Home :

You are here: home / staring experiment RESEARCH ON THE SENSE OF BEING STARED AT  Rupert Sheldrake, Ph.D. December 2000
Randomized Instruction Sheets A set of 20 Randomized Instruction Sheets for use with these experiments is now available: Click here for html version If you want to print the sheets, click here for the pdf version You will require Acrobat Reader for the pdf version.   Introduction The feeling of being looked at from behind is remarkably common.  So is the experience of looking at someone from behind and finding that they turn round.  Surveys show that about 90% of the population have personally experienced these phenomena.  If people really can tell when they are being looked at from behind, this suggests that an influence somehow reaches out from the looker.  If this influence cannot be explained in terms of normal sensory information, the implications are very far-ranging.  But there has been remarkably little research on this subject so far.  I discussed the background to this research in Chapter 4 of my book Seven Experiments That Could Change The World (Fourth Estate, London), gave preliminary results of my own experiments and described a simple experimental procedure that could be carried out at practically no cost.  Since my book was published, staring experiments have been conducted at more than 60 schools in Britain, America and Germany, and also by more than 20 other groups.   The results of these experiments have been positive in practically all cases.  The statistical significance of the staring effect in all the experiments taken together is overwhelming.   Currently, the odds against the positive results being due to chance are billions to one.  In the light of the results obtained so far, and as a result of recent research that I have carried out myself in schools in London, I have now developed improved methods for doing this research.  Below, I suggest several simple procedures that can be used with groups of adults, in family experiments at home and in classes in schools.  I have found these experiments can be carried out successfully with children as young as 8, and it may well be possible to do these experiments with children even younger.  I hope that it will be possible to repeat these experiments in a variety of schools and other settings because the results will be of much value for this research programme.  They are also fun to do, and provoke interesting discussions.  I would much appreciate receiving copies of the results and give my address at the end of this paper.
New: Online Method   Online Staring Experiment   Method 1.   Experiment with lookers and subjects within the same room. This experiment involves people working in pairs, one the looker and the other the subject. The subject sits with his or her back to the looker, and wears a blindfold (or if this is not possible keeps his or her eyes closed).  I have found it convenient to use blindfolds of the kind used on airlines and have been given some for my own research by Virgin Atlantic Airlines.   Lookers either look or do not look at the subjects in a series of 20 trials according to a random sequence.  This random sequence can be generated most simply by tossing a coin before each trial.  Heads means "look"; tails means "don't look".   Alternatively, random number tables or a random number generator on a calculator or computer can be used. A set of 20 randomized instruction sheets are available Click Here for the link. The instruction sheets give a sequence of randomized instructions telling you what to do in a series of 20 trials:  Even numbers mean "look", odd numbers "don't look". Each looker should keep a record on a sheet of paper, writing at the top the name of the looker and of the subject.  Then on the left there should be a column with the number of the trial, from 1 to 20; then a column saying what happened, "look" or "no", and then a tick or cross to indicate whether the subject's guess was correct or not, e.g.
1. Look	a
2. No	r
3.  No	a
4. Look	a
and so on .
Procedure The looker sits at least 1 metre behind the subject's back, and either looks or does not look at the subject in accordance with the random sequence of trials.  The lookers write their names on the score sheet and also the names of the subjects.  To signal the beginning of each trial, two methods are possible: Either Each looker has a clicker, bleeper or other mechanical signal to indicate when the trial begins.   (Mechanical signals are necessary, rather than the voice, to ensure than no subtle clues are given consciously or unconsciously)   Immediately after the signal has been given the looker either looks at the back of the neck of the subject, or looks away and thinks of something else.  Or   Instead of lookers proceeding at their own speed and giving their own signal, the teacher or experimenter can give a signal to the entire class, so that all trials are performed simultaneously.  This can be done by means of a bell, buzzer or simply by saying "Trial one: begin".  (Since all lookers are following different random sequences of instructions, the teacher's voice can give no relevant clues to the subjects). Then for the next trial saying, "Trial two: begin".  And so on.  The subject then says "looking" or "not looking", and the looker puts a tick or a cross against that trial depending on whether the subject is right or wrong.  The subject should not spend long thinking about it, but guess quite quickly; 10 seconds are long enough.   The looker records the subject's guess and then proceeds to the next trial.  The same procedure is repeated for all 20 trials. These trials are more interesting for the subjects and generally give higher success rates if feedback is given to the subject after each guess, in other words if the subject is told immediately each guess if it is correct or incorrect. After the series of 20 trials has been completed, the lookers hand in their score sheets.  The lookers and subjects then change places. Each new looker starts with a new sheet, and writes the looker's and subject's names at the top.  Obviously if pre-randomized sheets are used, the sheet should not be given to the looker until the subject is seated and has already got the blindfold on, so he or she cannot see the sequence in which the trials will be conducted. If more time is available then a third session can be conducted in which the looker becomes the subject again and the subject the looker, and so on. I have found by experience that it is possible to explain this experiment and carry it out with two sessions in about 45 minutes.  If a longer period is available, then more sessions are possible.
