CIA-RDP96-00789R003200100001-5

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
Psychological Bulletin
1994, Vol. 115, No. 1. 19-24
Copyright 1994 by the American Psychological Association, Inc.
0033-2909/94/$3.00
Anomaly or Artifact? Comments on Bem and Honorton
Ray Hyman
Bem and Honorton imply that the 1 1 autoganzfeld experiments demonstrate the existence of psi-
a communications anomaly. They claim that the autoganzfeld results are consistent with previous
parapsychological findings and constitute evidence for a replicable psi effect. Although the auto-
ganzfeld experiments are methodologically superior to previous parapsychological experiments. the
tests of their randomization procedures were inadequate. The autoganzfeld experiments consistently
produced positive hit rates, whose combined effect was highly significant. However, these experi-
ments produced important inconsistencies with the previous ganzfeld experiments. They also
showed a unique pattern in the data that may reflect a systematic artifact. Because of these unique
features, we have to wait for independent replications of these experiments before we can conclude
that a replicable anomaly or psi has been demonstrated.
Bem and Honorton (1994) imply that if psychologists were
familiar with the most recent parapsychological research, they
would be more willing to accept the possibility that a commu-
nications anomaly existed. In particular, Bem and Honorton
focus on the experiments that are based on the ganzfeld proce-
dure. They "believe that the replication rates and effect sizes
achieved by one particular experimental method, the ganzfeld
procedure, are now sufficient to warrant bringing this body of
data to the attention of the wider psychological community"
(Bem & Honorton, 1994, p. 4). They review the debate between
Honorton and me over the original ganzfeld experiments. Hy-
man (1985) found that these studies suffered from statistical,
methodological, and documentation problems. Honorton
(1985) responded that these flaws were not sufficient to account
for the observed hit rates. Bem and Honorton (1994) review
this controversy and cite reviewers who apparently agree with
Honorton's position. The implication is that despite the defi-
ciencies in the ganzfeld experiments, the results support the ex-
istence of psi-a communications anomaly.
To Honorton's credit, he initiated a new series of experiments
that would be free from the flaws of the earlier ganzfeld database
(Honorton et al., 1990). These 11 new experiments, called the
autoganzfeld studies yielded consistently positive hit rates and a
highly significant overall effect. Because these new experiments
showed positive results and allegedly were consistent with the
earlier ganzfeld database and other psi research. Bem and Hon-
orton implied that parapsychology had found its previously elu-
sive repeatable experiment.
Since the beginnings of psychical research in the mid-nine-
teenth century, its investigators have believed that they have sci-
entific evidence sufficiently strong to place before the general
scientific community. Each generation has tried to get the atten-
tion of the scientific community with findings that they claim
to be irrefutable. The particular evidence put forth has changed
from generation to generation. What a previous generation of
parapsychologists considered to be a solid case for psi was aban-
doned by later generations in favor of a more current candidate.
This shifting database for parapsychology's best case may be
why parapsychology still has not achieved the recognition it de-
sires from the general scientific community.
Now Bem and Honorton (1994) believe that they have a
strong case to put before the psychological community. They
admit that the autoganzfeld findings still require independent
confirmation. To their credit, they specify the conditions and
the required sample size needed to provide adequate power. The
informed critic of parapsychology might ask what makes the
current situation different from the past claims for psi? Why
should we now believe that Honorton and his colleagues have
finally found a way to consistently produce evidence for psi?
We must wait for future attempts at replication before we
have an answer to the question. Bem and Honorton appear con-
fident that this time is different. Their review of the ganzfeld and
autoganzfeld databases encourages them to believe that consis-
tent psi results are within reach. In this commentary, I provide
reasons for believing that the autoganzfeld results contain in-
consistencies and some unique patterns that raise doubts about
their replicability.1
Agreements and Differences
Although my commentary focuses on my disagreements with
Bem and Honorton's (1994) presentation, I would like to briefly
specify some points of agreement. The autoganzfeld studies do
comply with most of the "stringent standards" (p. 353) spelled
out in the joint communiqué by Hyman and Honorton (1986).
I commend Honorton and his colleagues (1990) for creating a
protocol that eliminates most of the flaws that plagued the orig-
inal ganzfeld experiments. The 1 I autoganzfeld studies consis-
tently yield positive effects that, taken together, are highly sig-
nificant. I concur with Bem and Honorton's admission that
Correspondence concerning this article should be addressed to Ray
Hyman, Department of Psychology, University of Oregon. Eugene, Or-
egon 97403.
' Although I take a pessimistic position regarding future replications.
I think it is good that Bem and the parapsychologists are optimistic.
Such optimism should encourage investigators to attempt replications.
These replications will eventually decide the issue.
19
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
20
RAY HYMAN
"the autoganzfeld studies by themselves cannot satisfy the re-
quirement that replications be conducted by a 'broader range of
investigators' " (p. 13). I also support their suggestion that sev-
eral parapsychologists pool their resources and plan a large-
scale ganzfeld replication in which each laboratory contributes
a set of trials to the total pool.
So what is there to disagree about? I disagree with Bem and
Honorton about how strongly the autoganzfeld studies support
the hope for a replicable psi experiment. Where they see consis-
tency between the autoganzfeld studies and previous parapsy-
chological findings, I see inconsistency. Although I agree that
the autoganzfeld studies meet most of the stringent standards
that Honorton and I spelled out, I disagree that they meet all of
those standards. Our disagreements are a matter of degree. The
value of discussing our disagreements is to help clarify what
should constitute adequate evidence for the existence of an
anomaly: The existence or nonexistence of psi will not be settled
by debate. The existence issue will be settled by independent
attempts at replication-at least four of which are currently un-
derway (McCrone, 1993).
