One group of parapsychological studies, the ganzfeld studies, have received more recent publicity, in terms of published articles examining the overall effect of the database, than has any other area of psi research. This attention is the result of detailed meta-analyses of the ganzfeld studies by a leading ganzfeld researcher, Honorton [9], and a critic of this work, Hyman [10]. Following the publication of these two meta-analyses, many other ``pro'' and ``con'' evaluations and commentaries have been published [11, 12, 13, 14]. The ganzfeld debate, often referred to as the ``Honorton--Hyman debate'', will be summarised below, but first a brief description of a ganzfeld study will be presented.
The ganzfeld technique consists of presenting a relaxed percipient with homogenous, unpatterned visual and auditory stimuli, which assists in increasing the mental imagery experienced by the percipient. While receiving this stimulus, the percipients verbalise all their experiences, their goal being to gain impressions which will relate to a sensorially isolated and remote target picture or short video clip. The ``target'' is being watched frequently by another person (a ``sender'' or ``agent'') who is attempting mentally to convey impressions of the target to the percipient or ``receiver''. These studies utilise a ``free-response'' methodology, in which the contents of the target material are unknown to the receiver (i.e., the percipient is ``free'' to respond with whatever impressions they generate, as he or she has no information regarding the specific contents of the possible target). The most common method of analysis used in ganzfeld studies is for the percipient or an independent judge(s) to compare the obtained impressions to four different target pictures/video clips, one of which is a duplicate of the actual target, looking for similarities. Using blind procedures, the judge has a one in four chance of correctly identifying the actual target (i.e., mean chance expectancy = 25 per cent ``hit'' rate). Study outcome is based upon whether similarities between the percipient's impressions and the actual target enabled the target to be correctly identified significantly more often than chance would allow. For further information regarding this experimental technique, procedural details and methods of analysis, see Honorton [9], and Honorton et al. [15].
A meta-analysis of twenty-eight ganzfeld studies was performed by Honorton (1985) [9], in response to a flaw analysis of the ganzfeld database conducted by Hyman (1985) [10]. Hyman found a highly significant overall effect in the database, but concluded that this effect was negated as he found a significant relationship between the study outcomes and procedural and statistical flaws contained in the studies. However, Hyman's flaw categorisations were severely criticised by Honorton, and a psychometrician, Saunders [16], found faults in Hyman's statistical analyses.
Honorton's meta-analysis found there were no significant relationships
between study outcomes and quality. The overall composite (Stouffer)
z score for the 28 ganzfeld studies included in the Honorton
meta-analysis was highly significant (
, two-tailed).
The effect sizes were homogeneous,
overall and across experimenters. The discrepancy between the
Honorton and Hyman analyses of the ganzfeld studies prompted
a further meta-analysis by Rosenthal [17], an independent specialist
in meta-analysis. Like Honorton, Rosenthal found an overall composite
z score of 6.60 for the twenty-eight ganzfeld studies.
His file drawer estimate agreed with that of Honorton, requiring
423 unreported, null studies to negate the significance of the
database. Here it is worth noting that another critic, Blackmore
[18] conducted a survey to discover the number of unreported
ganzfeld studies in 1980, prior to the Honorton/Hyman debate.
Her survey found 32 unreported studies, of which 12 were never
completed, and one could not be analysed. Of the remaining 19
studies, 14 were judged by Blackmore to have adequate methodology,
with 5 of these (36 percent) reporting significant results. She
concluded that ``the bias introduced by selective reporting of
ESP ganzfeld studies is not a major contributor to the overall
proportion of significant results'' (p. 217). Rosenthal, after
considering the possible influence of various flaws upon study
outcome, concluded that the overall hit rate of the studies could
be estimated to be 33 percent, whereas chance expectancy was
25 percent.
In 1986 Honorton and Hyman published a ``Joint Communiqué'' [19] in which they agreed that there was an overall effect in the database, but continued to disagree as to what extent this effect may have been influenced by methodological flaws. In their communiqué they outlined the necessary methodological precautions that should be taken to avoid the possibility of future studies giving rise to the same level of debate that had surrounded the previous ones. They concluded that more studies needed to be conducted, using the controls they had documented, before any final verdict about the database could be reached.
Honorton and his research team proceeded to design a new ganzfeld system which met the criteria he and Hyman had specified in their communiqué. This system, and studies using it, are referred to as ``autoganzfeld studies'', as much of the procedure is under automated computer control in order to avoid the problems found in some of the earlier studies. Before Honorton's lab closed in 1989, 11 experimental series, representing 355 sessions, conducted by eight experimenters, had been collected using the autoganzfeld. Honorton et al. [15] published a summary of the autoganzfeld studies and compared them with his earlier meta-analysis. The autoganzfeld sessions yielded overall significant results ( z = 3.89, p = 0.00005), with an obtained hit rate of 34.4 percent (with 25 percent being chance expectancy). The effect sizes by series and by experimenter were both homogeneous. Comparing the autoganzfeld outcomes to those of the 28 studies of the earlier meta-analysis revealed very similar outcomes, with the autoganzfeld showing slightly better ESP scoring than that obtained in the earlier studies (autoganzfeld results by series: effect size or es = .29, earlier 28 meta-analysis studies by experiment: es = .28).
Hyman, in 1991 [20] commenting upon a presentation of these results
by the statistician, Utts [12], concluded that ``Honorton's experiments
have produced intriguing results. If, as Utts suggests, independent
laboratories can produce similar results with the same relationships
and with the same attention to rigorous methodology, then parapsychology
may indeed have finally captured its elusive quarry.'' (p. 392).
Replications are currently being undertaken at various labs;
the only replication using a full autoganzfeld environment which
has been reported to date was conducted at the University of
Edinburgh [21], where the obtained significant, overall hit rate
was 33 percent (
). This outcome is consistent with Honorton's
autoganzfeld scoring rate of 34.4 percent, and replicates Rosenthal's
hit rate estimate based on the earlier ganzfeld studies. The
procedure for the Edinburgh study incorporated additional safeguards
against subject and experimenter fraud.