Exploring possible sender-to-experimenter acoustic leakage in the PRL autoganzfeld experiments.
We continue to differ over the degree to which the effect constitutes evidence for psi, but we agree that the final verdict awaits the outcome of future experiments conducted by a broader range of investigators and according to more stringent standards. (p. 351)
Hyman and Honorton then outlined these stringent standards, describing methodological and reporting recommendations for future ganzfeld studies. Several leading parapsychologists commented favorably on these recommendations (see the invited commentaries directly following Hyman & Honorton, 1986).
Bem and Honorton then discussed a later set of semi-automated ganzfeld experiments (the "autoganzfeld studies"). These studies were designed to overcome the methodological problems identified in the joint communique and were originally reported in a major parapsychological journal (Honorton et al., 1990). Hyman (1994) was asked to comment on the procedure and results of Honorton's autoganzfeld studies. Hyman commended "Honorton and his colleagues (1990) for creating a protocol that eliminates most of the flaws that plagued the original ganzfeld experiments" (p. 19), but he (1) noted that the results of the studies were inconsistent with the previous ganzfeld database and (2) challenged the adequacy of the randomization procedure used in the experiment. Bem (1994) argued against both of these notions.
The paper we now present provides an in-depth analysis of one non-psi hypothesis that could potentially account for the autoganzfeld results.
A BRIEF DESCRIPTION OF THE AUTOGANZFELD STUDIES
The autoganzfeld studies were designed and run by Charles Honorton and his colleagues at the Psychophysical Research Laboratories (PRL) in Princeton, New Jersey. The autoganzfeld procedure commonly used two participants, a "sender" and a "receiver." These individuals were placed in two separate rooms. The receiver was placed in a state of mild sensory deprivation. Many parapsychologists believe that if psi exists at all it is likely to be a weak signal that is easily masked by internal somatic and external sensory noise (see, e.g., Honorton, 1977). For this reason steps were taken to help the receiver minimize the effects of such noise. They consisted of having the receiver place ping-pong ball halves over his or her eyes and then bathing them in red light. This procedure has the effect of creating a homogeneous visual field. In addition, the receiver also heard white noise through headphones, creating an undifferentiated auditory field. Finally, the receiver usually engaged in some form of relaxation exercise to minimize any somatic interference. Once the receiver was in this state, the sender was repeatedly shown a video clip that had been randomly selected from a pool of 160 clips. The sender had been asked to psychically send this clip to the receiver. The receiver was asked to report all the ideas, images, and impressions that came into his or her mind (referred to as "mentation") during this sending period. Both the experimenter and the sender could hear the receiver's comments through headphones.
The receiver was then presented with a randomly ordered "target set," consisting of four video clips (the actual target and three decoys) and was asked to rate the amount of correspondence between his or her mentation and each of these clips. For many of the trials the experimenter assisted the receiver during this judging period. The sender could hear the judging process through headphones. The receiver scored a direct hit if the target receiving the highest rating was the target that had been viewed by the sender. The likelihood of scoring a hit by chance alone was 25%.
The autoganzfeld studies consisted of 11 separate series, with a total of 240 participants providing 354 trials. Each trial was run by one of eight experimenters. The studies started in 1983 and finished in 1989, when the laboratory was forced to close because of lack of funding. At the time of the laboratory's closure, many series remained unfinished; however, the results of the trials that had been run up until this time are impressive. Overall hitting rate for all trials was 34.5%. This is highly significant statistically. In addition, interesting patterns emerged in the data. Different types of targets seemed to have significantly different hit rates. The series had used two types of video clip: dynamic targets consisting of a moving image with accompanying soundtrack, and static targets consisting of a silent stationary picture. As predicted in advance of the studies, scoring was significantly higher for the dynamic targets than for the static targets. In addition, there was a significant correlation between receivers' introversion/extraversion scores and hitting, with extraverts having a higher hit rate than introverts.
POSSIBLE SENDER-TO-EXPERIMENTER ACOUSTIC LEAKAGE(1)
While the receiver was in the ganzfeld, he or she was asked to report any thoughts, feelings, and images that came to mind. This mentation was both recorded on audiotape and written down by the experimenter. After the sending period had terminated, the experimenter reminded the receiver of the various ideas and images that he or she had just mentioned. The experimenter paused between items as they were read back, to encourage the receiver to elaborate or comment on each item.(2) This occurred for all trials. The receiver then viewed four possible targets and was asked to rate the degree of correspondence between each potential target and his or her mentation. In 165 trials the experimenter interacted with the receiver during this judging process. During these trials, the experimenter pointed out potential correspondences between the receiver's mentation and each potential target.
Given that experimenters played an important role in the judging process, it is vital that they were completely isolated from the target or anybody who knew the target's identity. If this was not the case, they might inadvertently cue the receiver as to the correct target. The potentially large effect that unconscious experimenter cueing can have on psychological experiments has been well documented (see, e.g., Rosenthal, 1976). This cueing could have occurred when experimenters reminded receivers of their mentation, which occurred for all trials; it also may have taken place when experimenters actively helped receivers judge the four possible targets, which occurred for 165 of the 354 trials.
There are various ways in which information concerning the target could theoretically have leaked inadvertently to the experimenter. However, we will concentrate in this paper on just one, namely, the notion that the experimenter may have been able to unconsciously pick up and utilize any sound made by the sender.(3) Clearly, any possible channel of communication from the sender (who knew the target's identity) to the experimenter could compromise the autoganzfeld results. These two individuals were located in separate but adjoining rooms ([ILLUSTRATION FOR FIGURE 1 OMITTED], reproduced with permission from Honorton et al., 1990), and so any communication would have been acoustic rather than visual. During the 30-minute sending period, both the sender and the experimenter heard the receiver's ongoing mentation. If some senders vocally rewarded any of the receiver's relatively accurate comments, the experimenter would only have had to unconsciously register the presence (and possibly not even the content) of these noises to know that a certain section of mentation pertained to some aspect of the target. The fact that target sets were "constructed to minimize similarities among targets within a set" (Bem & Honorton, 1994, p. 9) would have made it easier for the experimenter to make implicit, unconscious use of this information. Thus, noise leakage during this period would be highly significant. Similarly, the sender heard the judging process; acoustic leakage at this point might have made the experimenter aware of more and/or different noises being made for one target than another.
The following sections examine the nature of any noises that could have been produced by some senders and the measures taken to acoustically isolate the experimenter from such sounds.
The sender's voice was not monitored or recorded at any point during the experiment, and thus there exists no objective record of the sounds that may have been made by some senders. As a result, one can only speculate as to the possibility of sender noise. It is important to know the instructions that were given to the sender and the likelihood that these instructions were understood, remembered, and obeyed. This section discusses both of these issues.
In an earlier version of this paper we stated that senders may have been encouraged to be vocally active in response to accurate mentation. This information came from Morris, Cunningham, McAlpine, and Taylor (1993). When describing the PRL autoganzfeld, they noted that senders "were encouraged to be vocally [italics added] supportive when they heard mentation that was on target" (p. 180). When a draft of the paper was shown to Rick Berger (one of the original autoganzfeld experimenters), he denied that this was the case, noting (Berger, April 20, 1994):
I vaguely remember you [the first author] asking me (while I was in Edinburgh) about whether we allowed them to vocalize during the sending period and I vaguely remember telling you that we did allow it. We DID in fact allow it during the debugging of the program, i.e., during the very early testing of the system. When we first discovered that the experimenter could sometimes hear the sender, we fixed the problem by soundproofing the sender's room [the effectiveness of the sound attenuation is examined more closely in the next section of this paper] and instituting the "silently communicate" rule.(4)
This "silently communicate" rule is described in Berger and Honorton (1985). They state that at the beginning of the sending period the sender's monitor displayed the following prompt:
You can adjust the sound level of both the receiver's voice and the target's soundtrack by turning the volume controls on the gray box.