In explaining my disagreements, I point to weaknesses in the
autoganzfeld experiments. I want to emphasize that as a single
contribution to the ganzfeld database, these are commendable
experiments of high quality. But no single experiment or set of
studies can be perfect in all respects. When such a series is given
the responsibility of carrying a burden beyond its original
purposes, then various deficiencies will inevitably become ap-
parent. This is the case, I believe, with the autoganzfeld studies.
Internal Consistency Within the Autoganzfeld Studies
Bem and Honorton describe the autoganzfeld studies as 11
separate experiments conducted by eight different experiment-
ers. The hit rates are positive and consistent across the studies
and the experimenters. Although this is encouraging, the con-
sistency tells us little about potential replicability. Neither the
studies nor the experimenters are independent. The studies vary
in whether they use naive or experienced subjects. However, the
target set, the selection and judging procedures, the laboratory,
the setting, and the procedures are identical across studies and
experimenters. No experimenter is associated with a single
study, nor does an experimenter have independent input into
the design and procedure as happens in an independent replica-
tion. Indeed, the term experimenter in this context simply refers
to a person who plays an already scripted role. Any unique fea-
tures of the autoganzfeld procedure-including possible arti-
facts-would be the same for all 11 studies and the eight differ-
ent experimenters. Consequently the autoganzfeld studies
should be looked on as I large experiment rather than 11 sepa-
rate contributions.
Consistency With the Original Ganzfeld Database
Bem and Honorton claim that "[the autoganzfeld] results are
statistically significant and consistent with those in the earlier
database" (p. 13). They cite only two reasons to support this
claim. The overall effect size or hit rate is approximately the
same in the two databases. This apparent agreement in overall
effect size is meaningless. The overall effect size in the auto-
ganzfeld studies is a composite of two significantly different
effect sizes-that for the static targets and that for the dynamic
targets. The overall effect size in the ganzfeld data base is an
arbitrary composite of heterogeneous effect sizes, contributed
in unequal numbers, from different laboratories. The fact that
the two composites yield approximately the same effect size is
accidental. Both numbers could easily have been larger or
smaller, depending on the mix of the arbitrary sources from
which they were composed.
The dynamic targets yielded a significantly higher hit rate
than did the static targets in the autoganzfeld studies. Bem and
Honorton argued that this was consistent with the finding that
the multiple-image targets (View Master stereoscopic slide reels)
in the ganzfeld database yielded significantly higher hit rates
than did the single-image targets. I do not believe that multiple
static images on a View Master reel can be equated to the dy-
namic moving image on a videoclip. However, I will not argue
this point.
Clearly the dynamic targets outperform the static targets in
the autoganzfeld studies. Even if this is consistent with the ap-
parent superiority of the View Master targets over the single-
image targets, Bem and Honorton (1994) overlook a serious dis-
crepancy. Single-image targets constituted 76% of the 835 ses-
sions in the ganzfeld experiments. Their average hit rate was
.346. Given this effect size and the 166 trials using static targets,
the power or probability of replicating this effect in the auto-
ganzfeld experiments was .82. This failure to find a significant
effect with the static targets was even more notable given that
these experiments were conducted in "the warm social ambi-
ence" (p. 14) of Honorton's laboratory.
Bem and Honorton acknowledge that the autoganzfeld stud-
ies failed to replicate the predicted sender-receiver pairing
effect. In the original ganzfeld database, the trials on which the
receiver chose a friend as a sender produced a hit rate of .44
compared with a hit rate of only .26 for those trials on which
the experimenter assigned a sender. I would emphasize that
given this size effect with the 198 trials with friends as senders
and 128 with someone else as senders, the power of getting a
significant replication of the effect is over .92. Again, given the
'psi conducive' atmosphere of Honorton's laboratory, this fail-
ure to get significance is a noteworthy inconsistency.
On two key comparisons with the original ganzfeld database,
the autoganzfeld fails to replicate even with adequate power.
The positive hit rate and overall significance of the autoganzfeld
studies are due to an essentially new type of target, presented in
a new way. Even if we agree that there is a kinship between the
View Master reels of the ganzfeld experiments and the dynamic
targets of the autoganzfeld, we cannot ignore the differences be-
tween multiple images of a travel scene presented statically with
a slide projector and excerpts from motion pictures presented
with their accompanying audio on videocassettes. The prob-
lems of selecting, presenting, and controlling such targets pres-
ent new challenges. During the judging procedure in the origi-
nal ganzfeld experiments, the target and the decoys were dis-
played simultaneously. The judging procedure for the
autoganzfeld involves presenting the target and its decoys one
at a time. Because the positive hit rate and significance are due
to an essentially new type of target presented in a new way, the
need for independent replication is especially urgent.
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
1.
ANOMALY OR ARTIFACT
21
Consistency With Previous Parapsychological Findings
Bem and Honorton (1994) also claimed that "there are reli-
able relationships between successful psi performance and con-
ceptually relevant experimental and subject variables, relation-
ships that also replicate previous findings" (p. 13). They point
to three such "replications." One is a small, but statistically sig-
nificant, correlation of .18 between a measure of extroversion
and "psi performance." This is consistent with a tendency found
in previous psi studies. Second, they report the strong psi per-
formance of the Julliard students that they see as consistent with
psi studies that found a relation between psi abilities and cre-
ative and artistic abilities. This latter replication is not so im-
pressive when one considers that only 20 students were involved
and that their performance was not significantly different from
the other participants in the two studies in which they partici-
pated (Fisher's exact p = . 262, two-tailed). In addition, as I
point out below, the Julliard students were exposed to just those
conditions that favored high hit rates-targets that were re-
peated, a preponderance of dynamic targets, and active prompt-
ing by the experimenter during judging. Thus, it is unclear
whether their high hit rate was a function of their creativity or a
function of the special targets and conditions with which they
happened to be associated.
