Journal of Parapsychology 67, 187-206, June 2003

(Corrected version: In the original version as published there are some errors in Tables 3 and 10 which have been rectified in this version)

Most people have had experiences with telephone calls that appear to be telepathic. Either they think of someone for no apparent reason, then that person calls; or they know who is calling when the phone is ringing, before picking it up; or they call someone who says, “I was just thinking about you!” (Brown & Sheldrake, 2001; Sheldrake, 2000, 2003).

We have developed a simple experimental procedure for testing whether people really can tell who is calling. A participant receives a call at a prearranged time from one of four potential callers. The participants know who the potential callers are but do not know which one will be calling in a given trial. The caller is picked at random by the experimenter.

When the telephone rings, the participant guesses who is calling. The guess is either right or wrong. By chance, participants would be right about 1 time in 4, or, in other words, have a 25% success rate.

We have described elsewhere the results of more than 570 such trials, involving 63 participants (Sheldrake & Smart, 2003). The average success rate was 40%, hugely significant statistically (by the binomial test, p = 4 x 10^-16). In these trials the callers and participants were miles apart, and in some cases thousands of miles. The results implied that the participants’ above-chance success rate was a result of telepathy from the callers.

An explanation in terms of telepathy was also favoured by the fact that the success rates depended on who was calling. In real life, apparent telephone telepathy generally occurs between people who are emotionally bonded (Sheldrake, 2003). We carried out tests in which some of the potential callers were familiar people, family members, or friends, nominated by the participants themselves. Others were unfamiliar people whose names the participants knew but whom they had never met.

In these tests, involving 37 participants and 322 trials, 53% of the guesses were correct with familiar callers
(p = 1 x 10^-16), whereas with unfamiliar callers the results were exactly at the chance level of 25% (Sheldrake & Smart, 2003). The difference between responses with familiar and unfamiliar callers was highly significant ( p = 3 x 10^-7 by Fisher’s exact test.

In view of the supposedly elusive nature of psi and the difficulty of obtaining repeatable results, these data may seem too good to be true. Although they cannot be explained in terms of normal sensory clues, could they be explained by cheating?

We think cheating is unlikely to explain this effect. In general terms, it seems improbable that so many apparently honest people would cheat. It is also possible to make a quantitative estimate of cheating from empirical data. In our earliest series of experiments, we inadvertently provided an opportunity for cheating. The experimental data make it possible to find out whether this opportunity was taken up or not.

In these early experiments, we carried out two trials per session. Both callers were notified in advance. In some cases (about 1 in 4), the same person was picked for both calls. When a caller knew he or she would be calling again in the second trial, in the first trial the caller could have conveyed information, consciously or unconsciously, when he or she was speaking to the participant.

If such a leakage of information had happened, the chances of guessing correctly in the second trial would have increased. It was in fact slightly higher, 43% as opposed to 35%, but this difference was not statistically significant. Only in the case of 1 participant was there a striking and statistically significant difference between the success rate with first and second calls (Sheldrake & Smart, 2003). The results from the first trials alone, in which there was no such possibility of information leakage, were still very significant statistically.

Nevertheless, in the experiments we have just described we were relying on the honesty of the participants and their callers. In this article we describe further series of experiments in which participants were videotaped continuously in such a way that we could rule out the possibility that they were giving us false information about their guesses, or receiving telephone or e-mail messages from potential callers. Before the callers were selected at random, the participants were already on camera, and any incoming telephone calls or e-mails would have been detected on the film. Their guesses were recorded on the videotape before they picked up the phone.

If the previous positive results had been a result of cheating, then in these videotaped trials the scores should have slumped to chance levels. This was not the case. As we describe in this article, the average success rate in the videotaped trials was higher, not lower, than that in our nonvideotaped experiments.

Recruiting Participants
We recruited participants through advertisements in the Part-Time Work section of newspapers or through a recruitment Web site called www.hotrecruit.co.uk. Our advertisements read: “Do you know who is ringing before you pick up the phone? Good pay for fun and simple experiments as part of psychic research project.” We initially offered a payment of Ł10 per two-trial session, and later Ł10 for a one-trial session. We sent details of the test procedure to the people who replied to these advertisements and asked them to nominate people to whose calls they thought they might respond. We asked them to check that these people would be willing to take part and asked them to supply us with their contact details and telephone numbers. We also asked participants to tell us when they would be able to take part in tests and to check that their callers would be free to call them at those times.