Silently communicate the contents and meaning of the target to [receiver's name]. (p. 24)
This "silently communicate" message is also reported in Honorton et al. (1990, p. 109). In addition, Berger's listing of the autoganzfeld computer program confirms that the above instructions were shown on the sender's monitor prior to target presentation and that the word silently was flashed on and off to reinforce this aspect of the message. Similarly, Honorton et al. (1990, p.109) report that at the start of the judging period the sender's monitor carried the following message: "Silently direct [R's first name] to select the target that you saw" (p. 109). Again, the word silently was flashed on and off to reinforce this aspect of the message. Obviously, the senders may have followed these instructions and remained silent throughout the sender period.
There is, however, evidence to suggest that the need for silence was not necessarily conveyed to the sender as a rigid requirement. The 1983 PRL Annual Report notes:
We have sound-attenuated the sender's room, which in our estimation achieves approximately 70dB of attenuation in the auditory frequency range, allowing the sender to become more vocally involved in "sending" without fear of accidentally giving the experimenter cues to the target [italics added] (although the senders are still instructed not to vocalize). (p. 46)
(Again, the effectiveness of the sound attenuation is examined more closely in the next section of this paper.)
In the absence of any evidence that the senders did remain quiet (even if instructed to), it is important to note that there are factors that could have resulted in their failure to follow instructions. Some senders (the vast majority of whom were very strong believers in psi(5)) may not have been able to prevent themselves from becoming excited when hearing accurate mentation and may thus have inadvertently given off short but noisy cues of delight or encouragement.(6) These may have been vocal, or they may, for example, have involved senders stamping their feet on the floor or striking their hands against the arms of the chair. Senders may also have made more or different noise during the judging of the correct target than during the judging for the decoy targets. Senders who did occasionally vocalize during these periods may not have realized, because of the headphones they were wearing, how loud a noise they were making.(7) All of this may sound speculative, but there is evidence to suggest that the experimenters themselves were not convinced that all senders would obey the "silently communicate" instruction. Two reviewers of this paper noted that part of the autoganzfeld protocol specified that trials were to be aborted if the experimenters heard any noises coming from the sender's room.(8)
Thus, the detection of noise from the sender's room during "debugging" trials led to modifications in the soundproofing of this room. The effectiveness of these measures will be discussed in the next section. This soundproofing is critical, given that there is some uncertainty concerning whether all senders continually followed the "silently communicate" rule.
SENDER-TO-EXPERIMENTER ACOUSTIC ISOLATION
The previous section suggested that some senders may have produced noise in response to accurate mentation and/or the judging of the correct target. For this reason, we now focus on the degree of sender-to-experimenter acoustic isolation.
Impressive measures were taken to isolate the receiver from the sender. The receiver was placed in a specially built, industrial-standard (Industrial Acoustics Corp., 1205A Sound Isolation Room) sound-isolation room (Honorton et al. 1990, pp. 104-105). In contrast, the measures taken to acoustically isolate sender from experimenter were far less impressive. Whereas the receiver and sender sat in rooms (one of which was the industry standard sound-isolation room referred to above) 14 feet apart, the experimenter sat 12 feet away from the sender's room, in an area that was not nearly as acoustically shielded as the receiver [ILLUSTRATION FOR FIGURE 1 OMITTED].
There are two types of sound that may have leaked from the sender's room: structure-borne sound (caused, for example, by senders stamping their feet on the floor or striking their hands against the arms of the chair) and airborne sound (caused, for example, by senders giving off vocal cues of delight or encouragement). Each will be discussed in turn.
Insulation Against Structure-Borne Sound
The floors of the experimenter's area and the sender's room were on the same level (i.e., parts of the same construction). This is a violation of the normal procedure used to acoustically isolate one area from another. Structure-borne sound can travel great distances through most construction materials with only a tiny amount of energy loss. As noted by Everest (1994),
a slammed door . . . can cause the structure to vibrate very significantly. These vibrations can travel great distances through solid structure with little loss. With wood, concrete, or brick beams, longitudinal vibrations are attenuated only about 2 dB in 100 ft. Sound travels in steel about 20 times as far for the same loss! Although joints and cross-bracing members increase the transmission loss, it is still very low in common structural configurations. (p. 139)
For these reasons a properly acoustically isolated chamber involves suspending one entire room (with its own floor) inside another. Estimating the degree that structure-borne sound may have carried between the two rooms is difficult, even when one has accurate knowledge of the precise construction of the floor and its coverings. Honorton et al. (1990) do not provide any of these details. Many individuals involved with the experiment have confirmed that the sender's room was carpeted. This would certainly have helped dampen the effects of structure-borne sound but, depending on the structure of the floor (unknown to date), may not have completely eliminated it.
Insulation Against Airborne Sound
Airborne sound could have traveled from sender to experimenter via several pathways (Parkin, Humphreys, & Cowell, 1979).
Direct transmission through the wall separating sender and experimenter. To assess this possibility it is vital to know the precise construction of the walls of the sender's room. Honorton et al. (1990, p. 105) provide only the briefest description of these walls, simply noting that they are "double." No other published sources provide further details. However, we asked Ephraim Schechter about the construction of the walls. Schechter (May 16, 1994) originally noted: "The walls were frame and wallboard with fiberglass insulation between the inner and outer wallboard panels." In response to our request for more information Schechter (May 25, 1994) later recalled:
As I remember it, the space between the wall board sheets was filled with some kind of foam-core insulation intended both for thermal insulation and some sound blockage, instead of simple fiberglass, but I'm not as clear about that as about the other details. It would be best to check with someone else who was there at the time - Rick Berger or Mario Varvoglis might be able to confirm or correct what I recall.(9)
The authors contacted Berger about this and received the following reply:
I also have no information on the construction of the sender's room. I'm sure that Chuck [Charles Honorton] knew. I don't know who's alive who knows. Don McCarthy or Ed May may have some details, as I believe that they were given some charge of the PRL stuff after PRL's demise. There may be blueprints or such. (Berger, May 26, 1994).
We then contacted Don McCarthy, Ed May, and several other parapsychologists about this matter (via an electronic discussion group). None of them were able to provide any additional information concerning this aspect of the PRL set-up. We have been unable to obtain any blueprints of the PRL set-up.
We then contacted Lawrence Tremmel.(10) Tremmel (September 5, 1994) noted that these walls "were constructed of standard wallboard on a wood framed structure." We then asked Tremmel if he recalled any sort of insulation between the two wallboards. Tremmel (September 23, 1994) replied:
Since I was responsible for cabling at PRL, I was present when the electrician installed all of the cabling through the "sender's room." If there was some kind of foam core insulation in the walls, he would have had to dig out a hole in the walls to run the cables. He didn't.
Tremmel does note, however, that such insulation may have been put in place after his departure from PRL.
In short, the sender's room walls appear to have been the standard type of internal wall used in the American construction industry(11) (Everest, 1994), which may or may not have been filled with some type of insulation material. It is obvious that the sender's room was certainly not built to the same high specifications as the receiver's room. A properly acoustically isolated room is constructed by building one room inside another, with a sizeable gap between the walls of the inner and outer rooms. This was the case with the receiver's room, but not the sender's.
The sound insulation provided by any wall can alter dramatically if it contains cracks, small holes, or any possible air paths. As noted by Parkin, Humphreys, and Cowell (1979):
An air path is a sound path, and the smallest hole can reduce insulation performance markedly. The gap around pipework passing through a partition, a door ajar, ill-fitting joints - all contribute to a reduction in sound isolation. A most important aspect of sound insulation, often overlooked, is that the total sound insulation of a composite construction is determined to a large extent by its weakest link. . . . [An] example of a loss of insulation due to a weakness is a 25 mm square hole in plastered 230 mm brick wall, 2.5 metres high and 2.5 metres long. The potential 50 dB average insulation of the brick wall will be reduced to a real value of approximately 40 dB. (pp. 142-143)
There is no reason to assume that the walls of the sender's room would have contained any large cracks or holes. However, it certainly had electrical connections to other parts of the experimental suite, and its wall may have had small cracks, ill-fitting joints (connecting it with the floor, ceiling, and other walls), ventilation, and plumbing/lighting connections running through it. Although there is no reason to assume that such holes or cracks existed, no written sources describe these connections or the effect that they might have had on sound insulation. However, in reply to our request for this type of information, Schechter (May 25, 1994) noted:
There was a pair of connector panels on the outside and inside walls, just as there was on the receiver's isolation chamber. Sockets on the inner and outer panels were connected with wires that ran through the insulation on the wall. I think all electrical connections into the sender's room were made through this panel, but I am not certain.
