyman admits to be significant. But the criticism of the static targets remains untouched."
Moura clearly missed Carter's (and Hyman's) point here. The debate is not about static targets or dynamic targets, but about the REPLICATON of the evidence for psi.
Certainly as Moura realizes, Hyman accepts that the results about dynamic targets were significant while denying that the original positive evidence of the static targets were replicated. But Hyman concludes from this fact that the autoganzfeld didn't offered any replication at all of the original database, which is false.Even agreeing with Hyman about the chance results of the static targets, the dynamic targets got positive results, and such positive results WERE PREDICTED by the original database, which is everything what Carter needs in order to refute Hyman's charge of lack of replication.
So, Moura's charge of irrelevance against Carter ("Carter wrote about the dynamic targets") is unjustified, because precisely the positive evidence in the case of dynamic targets replicates the prediction about dynamic targets (not about static targets in which Hyman misleadingly focused his criticisms) done on the grounds of the original database.
As Carter concluded this part of his reply to Hyman: "This prediction was in fact strongly corroborated, as Bem and Honorton (1994) reported:Dynamic versus static targets. The success of [these studies] raises the question of whether dynamic targets are, in general, more effective than static targets. This possibility was also suggested by earlier meta-analysis, which revealed that studies using multiple-image targets (View Master stereoscopic slide reels) obtained significantly higher hit rates than did studies using singleimage targets. By adding motion and sound, the video clips might be thought of as high-tech versions of the ViewMaster reels. The 10 autoganzfeld studies that randomly sampled from both dynamic and static target pools yielded 164 sessions with dynamic targets and 165 sessions with static targets. As predicted, sessions using dynamic targets yielded significantly more hits that did sessions using static targets (37 percent vs. 27 percent, p < .04). (p. 12)
As Hyman observed, “replicability implies the ability to predict successfully from the results of a meta-analysis to a new set of independent data.” And because of these results, virtually all ganzfeld studies ever since have used only dynamic targets. (p.158)
In conclusion, I agree with Carter and other parapsychologists that this evidence for psi has been scientifically replicated and therefore Hyman's objection that the evidence for psi is not replicable is demostrably fase.
0 comments:
Post a Comment