Some of the participants who had completed an initial series of tests, involving at least 10 trials, were then asked to take part in further trials that would be videotaped, as described below. The 4 participants who did so were Sue Hawksley (SH), in Wakefield, West Yorkshire (date of birth March 23, 1964); Scott Reeves (SR), in Mayfair, central London (date of birth March 26, 1966); Claire Morsman (CM) in Twickenham, Middlesex (date of birth April 6, 1976); and Thomas Marcovici (TM) in Ealing, West London (date of birth February 5, 1977).

Sue Hawksley took part in a prespecified series of 100 trials with four callers nominated by herself (SH Series 1 and 2), and then in a further series of 85 trials with two familiar and two unfamiliar callers (SH Series 3). She also took part in a series of 17 trials conducted on a single day, with all the callers filmed (SH Series 4). The other 3 participants took part in a prespecified series of 30 trials each.

Callers
For the first two series of trials, we asked Sue Hawksley to nominate all four callers, who were familiar people to whom she thought she might react telepathically. In fact, not all four callers she originally nominated were able to take part in all trials, and so she nominated a further two callers who substituted for those originally chosen in some of the trials. In the subsequent series, following Methods 3 and 4 (described below), we asked participants to nominate only two or three callers. We supplied the other one or two, who were unfamiliar to the participants, but whose names were told to the participants. These unfamiliar callers were Pam Smart (PS), who lives in Greater Manchester, and Carole Macaulay, who lives in London.

Test Procedures
For each trial, there were four potential callers. The participants knew which callers would be involved and also knew that one of them would be selected at random by the throw of a die. For the throw of the die, we used high-quality casino dice and a ribbed casino-style dice cup, purchased in Las Vegas, Nevada. Each of the potential callers was assigned a number from 1 to 4 and was selected by the die showing one of these numbers after being thrown. If the die showed 5 or 6, then it was thrown again until a number between 1 and 4 came up. The randomizations were tested statistically as described below.

In all tests, the participants used landline telephones rather than cell phones, and in all cases only telephones without a caller identification system were used. The video camera was set up in a fixed position so that the telephone was in full view. The participants themselves switched on the video camera at the beginning of the session and switched it off after the trial had been completed. When a cassette was full, they mailed it to PS. In all cases the trials were filmed on time-coded videotape, with the date and time burnt into the film.

In the first series of trials with Sue Hawksley, the camera focussed on the telephone, and thus recorded exactly when it rang and what Sue Hawksley said about her guess before answering it, and also what she said when she did answer it. But she herself was not on camera all the time. In all subsequent trials with Sue Hawksley, and in all trials with the other participants, the participant sat in a chair in full view of the camera so that all actions and activities during the session could be recorded.

In all cases, when a trial was taking place and the phone started ringing, the participant said his or her guess to the camera before picking up the phone. In addition, in most trials, the participants were also asked to rate the confidence they felt in their guess, either “confident,” “not very confident,” or “just guessing.” Immediately upon picking up the phone, the participants again stated their guess by saying that person’s name before the caller said anything. The caller then revealed his or her identity, so the participants received immediate feedback.

In most cases the participants were alone in the house or apartment while the trials were taking place. However, during some trials with Sue Hawksley, her daughter (then age 8) was present in the house, and in a few of the trials with Thomas Marcovici, his father was in the house but in a different room. In all trials, Scott Reeves and Claire Morsman were alone.

We used four different procedures, involving progressive simplifications and also progressive increases in rigor. The four methods are described below.