Again, it is difficult to assess the construction of the sender's room because precise information does not appear to be available.
As we said earlier, Berger (April 20, 1994) noted that, when the experimenters first discovered that they could sometimes hear the sender, they attempted to soundproof the sender's room. This soundproofing consisted for the most part of placing acoustical files around the room. Honorton et al. (1990) described this set-up, noting that
the inside walls and ceiling of the Se's [sender's] room are covered with 4-inch Sonex[TM] acoustical material, similar to that used in commercial broadcast studios. (p. 104)
Both the U.K. supplier (Canford Audio, Tyne & Wear, England) and the U.S. manufacturer (Illbruck USA, Minneapolis, MN), of Sonex tiles clearly state that these tiles are designed to stop sound from being reflected back into a room, rather than stop it from leaking from the room. Engineers informed us that the files have a high sound absorption coefficient but a very low sound reduction index. Thus, almost none of the sender's sounds would be reflected back into the sender's room. Instead, nearly all of it would be transmitted through the tiles to the walls.
Direct transmission through the door/door frame of the sender's room to the experimenter. Another possible pathway for leakage involves sound traveling through the sender's door/door frame and into the experimental area. The sender's room had only a single (versus the receiver's heavy double) door. Honorton et al. (1990, p. 105) describe the door as "acoustical" but provide no further details. Again, no published or unpublished sources present additional information concerning the structure of this door. There also appears to be some confusion regarding this issue among the PRL investigators. Schechter (May 31, 1994) reports the discussions that he has had with other PRL experimenters concerning this issue. Schechter first thought that the door was an "ordinary" indoor one (i.e., one obtained from a building contractor, rather than a door especially designed and constructed for the room). Another PRL autoganzfeld experimenter corrected him, noting that (because of the electrical shielding) the door was made in part from steel.(12) Berger (June 8, 1994) also recalled that the door was partly steel. Schechter (June 6, 1994) has since added that the term acoustical (used by Honorton et al.  describe the sender's door) is
commonly used in the US building trades to mean a wood or steel-clad door filled with sound-absorbent extruded foam. . . . The door used was at least steel-clad, as part of the electrical shielding. So, while I still am not sure (it's an old memory), the "acoustic" description makes me lean toward the notion that the sender's room door was the foam-filled steel clad type.
Again, precise information does not appear to be available.
Whatever the construction of the door, it appears that the door and door frame may have allowed sound to leak from the room. This is suggested by the fact that the experimenters felt it necessary to try to prevent this leakage by placing a special barrier behind the sender's chair. Honorton et al. (1990) note that "a free standing Sonex-covered plywood barrier (5 ft wide by 8 ft high) positioned inside the sender's room, between Se's chair and the acoustical door, blocks sound transmission through the door frame [italics added]" (p. 104). However, it is unclear whether Figure 1 correctly represents the positioning of the Sonex barrier. Schechter (May 20, 1994; May 25, 1994) has stated that it was positioned nearer (and possibly parallel) to the door. If this were the case, it would provide a greater barrier to sound leaking through the door frame, but less protection against sound leaking through the wall directly separating sender and experimenter. Daryl Bern(13)(April 21, 1994) has reported that Berger also confirmed this set-up, noting that
Rick Berger told me that the schematic diagram that appeared in the 1990 [Honorton et al.] paper is a bit misleading in that it shows the plywood barrier a bit of a distance from the door when, in fact, it was pressed closely (into the frame, Eph?(14)) thus providing much better attenuation through the door and frame than would be implied by the diagram.
However, in an electronic mail to us, Berger (June 8, 1994) later noted:
The last time I was at PRL was around 1986. I can't vouch for the position of the barrier after that time, but prior to that time it was at a 45-degree angle with the door, directly behind the sender's chair. It was a very large and imposing barrier, and I recall about 24" clearance as you walk around it to enter the room.
Schechter (May 16, 1994) has also provided more information about the supposed gaps between door and door frame. He recalled that
the four-inch Sonex on the door fitted tightly against the four-inch Sonex on the walls and the door frame, so that the door had to be firmly pushed shut against the binding of the Sonex surfaces. Even granted that Sonex is a sound absorbent rather than a sound blocking material, the door-to-frame gaps were pretty thoroughly muffled.
We asked Schechter how this would have worked on the hinge side of the door. If the two sets of Sonex were so closely connected here, either the door would not have opened or, over time, the Sonex would have become crushed. Schechter (May 31, 1994) recalls that the sender's room door may have actually opened outwards (i.e., into the experimenter's area) and thus would not have crushed the Sonex on its hinge side. However, Schechter was clearly uncertain about this, adding:
If my memory that the door opened outward is correct, you have the answer to why the Sonex didn't crumple. But I want to hear someone else say that the door opened outward before I trust that memory. It's so different from the diagram.
All of the other individuals quizzed by the authors recall that the door opened into the sender's room (i.e., in agreement with Figure 1).
Flanking transmission: Flanking transmission refers to any indirect sources of sound (i.e., those not traveling straight through connecting walls and doors) that can propagate from source to receiver room. Parkin, Humphreys, and Cowell (1979) note that such leakage can take many possible routes. For example, small amounts of sound could have propagated along the floor, ceiling, or back wall (i.e., the wall opposite the sender's room door) of the sender's room and emerged in the experimenter's area. These types of leakage can often result in non trivial reductions in sound insulation. For example, Parkin, Humphreys, and Cowell (1979) note that sound traveling over a partition (via a perforated, suspended ceiling) can reduce the partition's sound insulation by 10 dB or more. We asked Tremmel about the construction of the ceiling in the PRL set-up. Tremmel (September 5, 1994) replied:
The building in which PRL was housed was a leased building. This means the building was designed to allow (and expect) changing tenants. To maximize the portability of the building, air conditioning and electricity ducts were hidden in the ceiling of the building. These ducts were covered by drop ceilings, suspended by metal hangers. Thus the entire building contained a large common space above the suspended ceilings. [italics added]
We further inquired whether the walls of the rooms penetrated into this "common space." Tremmel (September 21, 1994) replied:
All constructed rooms were built only to the dropped ceiling, leaving a common space above for utilities. I saw the framing of the rooms during my frequent visits to the laboratory while it was under construction.
We also inquired about the possible presence of air conditioning ducts linking the sender's room to the experimental area. Tremmel (September 21, 1994) replied that
all rooms were air conditioned. . . . New Jersey in July and August has an unbearable temperature (90+) and humidity (90+). Anyone locked away in the sender's room without air conditioning would have sweated to death.
Various steps can be taken to reduce the effects of flanking transmission. Walls and floors can be made to contain structural breaks that reroute or disperse sound away from the receiver's room. Ceilings can be constructed from moderately heavy membranes, treated to become airtight, have fewer points of suspension, and so forth. Honorton et al. (1990) provide no details about the ceiling in either the sender's room or the experimenter's area or about the construction of the wall that ran between the two areas. Again, we have been unable to discover any additional information concerning these constructions.