The third correlate could not be demonstrated for the auto-
ganzfeld studies. Bem and Honorton (1994) pointed out that
the subjects in the autoganzfeld tended to believe in psi, re-
ported psychic experiences, and had practiced meditation or
related techniques. These variables were previously reported as
correlates of psi. However, I do not see how they can claim that
these attributes of their subject population are a replication of
previous findings. They report no correlations between these
variables and performance in the autoganzfeld studies. Indeed.
they cannot report any correlation because they did not have
subjects who lacked these properties. We do not know if nonbe-
lievers and people without psychic experiences would have per-
formed better or worse than the actual subjects.
In other words, they can justify only one of the correlates that
they use to claim consistency with previous psi studies. Even
here the relationship is weak and is just one of many previously
reported correlates that might have been found. At one time,
for example, parapsychologists claimed that the decline effect
was a pervasive and characteristic property of psi. However,
when no decline effect is found in a parapsychological study, it
does not deter the experimenter from pointing to some other
significant departure from chance as evidence for psi. Note that
in the autoganzfeld studies, there is no decline effect.
Randomization and Claims of Psi
As I already stated, I agree that the autoganzfeld studies meet
most of the requirements that Honorton and I specified in our
joint communiqué (Hyman & Honorton, 1986). One surprising
exception is the inadequate testing of the randomization proce-
dures. The issue of randomization was central to the debate con-
cerning the original ganzfeld findings (Hyman, 1985). Adequate
randomization procedures are critical for parapsychological re-
search because the evidence for psi is based on a low probability
value for a departure from a chance baseline. Such probability
values have meaning within an idealized statistical model of the
experimental situation. Whether this statistical model applies to
a given situation is an empirical matter that must be adequately
justified if the stated significance levels are to be taken seriously.
Appropriate randomization procedures are one way to help en-
sure that the statistical model applies to the experimental data.
With respect to the autoganzfeld studies, this would entail se-
lecting the targets for each trial and ordering the target and de-
coys during judging in a demonstrably random manner. In ad-
dition, following the practice of a few past researchers, the para-
psychologist can also provide some post hoc analyses to show
that the distributions of targets and judging orders are consis-
tent with the underlying probability model.
Unfortunately, the autoganzfeld studies fell short on this crit-
ical requirement. The tests for adequacy of randomization were
confined to showing a uniform distribution of outputs from 1
to 160 for target selection and a uniform distribution of the per-
mutations of all possible orderings during the judging proce-
dure. Emitting a uniform distribution of target choices is a nec-
essary but hardly sufficient requirement for an adequate ran-
dom generator.
These randomization procedures are critical because we can
expect strong systematic biases during the judging procedure.
The fact that the items to be judged have to be presented se-
quentially, when combined with what we know about subjective
validation (Marks & Kammann, 1980), would lead us to expect
a strong tendency to select the first or second items during the
judging series. We would also expect strong response biases
within each target pool. Bem and Honorton show such a bias in
the target pool used for Study 302. Both these biases may be
strengthened by the fact that the experimenter interacts with
the receiver during the judging process. Although most receivers
participate in one session, each experimenter participates in
several. The response biases of the experimenters can play an
important role, especially in those studies in which the experi-
menter deliberately prompts the receiver to choose a particular
item during the judging. Such active prompting occurred in 6
of the 11 studies (Honorton et al., 1990).2
If the randomizing of the selection of targets and of the order-
ing of items during judging is adequate, such response biases
should not affect the validity of the statistical tests. One way to
prevent response biases from distorting the hit rate is to use a
randomizing procedure that makes sure that each item within
a target pool occurs equally often. The simple randomizing pro-
cedure used in the autoganzfeld studies would guarantee that
each target occurred an equal proportion (not number) of times
only in the very long run. In any finite number of trials, the
individual targets would occur with varying frequencies. Again,
if the randomization was adequate, this inequality of occur-
rence would not bias the hit rate. The items in some target pools
that occurred most frequently would be those that were favored
2 One referee suggested that I make it clear that I am not claiming
that sensory leakage occurred because of experimenter prompting. I
agree. The experimenter, according to the protocol, was ignorant of
which member of the target pool was the target during the judging pro-
cedure. The point is that by actively helping the subject to rate the mem-
bers of the target pool. the experimenter let his or her own subjective
biases enter the selection procedure.
:
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
22
RAY HYMAN
Table 1
Hit Rate as a Function of the Frequency of Occurrence of Targets
Frequency
Variable
1
2
3
4
5
6
7
8
Total
Hits
12
25
16
20
19
4
4
6
106
Misses
36
65
26
48
36
8
3
2
224
n
48
90
42
68
55
12
7
8
330
Hit rate
.250
.278
.381
.294
346
333
571
.750
.321
by the response bias. This would bias the hit rate upward. The
items in other target pools, however, that occurred most fre-
quently would be those that were avoided by the response bias.
This would bias the hit rate downward. With adequate random-
ization, these two tendencies would balance each other.
Achieving adequate randomization is not easy. Much can go
wrong-as some parapsychologists, among others, have shown.
This is why it is disappointing that the autoganzfeld studies did
not show the same concern for randomizing that they showed
for other aspects of the methodology. This is also why, in my
role of devil's advocate, I was interested in directly checking the
actual distribution of target positions among the decoys during
judging. Daryl Bem kindly agreed to supply me with this infor-
mation along with other data from the autoganzfeld database.