1. Method 1 was used only in the first series of trials with Sue Hawksley. There were two trials per session. The two callers were selected at random by two throws of the die (ignoring 5 and 6). If the die showed the same number twice, then the same person was the caller in both trials.
   The times of the trials were also selected at random. Test sessions were usually an hour long, at times agreed in advance with the participants and their callers. To pick the call times at random, we divided the session into 6, and the beginning of one of these periods was selected by the throw of a die. For example, if the test session was from 10 to 11 a.m., the six periods began at 10-min intervals, starting at 10.10 a.m.. Thus if the die showed 4, then the test would be at 10.40 a.m.. The die was then thrown again to select the time for test with the other caller. If 1 came up, this meant 10.10 a.m. If the same number came up twice, then the die was cast again, to come up with a different time.
   The experimenter (PS) telephoned the randomly selected callers in advance, usually an hour or two beforehand, and asked them to call the participant, Sue Hawksley, at the randomly selected times and to think about her for about a minute beforehand. PS also rang the callers who had not been selected to tell them that they were not involved in this test session. This procedure was less rigorous than Methods 3 and 4, because it would have been possible for the callers to call Sue Hawksley before the tests began and inform her that they had, or had not, been selected.
   A few minutes after the tests, the experimenter rang the participant to ask what his or her guess had been, and in some cases also asked the callers, to make sure that that the participant had correctly reported the guess. In no cases did callers and participants disagree, nor did their reports differ from those recorded on the videotape, which was examined later. The experimenter recorded the result, noting down the date and times of each trial, the caller and the guess, and whether the guess was right or wrong.
2. In Method 2, also used only in the first series with Sue Hawksley, we simplified this procedure by using fixed times for the two trials in a session, for example at 10.15 and 10.30 a.m. As in Method 1, the callers were selected at random and informed an hour or two beforehand whether they had or had not been selected. The session itself began 15 min before the first trial, in this example, at 10.00 a.m. This is the time at which the participant switched on the video camera and sat in its field of view. The first caller to be selected was asked to call at 10.15 a.m., and the second (with a 1 in 4 chance that it would be the same person again) to call at 10.30. Potential callers who had not been selected were called and told they had not been selected. The experimenter found out and recorded what the guesses had been as in Method 1.
3. Method 3 was more rigorous than the previous methods. There was only one trial per session, and the session began 15 min before the time fixed for the trial, when the participant switched on the video camera. The experimenter (PS) selected the caller at random less than 15 min before the prearranged test time, while the participant was already on camera. Say the test time was 2.30 p.m., then the caller would turn on the video camera at 2.15, and the caller would be selected at random after 2.15 and notified before 2.20. Also, in Method 3, all videotapes were evaluated “blind” by a third party, as described below, who was not otherwise involved in the experiments.
   In Series 3 with Sue Hawksley, to start with (Method 3A) the experimenter called the three people who had not been selected to tell them so. From November 15, 2001, this step was eliminated (Method 3B). The potential callers were told that if they had not been notified at least 5 min before the test time, then they had not been selected. This simplification made it possible to carry out a series of separate, one-session trials in rapid succession, typically at 15-min intervals. Method 3B was used in all tests with the other 3 participants.
4. In Method 4, used in Series 4 with Sue Hawksley, all four callers were in the same location. The experimenter, Rupert Sheldrake (RS), selected the caller for each trial by the throw of a die. The selected caller then went to a separate room to make the call. The throwing of the die by RS and the subsequent movement of the caller to the separate room and then her or him making the call were filmed continuously by an independent cameraman. Meanwhile RS remained with the other three callers. Simultaneously, another cameraman was filming Sue Hawksley continuously in her house a mile away. The two cameramen were Nicholas and Mark Francis, who were brothers. We recruited them through the Institute of Communication Studies of Leeds University, where Nicholas Francis had been a student of broadcast journalism prior to this experiment. They received fees agreed in advance. They had no prior or subsequent connection with us, apart from the contacts needed for making the films, and then editing them into a synchronized split-screen format, in which Sue Hawksley can be seen on one side of the screen and the callers on the other.

Examination of the Videotapes
With Methods 1 and 2, the time-coded videotapes from these trials were viewed by PS, who noted down the participants’ guesses and the times at which they made them. She transcribed what they said about their confidence in their guesses and any other remarks. She also noted if there were any other telephone calls during the experimental period, or if the participant moved in such a way that he or she was off camera, even if only for a few seconds. Trials were disqualified if there were any other unscheduled telephone calls during the experimental period. Apart from the first series of trials with Sue Hawksley, in which she was not required to be on camera all the time, trials were also disqualified if a participant was not under observation for the whole of the experimental period.