Assessing Sound Insulation
None of the published journal articles report any steps taken to explicitly investigate how well the sender was acoustically isolated from the experimenter. However, as noted in the section "Sender Noise," the 1983 PRL Annual Report states that "we have sound-attenuated the sender's room, which in our estimation achieves approximately 70 dB of attenuation in the auditory frequency range" (p. 46). No additional details of this testing are provided in this or any other PRL report. Thus, it is difficult to know for certain how this testing was carried out (e.g., the meaning of the phrases in our estimation or auditory frequency range) or how "allowing the sender to become more vocally involved in sending" relates to the issues discussed in the section of this paper concerned with the levels of sender noise. We asked Schechter about this and received the following reply (Schechter, May 20, 1994):
I'm not certain what "in our estimation" means either. I have a memory of somebody - probably Chuck [Charles Honorton], Rick [Berger] or Mario [Varvoglis] - checking sound transmission with a commercial decibel meter after the Sonex was added, but I don't recall details. I've asked a couple of people but got only equally vague memories, and the folks who have access to Chuck's files haven't found records of these tests. (That may not mean much - Chuck's files are pretty scattered now, some in Edinburgh, some in Durham and possibly some still in New Jersey).
We contacted Berger, who recalled the following (Berger, June 8, 1994):
My memory (I wish there was documentation) was that we used a Radio Shack meter and did various tests like getting people to yell directly behind the door, behind the barrier, etc. We were satisfied from our testing that we were achieving very high attenuation at the outside of the door. Since our real concern was with the experimenter seated across the room, with headphones on, listening to the amplified voice of the receiver and transcribing everything madly - we felt that we had completely obviated this concern. We were not on an unlimited budget. The Sonex cost a lot of money, but was the most cost-effective means to produce what we felt was complete attenuation between sender and experimenter. (I seem to recall we couldn't get any readings on the dB meter outside the room, no matter what was going on inside.)
There are several well-respected and practical ways of measuring the sound insulation between two rooms (see, e.g., Parkin, Humphreys, & Cowell, 1979, Chapter 9). Briefly, these consist of using various loudspeakers to accurately create the hypothesized sound source. These speakers emit white noise (or a warble tone(15)) at each frequency and direction of interest. The inputs to microphones in the source and receiving room are then subtracted from one another to discover the difference in sound pressure levels between the two rooms. This is usually repeated for at least six locations of the microphones for frequencies up to 500 Hz and at three additional locations for higher frequencies (with further placements being necessary if the difference between the microphone inputs differ by more than 6 dB in any one frequency). Honorton et al. do not seem to have carried out these types of procedures. Similar tests can be used to assess the degree of insulation against structure-borne sound. These are important because there are considerable difficulties involved in estimating the quantity of such transmission theoretically. We have discovered no record of Honorton et al. carrying out such tests.
ESTIMATING SOUND LEAKAGE(16)
This section assesses the plausibility of the acoustic leakage hypothesis outlined above. Unfortunately, because the PRL autoganzfeld has now been dismantled and existing reports contain little detail, it is impossible to accurately assess the amount of possible sender-to-experimenter acoustic leakage. It is possible, however, to estimate whether, under the best of all possible "leakage" conditions, any meaningful levels of sender noise could have reached the experimenter. The following sections outline this estimation.
As already noted, there are three acoustic pathways between sender and experimenter. First, sound may have traveled through the freestanding barrier and out through the sender's room door and/or gaps around the door frame. However, Berger noted that individuals yelling on one side of this door failed to produce any measurable noise on the other side of the door. Although these tests were poorly conceived, this anecdotal evidence does suggest that this would be an unlikely pathway for sound. Second, structure-borne sound (e.g., senders stamping their feet) may have escaped from the room via the floor. The theoretical assessment of structure-borne sound is, at best, problematic (Parkin, Humphreys, & Cowell, 1979). This, combined with the lack of detail concerning the construction of the PRL floors, makes the analysis almost impossible. Instead, the following estimations of sender noise will assess what is probably the most likely explanation - namely, that sound may have traveled from the sender, through the part of the sender's room wall nearest the experimenter, to the experimenter.
Estimating the Amount of Sender Noise Hitting the Sonex Tiles
We carried out sound tests to establish the intensity of sound that may have hit the Sonex tiles in the sender's room. Clearly, it was important to reconstruct the main features of the autoganzfeld sender's room. For this reason, the tests took place in a high quality sound-insulated anechoic chamber (in the Department of Mechanical and Aeronautical Engineering, University of Hertfordshire). The walls, ceiling, and base of this chamber are lined with large foam wedges (2.5 feet long, with a 4-inch square base). Because of their large surface area, these wedges provide greater levels of absorption than the original Sonex material. Unlike the sender's room, this chamber does not have a sound-reflecting floor. Instead, a thin sheet of wire meshing is suspended above the wedges covering the base of the chamber. This provides minimum opportunity for reflection but supports the weight of any individuals and equipment placed in the chamber.
Schechter (May 20, 1994; May 25, 1994) has noted that autoganzfeld senders sat with their backs to the Sonex barrier and were facing a large (20-inch) television monitor. There were three routes by which sound would have traveled from senders to the wall located on their right. First, some sound would have traveled directly from the sender to the wall. Second, the TV monitor would have reflected additional sound to the wall. Third, the carpet-covered floor would have received some sound (either directly from the sender or reflected from the TV monitor) and reflected some of it toward the wall.
To reconstruct these conditions, we asked individuals (both male and female) to sit approximately 4 feet in front of a sheet of glass that was the same size as a 20-inch television screen (to simulate the reflection of the sender's television). A microphone was placed approximately 5 feet behind the individual (the construction of the chamber did not allow a microphone to be placed to the side of the individuals). Unfortunately, it was not possible to place a solid covering on the mesh, and thus any sound traveling toward the floor would have been absorbed by the foam wedges and not (as was the case with the autoganzfeld sender's room) reflected back into the room.
Individuals were asked to say "Yes" in a jubilant but controlled manner. Over several trials the microphone registered between 65-71 dB of noise. The levels in the original sender's room may have been slightly higher because of (1) the reflections provided by the floor, (2) the fact that the foam used in the test chamber would have provided higher levels of absorption than Sonex, and (3) the sound levels to the side of the individuals would have been slightly higher than those directly behind them.
It was therefore estimated that the Sonex files on the relevant section of the sender's wall would receive a maximum of 75 dB of noise spread evenly across the 125 Hz to 4 kHz frequency range.
As noted above, the acoustic tiles on the wall and barrier would have prevented only a small amount of the senders' sounds traveling through to the wall. Badi (1994) and the sound engineers at Illbruck USA (the manufacturers of Sonex files) estimate that the files would have a sound reduction index of 2-3 dB across the frequencies of interest. For this analysis it was assumed that the tiles would reduce sound traveling through them by 2 dB.
Sender's Room Wall
As noted in the previous section, we have been unable to determine the construction of the sender's room walls with any great certainty. For this reason, calculations for each of the possible standard types of double wall will be given. Everest (1994) notes that a double wall could have had a sound transmission class (STC)(17) of either 43 dB (no fiberglass filling), 55 dB (3.5 inches of fiberglass filling), or 58 dB (9.5 inches of fiberglass filling). However, these figures are determined under well-controlled laboratory conditions (i.e., using walls with no cracks, no small holes, no flanking transmission from other walls and ceiling, etc.). Acoustic engineers informed us that it would be reasonable to assume that these flaws, combined with flanking transmission, could have reduced sound insulation by as much as 10 dB. Thus, for the present analysis, the sound transmission classes of these walls were taken as 33 dB (no fiberglass filling), 45 dB (3.5 inches of fiberglass filling), and 48 dB (9.5 inches of fiberglass filling).
Sound Loss to Experimenter
Any noise emerging from the sender's room would have to travel from the outside of the wall to the experimenter (a distance of approximately 12 feet). The loss in sound energy over this distance depends heavily on the shape, structure, and surface materials of the experimenter's area. Sound tests carried out by us in a similar-sized room (carpeted but with bare walls) indicated that sound traveling this distance could incur a minimum loss of 7 dB.