Unfortunately, the variable labeled position on the data sheet
turned out to be the original position of the target in its target
pool rather than its position during judging. This latter infor-
mation was unavailable to either Bem or me at the time of this
writing.
Hit Rate and Target Frequency
Because I could not directly check the adequacy of the ran-
domization procedures, I tried to find some indirect indicators.
If randomizing was inadequate and targets occurred with vary-
ing frequency, possible biases might show up as differential hit
rates for targets occurring with various frequencies. For exam-
ple, if targets favored by response biases were also favored by a
deficient target selection procedure, then we would find a posi-
tive correlation between hit rate and target frequency. It would
be possible, of course, for a deficient randomization procedure
to yield a negative correlation. To see if actual repetitions of
targets had any observable consequences, I tabulated the pro-
portion of hits as a function of how many times a target oc-
curred in this database. 3
As Table I shows, the relation between hit rate and target
frequency was strong. The test for a linear trend among the pro-
portions (Snedecor & Cochran, 1967, pp. 246-248) was positive
and significant, (z = 2.49, p = . 013, two-tailed). An indication
of the strength of this trend is given by the Spearman rank order
correlation between the hit rate and target frequency, which was
.83. Another way to look at this relationship would be to com-
pare the hit rate of targets that occurred once or twice (.27) with
those that appeared three or more times (.36).
This pattern exists separately for the static and dynamic
targets, although it is stronger among the dynamic targets. The
static targets that occurred once or twice had a hit rate of .22
compared with a hit rate of .31 for those that occurred more
than twice. The hit rate was .32 for those dynamic targets that
occurred once or twice as compared with a hit rate of .41 for
those that occurred three or more times.
Target Occurrence and Experimenter Prompting
What accounts for this peculiar relationship? Is the correla-
tion between target frequency and hit rate determined by which
particular targets get repeated? Or does replication itself some-
how increase the hit rate? If the relation is due to response bi-
ases, I would expect experimenter prompting to affect the later
occurrences of targets rather than their first occurrences. With
these questions in mind, I conducted a multinomial analysis of
variance (Woodward, Bonett, & Brecht, 1990). In this analysis,
hit rate was the dependent variable, and 3 two-level factors were
the independent variables: target type (static, dynamic), target
occurrence (first, later), and experimenter prompting (no, yes).
Of the interactions, only that between target occurrence and
experimenter prompting was significant, x2(1, N = 330) = 6.83,
p= . 009. The two significant main effects were target type, x2(1,
N = 330) = 4.76, p = . 030, and target occurrence, x2(1, N =
330) = 11.56, p < . 001.
The difference between the hit rate for dynamic targets (.356)
and that for static targets (.249) does not interact with the other
two factors and will be ignored for the present discussion.4 The
meaning of the interaction between target occurrence and
prompting can be seen in the simple effects of target occurrence
within each level of prompting. With no experimenter prompt-
ing, the effects of target occurrence were minimal: The hit rate
for first occurrences of targets was .291 and that for later occur-
3 To be consistent with Bem and Honorton, I treated the basic data-
base as the 330 sessions in Studies 1 through 301. Study 302, which
used a single target pool, was treated as a special case.
4 These hit rates are slightly different from those used by Honorton et
al. (1990) and Bem and Honorton (1994). This is because they com-
puted hit rates for any category by simply dividing the number of hits
by the total number in that category. The hit rate for dynamic targets
obtained with this approach is 61/164 = . 372 and that for static targets
is 45/166 = . 271. These rates are means weighted by the number of
cases in the cells for each combination of levels of the factors. For the
purposes of additivity of effects, I am using the unweighted means (each
cell of the design is weighted equally). This removes distortions and con-
founds that are due to unequal cell sizes. In the present case, the differ-
ences are small and inconsequential. I am supplying this footnote to
explain some discrepancies that might confuse the reader.
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5

23
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
ANOMALY OR ARTIFACT
rences was .334, x2(1, N = 181) = . 396, p = . 534. The effect of
target repetition combined with experimenter prompting, how-
ever, was very large. The hit rate for first occurrences of targets
with experimenter prompting was only .140. The hit rate for
later occurrences of targets when combined with experimenter
prompting jumped to .445. This gain was significant. x2(1, N =
149) = 14.702, p = . 0001. These results suggest that experi-
menter prompting depresses hit rates for first occurrences of
targets and enhances hit rates for subsequent occurrences of
targets.
Internal Checks on the Validity of This Pattern
Is this peculiar relation among hit rate, target frequency, and
experimenter prompting merely a fluke? I broke the data into
subsets in several ways to see if the pattern was consistent in the
different subcategories. I checked this pattern within the dy-
namic and static targets separately. I compared Targets 1 to 80
with Targets 81 to 160. Likewise, I looked for the pattern within
Studies 101 through 103 taken as a group as compared with
Studies 104 through 301 considered as a group. I also checked
for this pattern for each of the five experimenters who contrib-
uted the most sessions. Although the numbers became small in
some of these comparisons, the hit rate was consistently larger
for later as opposed to first occurrences of a target. I found just
one nonsignificant exception in the trials for one experimenter.
Likewise, wherever meaningful comparisons were possible, the
interaction between prompting and target occurrence oc-
curred.' For this database, then, the dependence of hit rate on
frequency of target occurrence and experimenter prompting
was a robust effect.
Implications
As far as I know, this dependence of hit rate on target occur-
rence and experimenter coaching has never been previously re-
ported in parapsychological research. One referee suggests that
the dependence of hit rate on target frequency and prompting
may reveal important moderator variables rather than artifacts.