For trials conducted by Method 3, the time-coded videotapes were evaluated blind by Dr. Amanda Jacks, who knew no details of the tests and was unaware who was calling or when the calls would happen. She recorded the participants’ guesses and comments, and the times at which they made them. She also noted if there were any other telephone calls or if the participant went off camera at any stage.

In the first series of 60 trials with Sue Hawksley, 4 trials were disqualified because of unscheduled incoming telephone calls. In the third series of 85 trials, 6 trials were disqualified because of incoming telephone calls during the experimental period, and 9 because she went off camera briefly, usually to attend to her 8-year-old daughter. In the case of Thomas Marcovici, 1 trial was disqualified because of an unscheduled incoming telephone call and 1 because he went off camera briefly, leaving 28 trials for analysis. It was not necessary to disqualify any of the trials with Claire Morsman and Scott Reeves.

Statistics
For the testing of the hypothesis that the proportion of correct guesses would be above the chance level of .25, or 25%, we used the exact binomial test (Siegel & Castellan, 1988). The null hypothesis was that the probability of a correct guess was .25.
We used Fisher’s exact test (Siegel & Castellan, 1988) for the comparison of results with familiar or unfamiliar callers, or the results in first and second trials. For comparing success rates in trials with the participant’s estimate of his or her confidence, we used the Cochran and Armitage trend test (Agresti, 2002). We calculated exact 95% confidence limits for the true probability of a correct response as described by Hahn and Meeker (1991).

Randomization Tests
Trials were randomized by the throw of a die, and for each trial, each caller was assigned a number from 1 to 4. In Series 1, 2, and 3 with Sue Hawksley, there were more than four callers altogether, although only four took part in a given trial; when one was unable to participate, another took her place and was assigned the same number as the person for whom she was substituting. We tested the randomization provided by the throw of the die by comparing the actual frequency of the numbers 1 to 4 with the expected frequency, using the chi-square test (with 3 degrees of freedom). We also compared the actual with the expected number of repetitions, that is, the number of times the same number appeared twice in a row, using the binomial test. The expected number of repetitions was 0.25( n - 1).

The randomizations are summarized in Table 1, and the statistical tests showed that in all cases the randomizations did not differ significantly from chance expectation at the p = .10 level. In this table, we included all the trials we conducted, although some were later eliminated from the analyses shown in the Results section of this article because the participants went off camera or received other phone calls during the experimental session.

First Two Series of Videotaped Trials With Sue Hawksley
As we have described elsewhere (Sheldrake & Smart, 2003), we carried out an initial series of 30 trials with Sue Hawksley that were not videotaped. She guessed the caller correctly in 14 trials (47%), significantly above the 25% level expected by chance ( p = .008).

In the present experiment, we conducted a prespecified total of 100 videotaped trials with her, in two series, the first of 60 trials and the second of 40. The first series of trials was carried out between September 12, 2000 and January 9, 2001. In the first series, Sue was not required to be on camera throughout the experimental period but only when actually answering the telephone. The first 36 trials followed Method 1, in which there were two calls per session, at randomized times. The remaining trials followed Method 2, with two calls per session at prearranged fixed times. In two of the 2-trial sessions there were unscheduled telephone calls before either of the trials had taken place. These 4 trials were therefore disqualified, leaving a total of 56 trials. (In the disqualified trials, 3 guesses were right and 1 was wrong.) Sue was right in 28 out of 56 trials (50%). The exact binomial is highly significant ( p = .00005).

The second series of 40 trials was carried out between January 10, 2001 and March 19, 2001. In the second series Sue was on camera continuously. All of these trials followed Method 2, with 2 trials per session at fixed times. Sue was right in 18 trials out of 40 (45%), again significantly above the chance level ( p = .005). Under these more rigorous conditions, her performance was not significantly different from that in the first series.

Sue’s success rate when calls came at random times (Method 1) was higher than with Method 2 when the calls came at prearranged times. She was right in 18 out of 32 trials with calls at random times (56%), and right 28 times out of 64 trials with calls at fixed times (44%), but this difference was not statistically significant ( p = .17).