From Figure 1 it may appear that the experimenter's console would have obstructed the path of such a sound. However, this cannot be determined with any great sense of certainty. Schechter (May 20, 1994) has noted that his discussions with individuals involved in the PRL autoganzfeld have indicated "different memories of whether there was an open straight-line path from the sender's room walls to the experimenter's chair or whether the path was blocked by the experimenter's console." For this reason, the possible effects of the console have been excluded from the analysis.
Table 1 combines the above estimates. Averaged leakage figures are 35 dB (air-filled wall), 23 dB (wall filled with 3.5 inches of fiberglass), and 20 dB (wall filled with 9.5 inches of fiberglass).
[TABULAR DATA FOR TABLE 1 OMITTED]
UNCONSCIOUS EXPERIMENTER DETECTION AND UTILIZATION OF SOUNDS
To estimate whether the levels of sender sound might be unconsciously detected and utilized by the experimenter, it is necessary to estimate (1) the amount of noise that would have masked such a signal and (2) the signal-to-noise ratio currently thought to be involved in auditory subliminal perception. Each of these issues will be discussed in turn.
Estimating Masking Noise
There were two types of noise that would have helped mask any signals emerging from the sender's room. First, the experimenter was wearing "light" headphones (i.e., those resting on, as opposed to in or around, the ear [Schechter, June 3, 1994]). During the sending period, these carried the receiver's mentation. During the judging period, they carried both the mentation and (for dynamic targets only) a target soundtrack. When the receiver was speaking or the soundtrack of a target was playing, these would have carried 50-70 dB of noise. In the absence of these sounds, this figure could fall to a minimum of 30 dB. In addition, the experimenter's console contained some equipment (e.g., an Apple II Plus computer, a fan, and a videocassette recorder [Honorton et al., 1990, p. 104]) which would have produced a certain level of low frequency background noise. Richard Broughton (of the Institute for Parapsychology) has access to the original PRL equipment and has informed us (June 6, 1994) that the set-up currently produces 54-60 dB of noise at a distance of 2.5 to 3 feet from the machine.(18)
Combining these estimates is far from easy (see Parkin, Humphreys, & Cowell, 1979), especially because one was ambient and the other was delivered through headphones. In addition, the ambient noise in the experimenter's area would not have created an homogeneous field of sound. Instead, the complex absorption and reflecting properties of the room would result in certain areas having higher ambient noise levels than others (Parkin, Humphreys, & Cowell, 1979). However, the acoustic experts consulted by us believe it reasonable to set the minimum masking figure at approximately 50 dB.
Detection of Subliminal Auditory Stimuli
There is also a problem in estimating the signal-minus-noise (S-N) figure that might result if the experimenter subliminally detected sender signals. The literature on acoustic subliminal perception is both controversial and equivocal. Some theorists argue that the phenomenon has not been demonstrated, whereas others argue the opposite. However, in a recent review of this literature, Urban (1992, 1993) has noted that some studies have obtained positive results even when signals have been masked by as much as 30 dB of additional noise (i.e., a S-N ratio of minus 30 dB). This figure is based on auditory subliminal perception research that usually requires individuals to unconsciously register the meaning of a phrase or sentence. In the ganzfeld situation the experimenter may just have had to register the presence or absence of sender noise and possibly not its content: This detection may be possible under slightly higher values of masking. Table 2 contains the S-N ratios for the maximum leakage estimates for each type of wall. All of these estimates lie on or within the minus 30 dB S-N ratio. In short, if the maximum leakage estimates reflect the actual autoganzfeld set-up, it is possible that the experimenter may have been able to subliminally register sounds made by the receiver.
Utilization of Subliminal Stimuli
There is a problem in assessing the degree to which any signals, unconsciously detected by the experimenter, could have biased receivers' choice of target during the judging period. A large body of research has demonstrated the potentially large effect that such cueing can have within other types of psychological experiments (see, e.g., Rosenthal, 1976), but no previous work (in either parapsychology or subliminal auditory research) has explicitly tackled how this might work in the ganzfeld situation. However, Bem and Honorton (1994) report that one study in the ganzfeld database allowed the experimenter to know the identity of the target and also to interact with a receiver before judging. They noted that
TABLE 2 POSSIBLE LEVELS OF MASKING PROVIDED BY BACKGROUND NOISE FOR EACH WALL TYPE Wall Type Signal Mask Signal-Mask Air-filled 35 50 -15 3.5 inches fiberglass 23 50 -27 9.5 inches fiberglass 20 50 -30
if the experimenter who interacts with the receiver knows the identity of the target, he or she could bias the receiver's similarity ratings in favor of Correct identification. Only one study in the database contained this flaw, a study in which subjects actually performed slightly below chance expectation. (p. 7)
We contacted Bern and asked for additional information concerning this study. Bem (July 6, 1994) replied, noting that these comments were based upon a study by Palmer and Aued (1975). Bem and Honorton misunderstood the nature of this experiment: Palmer has assured us that the experimenter was kept blind to the identity of the target until after the completion of judging (Palmer, June 27, 1994).
It is very difficult to establish the plausibility of the sender-to-experimenter leakage hypothesis. This is due to (1) the dearth of accurate information regarding many aspects of the PRL autoganzfeld set-up and (2) the lack of unequivocal and predictive information within research on acoustics, psychoacoustics, and auditory subliminal perception. However, the analyses presented in these sections have argued that sender-to-experimenter leakage could, under certain conditions, have taken place. The figures presented here are not unreasonable, given what is known about the PRL set-up. They do, however, represent the most optimistic estimates and interpretations of subliminal literature. Both of these points should be borne in mind when attempting to decide upon the plausibility of sender-to-experimenter leakage.
ASSESSING THE ACTUALITY OF SENDER-TO-EXPERIMENTER LEAKAGE
In this section we discuss the question of whether there is any evidence to suggest that the acoustic leakage proposed above actually occurred.
Internal Patterns in the Database
The likelihood of potential nonpsi explanations can be assessed on the basis of whether they can account for internal effects found in the data. The sensory leakage problem, if valid, could help explain many of the patterns in the autoganzfeld database.
If the sensory leakage hypothesis were valid, one would expect the targets of the 165 sessions in which the experimenter interacted with the receiver (labeled "prompt trials") to obtain significantly lower ranks (i.e., more hits) than the targets in the 189 trials in which receivers judged on their own (labeled "no-prompt" trials). A Mann-Whitney U test comparing the two sets of trials reveals that this did indeed happen (z [corrected for ties] = 1.94, p = .026, one-tailed). However, it should be noted that this is also consistent with psi-type hypotheses,(19) and so it is difficult to tease apart normal from paranormal interpretations of the result.