The referee may be correct. The skeptic, however, might point
to the long history of alleged "moderator" variables in parapsy-
chology-such as the decline effect, displacement effects,
sheep-goat effects, and others. The problem is that when such
moderators are discovered in the data, they are put forth as im-
portant indicators or characteristics of psi. The absence of such
characteristics in subsequent data, however, does not deter para-
psychologists from claiming evidence for psi if they find a sig-
nificant hit rate. This is the troublesome problem of boundary
conditions. The parapsychologists have been unable to specify
what would constitute the absence of psi.
The positive effect for repeated occurrences of a target may
eventually turn out to be an important property of psi-if psi
exists. However, the fact that first occurrences of a target pro-
duce a hit rate consistent with chance raises questions. All of
the positive effect in the ganzfeld experiments rests on those
targets that have occurred more than once. The prompting
effect is even more curious. On first occurrences of a target,
active coaching by the experimenter seems to depress the hit
rate -. 28 without prompting versus .15 with prompting. For
second or later occurrences of a target, active coaching appears
to enhance the hit rate -. 33 without prompting versus .45 with
prompting. If the prompting by the experimenter is intended to
increase reliability by reminding the receiver of ganzfeld asso-
ciations that he or she might overlook during judging, why
should the effects of such prompting show up only for the sub-
sequent occurrences of a target?
That hit rate correlates with frequency of target occurrence
could mean that the "better" targets are somehow selected more
often by the randomizing procedure. Or it could mean that fre-
quency of occurrence, itself, is the determinant of a higher hit
rate. The data suggest the latter possibility. The 48 targets that
occurred exactly once in the database had a hit rate of .22. The
first occurrence of the targets that occurred more than once had
a hit rate of .23. The combined hit rate for second or later oc-
currences of targets was .39. Another way of examining this re-
lation would be to look separately at the hit rates for first and
subsequent occurrences of targets that appeared exactly twice,
three times, four times, and so on. Only the targets that oc-
curred from two to five times could be used because only one or
two targets appeared with frequencies of six or more. In all these
comparisons, the first occurrences consistently had a lower hit
rate than subsequent occurrences of the same targets.
Whatever the source for this pattern, it raises questions about
interpretations of other findings in the database. For example,
Bem and Honorton (1994) pointed to the high rate of .50 for
the 20 Julliard students as evidence for the effect of artistic cre-
ativity on hit rate. However, all of the sessions in which the Jul-
liard students appeared were prompted, and 15 of the 20 used
second or later occurrences of a target. On the five targets that
were occurring for the first time, these students got one hit. Con-
sequently, we cannot tell if the hit rate for these students reflect
any special abilities or if they are due to whatever makes hit rate
a function of target frequency and coaching in this database.
Are these findings due to an artifact, or do they point to some
new, hitherto unrecognized property of psi? We cannot say. The
existence of this pattern in the database. however, strongly sup-
ports the need to replicate the findings before we can be confi-
dent that the parapsychologists have finally found a way to cap-
ture and tame their elusive quarry.
Conclusions
The autoganzfeld experiments are a praiseworthy improve-
ment in methodological sophistication and experimental rigor
over the previous ganzfeld experiments. Despite these improve-
ments, the experiments fall disappointingly short in the critical
area of justifying the randomization procedures. Even though
all but one of the individual studies produced a positive effect
size and the overall effect was significant, the autoganzfeld ex-
periments do not constitute a successful replication of the orig-
inal ganzfeld experiments.
' As noted in Footnote 5 of the Bem-Honorton article, a recent review
of the original computer files uncovered a duplicate record in the auto-
ganzfeld database. This has now been eliminated, reducing by one the
number of sessions on which my analysis was based. Some experiment-
ers contributed only unprompted sessions, and some contributed
mainly prompted sessions.
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5

.
1
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
24
RAY HYMAN
Although Bem and Honorton point to consistencies between
the autoganzfeld results and those of previous parapsychologi-
cal research, these consistencies are more apparent than real.
On the other hand, as I have argued, important inconsistencies
exist between the two databases.
Three robust effects in the autoganzfeld database are the de-
pendence of hit rate on type of target (dynamic or static), target
occurrence (first or subsequent), and experimenter prompting
(yes or no). Although my looking for effects of the latter two
factors was motivated by my concern for possible randomiza-
tion deficiencies, their existence should interest both parapsy-
chologists and critics. This is because the existence of an effect
depends on these factors. The combination of dynamic targets,
repeated occurrences of a target, and experimenter prompting
produces a hit rate of .471 with 95% confidence limits from
305 to .629. The combination of static targets, first occurrences
of a target, and no prompting yields a hit rate of .178 with 95%
confidence limits from .066 to .336.
We do not have enough information to know if the depen-
dence of hit rate on target frequency and experimenter prompt-
ing involves response preferences for items within a target pool.
One way to ensure that such preferences do not bias the hit rate
is to present each member of a target pool equally often. I tried
to get some idea what the hit rate might be if each member of a
target pool had occurred equally often. I restricted myself just
to those target pools in which each member occurred at least
once.º The hit rate for first occurrences of a target in these target
pools was .275 with 95% confidence limits from .167 to .399.
(The hit rate for the targets in these target pools that occurred
a second or later time was .427.) This finding does not prove
anything, but it suggests that if the targets within each target
pool had occurred equally often, the results might have been
consistent with chance.
The autoganzfeld studies failed to replicate key findings of
the original ganzfeld experiments, even though the power was
sufficient. The positive effect size and significance depended on
a new type of target whose presentation involves a new technol-
ogy and on target repetition and experimenter coaching. What-
ever their source, these effects are new to the psi literature. We
do not know how much of this is unique to this experimental
setup and laboratory. For these reasons, we have to wait for fu-
ture attempts at replication to see if a replicable psi effect is at
hand.