With Methods 1 and 2, the use of 2 trials per session left open the possibility that there could have been a leakage of information from the first caller to the participant. At the time the first trial took place, the caller knew whether or not she would be calling again in the second trial and could therefore have consciously or unconsciously given some indication of this fact. If this were the case, then we would expect there to be more correct guesses in the second trials than in the first. This was not the case. In 45 sessions there were 2 trials per session, accounting for 90 of the 96 trials. The remaining 6 trials in the series had only 1 trial per session. In these 45 sessions, Sue Hawksley guessed correctly in 21 out of 45 first trials (47%), and in 22 out of 45 second trials (49%). There was practically no difference between these scores, and thus there seems to have been no significant leakage of information from the first callers.

To be additionally sure that no leakage of this kind could have accounted for Sue’s success, we could take the results only from the 45 first trials, plus those from the six sessions in which there was a single trial, in which there were 3 correct guesses. Adding together the first trials and single trials, there were 24 correct guesses out of 51 (47%), very significantly above chance ( p = .0005).

Altogether six callers took part in this series of 100 trials, although only four were involved in any given trial. All were either friends or family members and lived within a 5-mile radius of Sue Hawksley’s home. Her success rate with these callers varied considerably, being highest with her friends Jayne and Gayle (see Table 2). Her lowest success rates were with her sister, Gillian, and her mother, Emma.

The Third Series of Trials With Sue Hawksley, Comparing Familiar and Unfamiliar Callers, Using a More Rigorous Method

In our third series of videotaped experiments with Sue Hawksley we used Method 3, which was more rigorous than Methods 1 and 2, used in the preceding series. In Method 3, there was only one trial per session, the participant remained on camera throughout the entire experimental session, the callers were selected only after the participant was on camera, and all videos were evaluated blind.

There were 85 trials in this series, commencing on April 2, 2001 and finishing on October 2, 2001. All of these trials followed Method 3A. Two of the callers were friends who had already participated in the previous 100-trial series, Gayle and Jayne. The other callers were unfamiliar: one was PS, whom Sue Hawksley knew only as an experimenter, and the other callers were a computer, in the form of British Telecom (BT) alarm calls, and then, in place of the computer, Carole Macaulay, whom Sue had never met.

Of these 85 trials, 15 were disqualified, 6 because of unscheduled telephone calls during the experimental period and 9 because Sue went off camera briefly, usually to attend to her young daughter or to answer the doorbell when the postman came. (Her guesses were right in 7 [47%] of these 15 excluded trials.)

In the remaining 70 trials there were no interruptions or other telephone calls, and Sue was on camera continuously throughout the 15-min experimental period preceding the telephone call. Sue was right in 30 (43%) of these 70 trials, very significantly above the chance level ( p =.0008).

There was a striking difference between Sue’s performance with familiar and unfamiliar callers. With the two familiar callers, she was right 25 times out of 35 (71%; p = .00000001). With the unfamiliar callers, she was right only 5 times out of 35 (14%), not significantly different from the chance level (see Figure 1). The difference between success rates with familiar and unfamiliar callers was very significant statistically ( p = .000001).

Figure 1.   Percentage success in telephone telepathy trials with familiar and unfamiliar callers. The combined results from a total of 175 videotaped trials are shown first (total), then the data from the third and fourth series of trials with Sue Hawksley (S.H. 3 and 4), with Scott Reeves (S.R), Claire Morsman (C.M.), and Thomas Marcovici (T.M.). For comparison, the combined results are also shown from 190 trials involving 37 participants (37 P) in nonvideotaped trials (Sheldrake & Smart, 2003). The horizontal line indicates the chance expectation level of 25%.

The success rates with individual callers showed a wide variation, ranging from 81% with Jayne to 0% with Carole and the BT computer (see Table 3). An analysis of the pattern of guesses showed that Sue never guessed “Carole” or “BT” at all, and so there was no possibility of her being right when Carole or the computer called (see Table 3). She guessed “Gayle” 28 times, “Jayne” 32 times, but guessed “Pam” only 10 times. A higher proportion of “Jayne” guesses than “Gayle” guesses was correct.