One reviewer of this paper suggested that one might also predict that the ranks given to the targets from the 189 no-prompt trials should be at chance. In fact, these trials are independently significant. However, as noted at the start of this paper, during these trials the experimenter commenced the judging period by reading back the receiver's mentation, and therefore there was an opportunity for the experimenter to bias the receiver via subtle cues (e.g., unconsciously emphasizing certain parts of the mentation). In addition, there is no reason to assume that these trials were uncontaminated by other potential artifacts. For example, after running 80% of the autoganzfeld trials, Honorton noticed that dynamic targets were scoring much higher than static targets. Because only dynamic targets carried a soundtrack, Honorton wondered if the effect could be due to auditory leakage from target to receiver. To assess this possibility Honorton carefully monitored the input to the receiver's headphones. Honorton et al. (1990) describe the results of this testing:
With the VCR audio set to normal amplification, no auditory signal could be detected through R's [receiver's] headphones, with or without white noise. When an external amplifier was added between the VCR and R's headphones and with the white noise turned completely off, the soundtrack could sometimes be faintly detected [italics added]. (p. 132)
Thus, for 80% of the trials, a small auditory signal was being directly transmitted from the target soundtrack to the receiver's headphones.(20) This signal was weak and could be consciously detected only with the aid of an external amplifier. Nevertheless, it is another potential artifact, and, because the 189 no-prompt trials were run before this leakage was detected, it helps to illustrate one difficulty with attempting to assess the actuality of a potential artefact via this approach.(21)
The presence of experimenter cueing could also account for some of the other patterns in the autoganzfeld database. As noted above, dynamic targets yielded significantly higher scores than static ones. This difference was one of the main experimental hypotheses.(22) Thus, experimenters would have entered the studies highly motivated to find a difference between the two types of targets. This hypothesis could have become a self-fulfilling prophecy, with experimenters inadvertently providing more cues during the judging of dynamic, as opposed to static, targets. In addition, senders may have found dynamic targets more absorbing or exciting than static ones and thus made more noise during the sending and judging of dynamic as opposed to static video clips.(23) In fact, Mann-Whitney tests comparing static/dynamic target performance showed that dynamic targets outperformed static targets to roughly the same extent in both prompt and no-prompt trials, but neither difference reached statistical significance (prompt trials: z [corrected for ties] = -1.34, p = .090, one-tailed; no-prompt trials: z [corrected for ties] = -1.31, p = .095, one-tailed). This result is difficult to interpret because all of the no-prompt trials took place before the soundtrack leakage problem (outlined above) was discovered and eliminated. Clearly, this problem, if valid, could have led to increased hitting of dynamic targets in the no-prompt trials. The presence of experimenter cueing could also explain why certain groups of receivers (e.g., extraverts) scored higher than others (e.g., introverts). Perhaps some types of receivers (e.g., those who were more outgoing or socially oriented) were able to elicit more cues from the experimenter and/or made more effective use of these cues. These ideas can be tested by examining whether these patterns exist only in the prompt trials (a result that might then support the existence of the acoustic leakage) or also in the no-prompt trials (a result that would not support the hypothesis). The differential scoring of extraverts and introverts was found in the no-prompt trials (z [corrected for ties] = -1.89, p = .028, one-tailed) but not the prompt trials (z [corrected for ties] = -1.15, p = .124, one-tailed). This finding does not support the sender-to-experimenter acoustic leakage hypothesis.
Running Additional Autoganzfeld Sessions
One reviewer of this paper suggested that the leakage hypothesis could be tested by running another autoganzfeld study but randomly assigning trials to either (1) the old procedure (i.e., one including the possibility of sender leakage) or (2) a new (and flaw-free) procedure. Although we support the notion that critics should try to experimentally test their hypotheses, it would be difficult to accurately reconstruct the PRL set-up for various reasons. First, some of the information concerning the physical construction of the PRL is not present in any published sources and cannot be recalled by the autoganzfeld experimenters contacted by us. Second, it would prove difficult to match various attributes (e.g., hearing ability, belief in psi, experience in running the autoganzfeld) of any new experimenters and senders with those involved in the original autoganzfeld. The original experimenters (and some participants) could not be used, in part because they are now aware of the hypotheses being tested. In addition, running a ganzfeld study is very time-consuming and requires great dedication. A single ganzfeld session can take between 1 and 3 hours, with many trials being needed to have a realistic chance of obtaining a statistically significant result. Also, the ganzfeld procedure needs a relatively large amount of specialized equipment and laboratory space. Investing time and effort in a possibly inconclusive testing of sender-to-experimenter leakage would be seen by many as a waste of these limited resources.
Re-judging Existing Autoganzfeld Sessions
The transcripts of the receiver's mentation could be blind judged. Such judging would help remove any influence provided by the experimenter during the judging period. Assuming the data were not contaminated by any additional methodological weaknesses, a significant above-chance result would provide tentative evidence against the actuality of sender-to-experimenter leakage; a chance result would provide the opposite. Clearly, such an analysis could at best be only indicative because the judges (i.e., the subjects themselves and an experienced ganzfeld experimenter) used for the reanalysis might not be not as skilled or motivated as those used in the original studies. Despite these problems, we believe that such rejudging could prove both useful and constructive. For these reasons, we hope that such judging will be undertaken in the near future.
In this paper we have assessed the possibility of sender-to-experimenter acoustic leakage in the PRL autoganzfeld and, in doing so, have raised a number of questions related to methodology and to reporting.
First, to assess the possibility of acoustic leakage it was necessary to know exactly how the sender's room was constructed. Unfortunately, the written sources describing the autoganzfeld (e.g., Berger & Honorton, 1985; Honorton et al., 1990; the PRL annual reports) contained few details about this subject. The Hyman-Honorton joint communique recommended that research analysts should be able to reconstruct experimental procedures from the descriptions provided in written reports (Hyman & Honorton, 1986, p. 360). The autoganzfeld investigators were in clear agreement with this notion and provided an admirable description of many other aspects of their studies. However, they did not provide enough information about the sender's room. Future research should concentrate on developing strategies that would help parapsychologists provide a more complete, unambiguous, and reliable description of their studies. This research could involve, for example, learning how best to record the set-up and procedure of these studies using different types of media (e.g., still photographs, videotape, floor plans, etc.). As noted by Hyman and Honorton (1986), such detailed recording is not the norm in science. However, it is important within parapsychology if future analysts (and replicators) are to be able to accurately reconstruct past experiments.
Second, although the written reports did not provide very much information concerning the sender's room, additional details were available from some of the original autoganzfeld experimenters. The information provided by these individuals proved extremely valuable, and we are indebted to the experimenters for taking the time and trouble to make these details available. However, the autoganzfeld was constructed over 10 years ago, and thus it was not surprising that these individuals were unable to clearly recall all of the details needed to fully assess the leakage hypothesis (e.g., the exact materials used in the construction of the sender's room walls and the methods used to assess the acoustic properties of the room are still unknown). In addition, these memories often appeared vague and uncertain. This further emphasizes the need for investigators to make accurate records of studies when they happen, so that they will not have to depend on their memories (which are likely to be influenced by the effects of bias and difficulties of recall) at a later date.
Third, the information recalled by the experimenters revealed that some of the details in the written sources were inaccurate (e.g., the notion that senders were encouraged to be vocally supportive when they heard mentation that was on target, and possibly the way in which Honorton et al.  represented the position and angle of the barrier in the sender's room). Again, this situation highlights the need for researchers to ensure that information presented in written sources is both accurate and complete. In addition, researchers involved in multi-author studies should perhaps attempt to identify and comment on any discrepancies between the report of a study as agreed upon by all authors and any summaries of that study presented by other writers.
Fourth, when the reporting deficiencies discussed above were resolved (as far as it was possible), and the autoganzfeld procedure was reconstructed, the potential for sender-to-experimenter acoustic leakage appeared possible but improbable. Leakage seems possible, in part because (1) the acoustical tiles used by Honorton et al. would have prevented only a small amount of sound from escaping from the room, (2) the construction of the sender's "double" wall cannot be established with any great degree of certainty, and (3) the sender's room and experimenter's area shared the same floor level. The autoganzfeld experimenters were clearly aware of the potential problems that could be caused by sender-to-experimenter leakage and took some measures to minimize its occurrence. However, these measures do not appear to have been as effective as the experimenters may have believed them to be. Again, future researchers should try to prevent this type of problem from occurring. For example, researchers running these types of experiments might find it helpful to have a working knowledge of relevant areas of acoustics or be prepared to closely interact with acoustical engineers. The acoustics engineers consulted by us have suggested that the necessary level of sound isolation would only have been achieved for certain either by placing the sender in a custom-built sound isolation chamber (of the type used to house the receiver) or by retaining the sender's room but moving it much farther away from the experimenter.