6 I could not use higher frequency of occurrence because only three
target pools existed in this database that had at least two occurrences of
each of its members.
References
Bem, D.J., & Honorton, C. (1994). Does psi exist? Replicable evidence
for an anomalous process of information transfer. Psychological Bul-
letin, 115. 4-18.
Honorton, C. (1985). Meta-analysis of psi ganzfeld research: A response
to Hyman. Journal of Parapsychology, 49, 51-91.
Honorton, C., Berger, R.E., Varvoglis, M.P., M., Derr, P., Schechter, E.I.,
& Ferrari, D.C. (1990). Psi communication in the ganzfeld: Experi-
ments with an automated testing system and a comparison with a
meta-analysis of earlier studies. Journal of Parapsychology, 54, 99-
139.
Hyman, R. (1985). The ganzfeld psi experiment: A critical appraisal.
Journal of Parapsychology, 49, 3-49.
Hyman, R., & Honorton, C. (1986). A joint communiqué: The psi ganz-
feld controversy. Journal of Parapsychology, 50, 351-364.
Marks, D., & Kammann, R. (1980). The psychology of the psychic.
Buffalo, NY: Prometheus Books.
McCrone, J. (1993, May 1). Roll up for the telepathy test. New Scientist,
pp. 29-33.
Snedecor, G.W., & Cochran, W.G. (1967). Statistical methods (6th ed.,
pp. 246-248). Ames: Iowa State University Press.
Woodward, J.A., Bonett, D.G., & Brecht, M. L. (1990). Introduction
to linear models and experimental design. San Diego, CA: Harcourt
Brace Jovanovich.
Received July 26, 1993
Accepted July 27, 1993
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
Psychological Bulletin
1994, Vol. 115, No. 1, 25-27
Copyright 1994 by the American Psychological Association, Inc.
0033-2909/94/$3.00
Response to Hyman
Daryl J. Bem
R. Hyman (1994) raises two major points about D. J. Bem and C. Honorton's (1994) article on the
psi ganzfeld experiments. First, he challenges the claim that the results of the autoganzfeld experi-
ments are consistent with the earlier database. Second, he expresses concerns about the adequacy of
the randomization procedures. In response to the first point, I argue that our claims about the con-
sistency of the autoganzfeld results with the earlier database are quite modest and challenge his
counterclaim that the results are inconsistent with it. In response to his methodological point, I
present new analyses that should allay apprehensions about the adequacy of the randomization pro-
cedures.
I am pleased that Ray Hyman, one of parapsychology's most
knowledgeable and skeptical critics, concurs with Charles Hon-
orton and me on so many aspects of the autoganzfeld experi-
ments: the soundness of their methodology, the clear rejection
of the null hypothesis. and. of course, the need for further repli-
cation. I hope this brief response will further augment our areas
of agreement.
Hyman raises two major points about our article. First, he
challenges our claim that the results of the autoganzfeld studies
are consistent with those in the earlier database. Second, he ex-
presses concerns about the "incomplete justification of the ade-
quacy of the randomization procedures" and speculates that
inadequate randomization may have interacted with subject or
experimenter response biases to produce artifactual results.
Consistency With the Earlier Database
The earlier ganzfeld database comprised studies whose meth-
ods and results were quite heterogeneous. Consequently, one
-
cannot justify any strong claims that some subsequent finding
is either consistent or inconsistent with that database. For this
reason, Honorton and I were careful not to make such claims.
With regard to the major finding, we simply observed that ear-
lier studies had achieved an overall hit rate of about 33% (25%
would be expected by chance) and noted that the autoganzfeld
experiments achieved approximately the same effect size. End
of claim.
In general, the earlier database served primarily to suggest the
kinds of variables that needed to be examined more systemati-
cally or more rigorously in the new studies. For example, previ-
ous ganzfeld studies that had used multi-image View Master
slide reels as target stimuli obtained significantly higher hit rates
I am grateful to Richard Broughton of the Institute for Parapsychol-
ogy in Durham, North Carolina, for going through the original auto-
ganzfeld computer files with me to unearth the data necessary for the
additional analyses presented in this response.
Correspondence concerning this article should be addressed to Daryl
J. Bem, Department of Psychology, Uris Hall, Cornell University, Ith-
aca, New York 14853. Electronic mail may be sent to d.bem@cor-
nell.edu.
than did studies that had used single-image photographs. This
finding prompted Honorton and his colleagues to include both
video film clips and single-image photographs in the autoganz-
feld experiments to determine whether the former were supe-
rior. They were. Our only claim about methodological compa-
rability was the modest observation that "by adding motion and
sound, the video clips might be thought of as high-tech versions
of the View Master reels."
But Hyman argues at length that video clips are not really
like View Master reels. Surely this is a matter of interpretation,
but does it really matter? Usually in psychology, successful con-
ceptual replications inspire more confidence about the reality
of the underlying phenomenon than do exact replications. I be-
lieve that to be the case here.
An example of a variable selected from the earlier database
for more rigorous reexamination was sender-receiver pairing.
Previous ganzfeld studies that permitted receivers to bring in
friends to serve as senders obtained significantly higher hit rates
than did studies that used only laboratory-assigned senders. But
as we emphasized in our article, "there is no record of how
many participants in the former studies actually brought in
friends," and hence these studies do not provide a clean test of
the sender-receiver variable. Moreover, the two kinds of studies
differed on many other variables as well.