In the final section of this paper we discussed some of the issues involved in establishing whether the potential sender-to-experimenter leakage actually occurred. The possibility of sender-to-experimenter leakage suggested that a cued experimenter might subsequently cue subjects about the correct target, particularly when the experimenter prompted the subject during the judging session. Analyses were undertaken to learn whether experimenter-prompted trials, as contrasted with those not thus prompted, were associated with a higher overall hit rate and whether they provided more evidence in support of two substantive hypotheses championed by Honorton, namely, better performance with dynamic than static targets and superior performance among more extraverted subjects. All of these analyses are correlational in nature, and their interpretation therefore is ambiguous. The analyses related to overall hit rate provided significant evidence in accord with the experimenter-cueing hypothesis, whereas those related to Honorton's favored hypotheses provided no support for cueing in the first case and contradictory evidence in the latter. Better evidence for or against the experimenter-cueing hypothesis could be sought through further research. For example, the database might be rejudged by external judges who have no opportunity to be prompted by a potentially knowledgeable experimenter. As such, it should be clearly noted that this paper has outlined the potential for the sender-to-experimenter leakage; it has not established that such an artifact actually occurred.
Future analysis of the autoganzfeld data may suggest that this potential artifact did not actually happen. This would not, however, alter the focus of this paper, namely, that important lessons should be learned from the potential for this artifact and from the way in which some aspects of the experiment have been reported. The authors feel sure that earlier commentators such as Bem and Honorton (1994) or Hyman and Honorton (1986) would have agreed with the need for further documentation about sensory shielding in this type of studies. Indeed, both have noted that studies need to be documented as fully as possible. The Hyman-Honorton joint communique noted that
we believe that readers (including research analysts and prospective replicators) should be able to reconstruct the author's procedures from the descriptions provided in the experimental report. Although this is not common practice in science generally, we believe it is important in areas such as parapsychology where routine replicability cannot be taken for granted. More detailed exposition of methods and procedures should serve not only to aid evaluation of research quality, but also to increase the likelihood that other investigators will be able to replicate the original investigator's results successfully. (Hyman & Honorton, 1986, p. 360)
The difficulty in achieving these standards is reflected in the fact that Bem and Honorton (1994) believed that the Honorton et al. (1990) publication did exactly this, as they note:
Because Honorton and his colleagues have complied with the Hyman-Honorton specification that experimental reports be sufficiently complete to permit others to reconstruct the investigator's procedures, readers who wish to know more details than we provide here are likely to find whatever they need to know in the archival publication of these studies in the Journal of Parapsychology (Honorton et al., 1990). (Bem & Honorton, 1994, p. 9)
Despite the problems outlined in this paper, we believe that the autoganzfeld studies (and the resulting database) represent an impressive achievement. The studies achieved a very high level of methodological sophistication. However, just as the autoganzfeld studies built upon the shortcomings of past studies, so future work should aim to identify and eradicate any errors contained in the autoganzfeld studies. Indeed, to a limited extent this has already started to happen. Discussions with one laboratory currently attempting to replicate the autoganzfeld studies (the Koestler Chair of Parapsychology at the University of Edinburgh) have resulted in various design modifications (see Dalton, Morris, Delanoy, Radin, Taylor, & Wiseman, 1996).
Parapsychologists may indeed be starting to corner their elusive quarry. However, future attempts to actually capture their prey will require further improvements in the way in which this type of experiments is designed, reported, and criticized. We hope that these improvements will continue to develop from the type of open and constructive debate currently surrounding the autoganzfeld studies.
We thank Daryl Bem, Kathy Dalton, and Richard Broughton for making available the autoganzfeld database. We are also grateful for the helpful information kindly provided by those individuals directly involved in the design, set-up, and running of the PRL autoganzfeld, including Rick Berger, Ephraim Schechter, Mario Varvoglis, Ed May, Lawence Tremmel, and Don McCarthy. Special thanks also go to Robert Morris, Katie Moore, Robin Taylor, M.N.M. Badi, Michael Adams, Peter Howell, Dean Radin, Rex Stanford, and two anonymous referees for their helpful comments on the work presented in this paper. An earlier version of the paper was circulated on a parapsychological electronic discussion network and was modified as a result of feedback received from Daryl Bem, Richard Broughton, John Palmer, Dick Bierman, Charles Tart, Topher Cooper, Brian Josephson, Roger Nelson, Adrian Parker, and James Spottiswoode. Finally, the first author gratefully acknowledges help given by the late Charles Honorton in conversations about the autoganzfeld. As a result of his untimely death in November 1992, parapsychology has lost one of its most creative and productive experimentalists.
1 Details of the autoganzfeld design, procedure, and results referred to in the following sections are based upon several sources. Published sources consisted of Berger and Honorton (1985), Honorton and Schechter (1986), Honorton et al. (1990), and Bem and Honorton (1994). Unpublished sources consisted of the autoganzfeld computer listing (R. Berger, Auto-ganzfeld program documentation. Princeton, NJ: Psychophysical Research Laboratories), experimenters' instructions for the autoganzfeld (attached to the autoganzfeld computer listing), and the PRL annual reports. Honorton's personal files concerning the PRL autoganzfeld are now at the University of Edinburgh and the Institute for Parapsychology, and some are possibly with Honorton's son in New Jersey. We have inspected the first two locations and have been informed that items in the third location should be classified as either destroyed or lost (D. McCarthy, June 23, 1994). We also contacted individuals directly involved in setting up and running the autoganzfeld studies (see the authors' note on p. 97) and have amassed a considerable number of personal communications concerning this topic. All personal communications cited are identified in the text by the person's name and the date of the fax, telephone call, letter, or electronic mail. Copies of all faxes, letters, and electronic mails cited are available upon request.
2 This information is to be found only in two sheets of paper entitled simply "Instructions," attached to Berger's listing of the autoganzfeld computer program.
3 Other possibilities revolve around possible cues provided by the videotaped targets. For example, the autoganzfeld studies used the same copy of the target during the sending period as the judging period. The repeated playing of the videotape during the sending period may have caused the sound/picture quality of that part of the videotape to slightly degenerate. If this were the case, the target may then have looked slightly different to the three decoys during the judging period. Alternatively, noises made by the videocassette recorder (VCR) could have cued the experimenter as to the target's identity. The VCR responsible for playing the target/decoys during the sending/judging periods was located just a few inches from the experimenter. Any idiosyncratic sounds made by the VCR during the sending period (made, e.g., while the videotape was playing or as it wound or rewound to the target clip) might have been unconsciously picked up by the experimenter and inadvertently used to cue the receiver when the same sounds were heard from the VCR during the judging period. The authors plan to examine the plausibility of these notions in future publications.
4 The information from Morris et al. was published in the 1993 conference proceedings of the Parapsychological Association and was clearly mentioned during the paper's oral presentation at that conference. The validity of this information was not questioned by the paper's referees, those attending the conference, or anybody after the conference (Morris, May 1994). This information was apparently based on conversations between Robert Morris and either Rick Berger, Mario Varvoglis, or Charles Honorton (Morris, May 1994); the latter three served as experimenters during the PRL autoganzfeld studies. Varvoglis (May 22, 1994) does not recall telling Morris this information. However, Berger (June 8, 1994) noted: "I may very well have told Bob [Morris] or anybody else that sender vocalization was not forbidden in the formal series. This was until I went back and probed my notes, in particular the program listings."
5 Honorton et al. (1990) note: "Belief is strong in this population. On a seven-point scale where '1' indicates strong disbelief and '7' indicates strong belief in psi, the mean is 6.20 (SD = 1.03); only two participants rated their belief in psi below the midpoint of the scale" (p. 102).
6 One reviewer of this paper noted that senders might have been equally likely to give off noises of disappointment when the mentation was not on target. We believe that sounds of delight were likely to have been louder than sounds of disappointment. If this were not the case, the experimenter would need to have subliminally registered the approximate content of the sender's noise, rather than simply its presence or absence.
7 Senders' headphones carried the receiver's mentation and, in the case of dynamic targets, soundtracks.
8 We asked these reviewers to specify the source of this information. Both believed that the autoganzfeld experimenters had stated that this was the case. We contacted three of the main experimenters (Rick Berger, Ephraim Schechter, and Mario Varvoglis) and asked for further information. All three confirmed that they believed this was the case (see the discussion for details), but neither they nor we were able to locate any written source that described this aspect of the autoganzfeld protocol. None of these individuals reported aborting any trials for this reason.
9 The citation of both of these quotes was requested by Schechter (May 31, 1994), who noted: "If you're going to use me as the source, I'd prefer that you give both possibilities for insulation [i.e., fiberglass and some type of 'foam-core' material] and make it clear that I don't claim to recall accurately. It does seem appropriate to stick with the conservative assumption (fiberglass) for the calculations, though." (See the acoustic analyses presented later in this paper.)
10 Tremmel was involved in the very early stages of setting up the PRL. Tremmel (September 5, 1994) described his role, noting: "Chuck [Charles Honorton] asked me to help design the new laboratory (PRL) layout in Princeton. Additionally, I was responsible for specifying the wiring for the experimental and monitoring rooms for audio, biofeedback and computer communications at PRL. It is interesting to note that PRL was established in a brand new research building in Princeton. We waited over a year for construction to begin. Because of the delay, I moved out to the Princeton area over a year prior to the actual establishment of the laboratory (and a year before Chuck). One benefit of my early move was that I was able to monitor the construction of the research building and the laboratory from the first day."
11 A double wall partition consists of two separate sheets of 5/8-in. Gypsum board. These are mounted on two separate 2 in. x 4 in. plates, and the gap between the sheets can be left empty or filled with varying thicknesses of fiberglass.
12 Schechter (June 8, 1994) has since asked us to note that his original memory was caused by his having "not been thinking in terms of electrical shielding."
13 Bem has asked us to make it clear that "I am serving as a secondary source. I was not an experimenter and am relying on communications with those who were, for my own understanding of the exact experimental setup."
14 Bem's message was part of an electronic discussion group. In this part of the message he was asking Ephraim Schechter for more information about this issue. In a note to us Schechter (May 20, 1994) wrote: "The Sonex-covered barrier in the sender's room, by the way, was fairly close to the door rather than directly back of the sender's chair." In reply to our request for more information, Schechter (May 25, 1994) further added: "My memory doesn't gibe with the diagram. I remember the panel as being parallel to the door. Check with Rick or Mario, though, and with someone who participated in sessions after I left. Maybe the position was changed sometime after I left."
15 A warble tone is a sound whose frequency is continuously varying in a regular manner within fixed limits.
16 The figures cited in this section are based on advice and help from M.N.M. Badi (Department of Mechanical and Aeronautical Engineering, University of Hertfordshire), Michael Adams (Canford Audio), sound engineers at Illbruck USA, and Peter Howell (psychoacoustics specialist, University College, London).
17 The STC is a value of the transmission loss (TL) at 500 Hz with attenuations of -16 dB at 125 Hz and +4 dB at 4 kHz.
18 This is higher than most Apple II computers because Honorton's machine is equipped with a Kensington System Saver fan.
19 For example, experimenter prompting during the judging period "was initiated because participants frequently failed to identify obvious correspondences between their mentation and target elements" (Honorton et al., 1990, p. 110).
20 Honorton et al. (1990) have argued that this acoustic leakage is unlikely to have affected the subjects' scores, noting that after the leakage was eliminated the hit rate actually increased. Unfortunately, all of these later trials were ones in which the experimenter helped with judging, and the increase in ganzfeld performance could be attributed to receivers' picking up on cues from the experimenter during the judging period. Indeed, during these later trials, the experimenters may have been highly motivated to have receivers score hits, because a reduction in ganzfeld performance could have indicated that the success in earlier trials was due to the audio leakage. This motivation may have made the experimenters better able to pick up any subliminal cues emerging from the sender's room and particularly eager to employ them during the judging period.
21 Interestingly, the hit rate in the 100 trials in which the opportunity for either type of potential sensory leakage is minimized (i.e., those trials employing static targets and no experimenter prompting during the judging period) do not deviate significantly from chance (n = 100, 26 hits, hit rate = 26%, p = .41).
22 This hypothesis stems from the fact that previous ganzfeld experiments using multiple image targets (e.g., from View Master slide reels) produced significantly higher hit rates than studies using single images as targets.
23 It should be remembered that the soundtracks of the dynamic targets would have been heard by the experimenter during the judging period and possibly could have helped to mask any noise emanating from the sender's room.
BADI, M. N. M. (1994). Sound transmission from the sender room in parapsychology experiment. (Technical Report.) Hatfield: University of Hertfordshire.
BEM, D. J. (1994). Response to Hyman. Psychological Bulletin, 115, 25-27.
BEM, D. J., & HONORTON, C. (1994). Does psi exist? Replicable evidence for an anomalous process of information transfer. Psychological Bulletin, 115, 4-18.
BERGER, R. E., & HONORTON, C. (1985). An automated psi ganzfeld testing system. Proceedings of Presented Papers: The Parapsychological Association 28th Annual Convention, vol. 1, 3-36.
DALTON, K. S., MORRIS, R. L., DELANOY, D. L., RADIN, D., TAYLOR, R., & WISEMAN, R. (1996). Security measures in an automated ganzfeld system. Journal of Parapsychology, 60, 129-147.
EVEREST, F. A. (1994). The master handbook of acoustics. New York: McGraw Hill.
HONORTON, C. (1977). Psi and internal attention states. In B.B. Wolman (Ed.), Handbook of parapsychology (pp. 435-472). New York: Van Nostrand Reinhold.
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., QUANT, M., DERR, P., SCHECHTER, E. I., & FERRARI, D. C. (1990). Psi communication in the ganzfeld: Experiments with an automated testing system and a comparison with a meta-analysis of earlier studies. Journal of Parapsychology, 54, 99-139.
HONORTON, C., & SCHECHTER, E. I. (1986). Ganzfeld target retrieval with an automated testing system: A model for initial ganzfeld success. Proceedings of Presented Papers: The Parapsychological Association 29h Annual Convention, 171-187.
HYMAN, R. (1985). The psi ganzfeld experiment: A critical appraisal. Journal of Parapsychology, 49, 3-49.
HYMAN, R. (1994). Anomaly or artifact? Comments on Bern and Honorton. Psychological Bulletin, 115, 19-24.
HYMAN, R., & HONORTON, C. (1986). A joint communique: The psi ganzfeld controversy. Journal of Parapsychology, 50, 351-364.
MORRIS, R. L., CUNNINGHAM, S., MCALPINE, S, & TAYLOR, R. (1993). Toward replication and extension of autoganzfeld results. Proceedings of Presented Papers: The Parapsychological Association 36th Annual Convention, 177-191.
PALMER, J., & AUED, I. (1975). An ESP test with psychometric objects and the ganzfeld: Negative findings. In J. D. Morris, W. G. Roll, & R. L. Morris (Eds.), Research in parapsychology 1974 (pp. 50-53). Metuchen, NJ: Scarecrow.
PARKIN, P. H., HUMPHREYS, H. R., & COWELL, J. R. (1979). Acoustics, noise and buildings. London: Faber and Faber.
ROSENTHAL, R. (1976). Experimenter effects in behavioral research. (Enlarged ed.) New York: Irvington Publishers.
URBAN, M. J. (1992). Auditory subliminal stimulation: A re-examination. Perceptual and Motor Skills, 74, 515-541.
URBAN, M. J. (1993). Auditory subliminal stimulation: Methods. Perceptual and Motor Skills, 76, 1103-1106.
Department of Psychology University of Hertfordshire College Lane Hatfield Hertfordshire AL10 9AB United Kingdom
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Psychophysical Research Laboratories|
|Author:||Wiseman, Richard; Smith, Matthew; Kornbrot, Diana|
|Publication:||The Journal of Parapsychology|
|Date:||Jun 1, 1996|
|Previous Article:||Alternative Realities: The Paranormal, the Mystic and the Transcendent in Human Experience.|
|Next Article:||Security measures in an automated ganzfeld system.|