In the autoganzfeld studies, all participants were free to bring
in friends, and it was found that sender-receiver pairs who were
friends did, in fact, achieve higher hit rates than did sender-
receiver pairs who were not friends (35% vs. 29%). But the reli-
ability of this finding is equivocal. In the archival publication of
the autoganzfeld studies, Honorton et al. (1990) presented this
finding as a marginally significant point-biserial correlation of
.36 (p = . 06). In our article, however, we chose to apply Fisher's
exact test to the hit rates themselves. Because this yielded a non-
significant p value, we thought it prudent simply to conclude
that "sender-receiver pairing was not a significant correlate of
psi performance in the autoganzfeld studies."
But to Hyman, "this failure to get significance is a noteworthy
inconsistency." (In part, he makes it appear more inconsistent
than it is by erroneously stating that the earlier database yielded
a significant difference in performance between friend pairs and
nonfriend pairs. As noted earlier, this is an indirect inference at
best.)
25
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
-

(
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
26
DARYL J. BEM
I submit that Hyman is using a double standard here. If the
successful replication of the relation between target type and psi
performance is not analogous to the earlier finding with the
View Master reels, then why is this near miss with a method-
ologically cleaner assessment of the sender-receiver variable a
"noteworthy inconsistency"?
Hyman cannot have it both ways. If the heterogeneity of the
original database and the methodological dissimilarities be-
tween its variables and those in the autoganzfeld studies pre-
clude strong claims of consistency, then these same factors pre-
clude strong claims of inconsistency.
Randomization
As we noted in our article, the issue of target randomization
is critical in many psi experiments because systematic patterns
in inadequately randomized target sequences might be detected
by subjects during a session or might match their preexisting
response biases. In a ganzfeld study, however, randomization is
less problematic because only one target is selected during the
session and most subjects serve in only one session. The primary
concern is simply that all the stimuli within each judging set be
sampled uniformly over the course of the study. Similar consid-
erations govern the second randomization, which takes place
after the ganzfeld period and determines the sequence in which
the target and decoys are presented to the receiver for judging.
In the 10 basic autoganzfeld experiments, 160 film clips were
sampled for a total of 329 sessions; accordingly, a particular clip
would be expected to appear as the target in only about 2 ses-
sions. This low expected frequency means that it is not possible
to statistically assess the randomness of the actual distribution
observed. Accordingly, Honorton et al. (1990) ran several large-
scale control series to test the output of the random number
generator. These control series confirmed that it was providing
a uniform distribution of values through the full target range.
Statistical tests that could legitimately be performed on the ac-
tual frequencies observed confirmed that targets were, on aver-
age, selected uniformly from among the four film clips within
each judging set and that the four possible judging sequences
were uniformly distributed across the sessions.
Nevertheless, Hyman remains legitimately concerned about
the adequacy of the randomizations and their potential interac-
tions with possible receiver or experimenter response biases.
Two kinds of response bias are involved: differential preferences
for video clips on the basis of their content and differential pref-
erences for clips on the basis of their position in the judging
sequence.
Content-Related Response Bias
Because the adequacy of target randomization cannot be sta-
tistically assessed owing to the low expected frequencies, the
possibility remains open that an unequal distribution of targets
could interact with receivers' content preferences to produce
artifactually high hit rates. As we reported in our article, Hon-
orton and I encountered this problem in an autoganzfeld study
that used a single judging set for all sessions (Study 302), a prob-
lem we dealt with in two ways. To respond to Hyman's concerns,
I have now performed the same two analyses on the remainder
of the database. Both treat the four-clip judging set as the unit
of analysis, and neither requires the assumption that the null
baseline is fixed at 25% or at any other particular value.
In the first analysis, the actual target frequencies observed are
used in conjunction with receivers' actual judgments to derive
a new, empirical baseline for each judging set. In particular, I
multiplied the proportion of times each clip in a set was the
target by the proportion of times that a receiver rated it as the
target. This product represents the probability that a receiver
would score a hit if there were no psi effect. The sum of these
products across the four clips in the set thus constitutes the em-
pirical null baseline for that set. Next, I computed Cohen's mea-
sure of effect size (h) on the difference between the overall hit
rate observed within that set and this empirical baseline. For
purposes of comparison, I then reconverted Cohen's h back to
its equivalent hit rate for a uniformly distributed judging set in
which the null baseline would, in fact, be 25%.
Across the 40 sets, the mean unadjusted hit rate was 31.5%,
significantly higher than 25%, one-sample :(39) = 2.44, p = . 01,
one-tailed. The new, bias-adjusted hit rate was virtually identi-
cal (30.7%), ((39) = 2.37, p = . 01, faifr(39) = 0.85, p = . 40, indi-
cating that unequal target frequencies were not significantly in-
flating the hit rate.
The second analysis treats each film clip as its own control by
comparing the proportion of times it was rated as the target
when it actually was the target and the proportion of times it was
rated as the target when it was one of the decoys. This procedure
automatically cancels out any content-related target prefer-
ences that receivers (or experimenters) might have. First, I cal-
culated these two proportions for each clip and then averaged
them across the four clips within each judging set. The results
show that across the 40 judging sets, clips were rated as targets
significantly more frequently when they were targets than when
they were decoys (29% and 22%, respectively), paired ((39) =
2.03, p = . 025, one-tailed. Both of these analyses indicate that
the observed psi effect cannot be attributed to the conjunction
of unequal target distributions and content-related response bi-
ases.
Sequence-Related Response Bias
Hyman is also concerned about the randomization of the
judging sequence
because we can expect strong systematic biases during the judging
procedure. The fact that the items to be judged have to be presented
sequentially, when combined with what we know about subjective
validation . . . would lead us to expect a strong tendency to select
the first or second items during the judging series.
Hyman's hypothesis is correct: As shown in Table 1, receivers
do display a position bias in their judgments x2(3, N = 354) =
8.64, p <. 05, tending to identify as targets clips appearing either
first or last in the judging sequence. Moreover, the actual distri-
bution of targets across the judging positions also departs sig-
nificantly from a uniform distribution, x2(3, N = 354) = 7.83,
p < . 05, with targets appearing most frequently in the third po-
sition.
To determine whether the conjunction of these two unequal
distributions might contribute artifactually to the hit rate, one
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5

,
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5
RESPONSE TO HYMAN
27
can again combine the observed frequencies to derive an empir-
ical null baseline. As shown in Table 1, each proportion in the
second column can be multiplied by the corresponding propor-
tion in the third column to yield the hit rate expected if there
were no psi effect. As shown, the expected hit rate across all four
judging positions is 24.7%.
The pertinent fact here is that this is lower than the 25% that
would have been obtained if the target positions had been uni-
formly distributed across the sessions. In other words, the con-
junction of receivers' position biases with the imperfect ran-
domization of target positions works against successful psi per-
formance in these data. Again, inadequate randomization has
not contributed artifactually to the hit rates.
Alternative Randomizing Strategies?
Hyman suggests that "one way to prevent response biases
from distorting the hit rate is to use a randomizing procedure
that makes sure that each item within a target pool occurs
equally often." Coming from a critic as sophisticated as Hyman,
this is a very puzzling suggestion, because he appears to be sug-
gesting some variant of sampling without replacement, a proce-
dure that would virtually guarantee response-bias artifacts. For
example, if receivers tend to avoid selecting targets that ap-
peared in previous sessions, this response bias would coincide
with the actual diminishing probabilities that a previously seen
target would reappear. The experimenters-who participate in
many sessions and discuss them with one another-are in an
even better position to detect and possibly to exploit the dimin-
ishing probabilities of target repetition. Sampling without re-
placement is precisely what enables card counters to improve
their odds at blackjack.
Alternatively, perhaps Hyman is advocating a procedure in
which the experiment continues until each clip within a judging
set appears as a target a predesignated minimum number of
times. For purposes of analysis, the investigator then randomly
discards excess sessions until the target frequencies are equal-
ized at that minimum number. This would solve the response-
bias problem but would be enormously wasteful. Suppose, for
example, that only 4 sessions from each judging set would have
to be discarded, on average, to equalize the target frequencies.
With 40 judging sets, the investigator would end up discarding
160 sessions, equal to nearly half of the sessions that took
Honorton and his colleagues 61/2 years to collect! Only a study
with many fewer judging sets could reasonably implement this
strategy.
Hit Rates as a Function of Target Repetition
In his post hoc excursion through the autoganzfeld data, Hy-
man uncovered an unexpected positive relationship between hit
rates and the number of times targets had been targets in previ-
ous sessions. (Ironically, Hyman has been one of the most out-
spoken critics of parapsychologists who search through their
data without specific hypotheses and then emerge with unex-
pected "findings.")
If this finding is reliable and not just a fluke of post hoc ex-
ploration, then it is difficult to interpret because target repeti-
tion is confounded with the chronological sequence of sessions:
Table 1
Proportion of Sessions in Which Each Clip Was Selected as the
Target and Proportion in Which It Appeared as the Target
Position in
judging
sequence
Selected
as target
Appeared
as target
Expected
hit rate
(%)
1
.30
.25
7.5
2
.20
.24
4.9
3
.22
.31
6.7
4
.28
.20
5.6
Total
1.00
1.00
24.7
Note. N = 354 sessions.
Higher repetitions of a target necessarily occur later in the se-
quence than lower repetitions. In turn, the chronological se-
quence of sessions is confounded with several other variables,
including more experienced experimenters, more "talented"
receivers (e.g., Juilliard students and receivers being retested be-
cause of earlier successes), and methodological refinements in-
troduced in the course of the program in an effort to enhance
psi performance (e.g., experimenter "prompting").
Again, however, Hyman's major concern is that this pattern
might reflect an interaction between inadequate target random-
ization and possible response biases on the part of those receiv-
ers or experimenters who encounter the same judging set more
than once. This seems highly unlikely. In the entire database,
only 8 subjects saw the same judging set twice, and none of them
performed better on the repetition than on the initial session.
Similar arithmetic applies to experimenters: On average, each
of the eight experimenters encountered a given judging set only
1.03 times. The worst case is an experimenter who encountered
the same judging set 6 times over the 61/2 years of the program.
These six sessions yielded three hits, two of them in the first two
sessions.
At the end of his discussion, Hyman wonders whether this
relationship between target repetition and hit rates is "due to an
artifact or [does it] point to some new, hitherto unrecognized
property of psi?" If it should turn out to be the latter, then I
believe it only appropriate that parapsychologists reward his
serendipity by calling it the Hyman Effect.
References
Bem, D. J., & Honorton, C. (1994). Does psi exist? Replicable evidence
for an anomalous process of information transfer. Psychological Bul-
letin. 115. 4-18.
Honorton, C., Berger, R. E., Varvoglis, M. P .. Quant, M., Derr, P.,
Schechter, E. I., & Ferrari, D. C. (1990). Psi communication in the
ganzfeld: Experiments with an automated testing system and a com-
parison with a meta-analysis of earlier studies. Journal of Parapsy-
chology. 54. 99-139.
Hyman. R. (1994). Anomaly or artifact? Comments on Bem and Hon-
orton. Psychological Bulletin. 115, 19-24.
Received August 9, 1993
Accepted August 9, 1993 =
Approved For Release 2000/08/08 : CIA-RDP96-00789R003200100001-5