Method 2.   Experiments involving looking through windows. The experiment described above eliminates the possibility of the subjects knowing when they are being looked at through visual clues, especially if blind-folds are used.  But it is still possible that there is some leakage of information through sounds which may be given consciously or unconsciously by the looker.  Therefore a more rigorous form of this experiment is to do it under conditions in which sounds can be eliminated as well.  As in Method 1, the participants work in pairs, one the looker and the other the subject. The lookers look at the subjects through windows.   The lookers are inside, and have a list of instructions for the 20 trials.  The subjects are outside, sitting with their backs to the window and wearing blindfolds (or if no blindfolds are available with their eyes shut).  As described above for Method 1, the lookers either carry out the randomization themselves trial by trial by tossing a coin. Alternatively, a set of 20 randomized instruction sheets are available Click Here for the link. The instruction sheets give a sequence of randomized instructions telling you what to do in a series of 20 trials: either look ("look") or don't look ("no").  Each looker has a different random sequence of instructions.  The subjects sit outdoors with their backs to the window, wearing blindfolds if possible.  Each is given a score sheet which simply has numbers from 1 to 20 on it for each of the trials.  A specimen score sheet is shown below.  (Before the session begins, it is important to ensure that each subject has a book or other hard surface on which to rest the score sheet and a pen or pencil with which to record their guesses).   Before the session begins, each looker and subject writes his or her name and that of their partner on their sheet they have been given.  The subjects sit in a row with their backs to the windows, and the lookers are in a row indoors in the same order so that they are looking through the windows directly at their  partner.    For this experiment it is most convenient to have two adults supervising.  One is inside with the lookers, the other outside with the subjects.  When both lookers and scorers have written their names on their sheets and everyone is ready, the trial begins.  The teacher inside says 'Trial 1. Ready. Go' and at that stage signals to the teacher outside by tapping on a window, or by means of a bell or buzzer.  To avoid any possible confusion about the trial number, the teacher indoors can hold up to the window a piece of paper with a large number 1 written on it.  The teacher outdoors, who is the only person facing the window, then calls out to the subjects and scorers "Trial 1. Begin".   Alternatively, the signal can be given by means of a walkie-talkie type intercom, over which the teacher indoors can say the trial number. After 10 seconds, the teacher outdoors then tells the lookers to write down their guess.  To do this they push up their blindfolds, write "yes" if they think they were being looked at, or "no" if they think they were not, and then put their blindfolds back in place.   They should not at any stage turn around. When all are ready, the adult outdoors gives a thumbs up signal to the adult indoors who then says, "Trial 2. Ready. Begin", signalling with a bell, buzz, tap or intercom to the teacher outdoors that the trial is beginning, and which trial it is. And so on.  This procedure is repeated until all 20 trials have been completed.  The lookers then go outdoors to their partners,  and the two sheets from each group are collected by the supervising teacher and stapled together so that they can be scored later to see how many of the guesses were right and wrong. After the first session of 20 trials has been completed, the children change places, with the subject becoming the looker, and the looker the subject.  Each session need last no more than 10-15 minutes, and so with an initial period for explaining the experiment, the time needed for handing out sheets, moving chairs outdoors and changing over between sessions, I have found that the entire procedure need take no longer than one hour, during which time two sessions can be completed.  If more time is available then additional sessions are of course possible. With this experimental method, it is not possible to give immediate feedback to the subjects as to whether their guesses are correct or incorrect.  Generally speaking, subjects do not perform as well without feedback.   So if you want to compare Method 1 and Method 2,  Method 1 should also be carried out without feedback.
Using computers A more sophisticated way of doing these tests is to have the randomized instructions appear on the screen of a computer, together with spaces for filling in the names of the subject and looker and for recording the guesses.  In Method 1, the looker directly records the subject's guesses trial by trial on the computer.  For Method 2, these would have to be entered after the series of 20 trials from the subject's record sheet, unless the subjects are equipped with a remote keyboard on which they can key in their "yes" or "no" guesses trial by trial, which are then   recorded automatically on the looker's computer.  Additionally, if subjects have a computer monitor in front of them as well as a keyboard, they can be given immediate feedback about their guesses through the monitor.  Several people have already carried out staring experiments of this kind using computers, with positive and significant results.
Adding up the scores For each score sheet, the number of correct and wrong guesses in the looking trials is added up, the same is done for the not looking trials, and these subtotals are added together to give the totals.  The results for each subject can then be tabulated as follows:
Looker/subject Looking right Looking wrong Not looking right Not looking wrong Total right Total wrong NL/PJ      7     3     5      5     12       8 PJ/NL      5     4     5      6     10     10 JR/SW      8     2     6      4     14       6 SW/JR      6     4     5      5     11       9 AD/HK      5     6     4      5       9     11 Totals    31    19    25    25     56     44
All the columns of figures are then added up to give the totals for a given group of subjects.  Examples of data tabulated in this way from my own experiments, and from many independent experiments in colleges and schools are given in my recent papers on the sense of being stared at:  The sense of being stared at: Experiments in schools (1998) and The "sense of being stared at" confirmed by simple experiments (1999).
These papers also give examples of different methods of statistical analysis of the data. In my research so far, I have found that there is a characteristic pattern of results that seems remarkably repeatable.  The score in looking trials is usually above the 50% chance level, often as high as 60%, whereas that in not-looking trials is close to the chance level of 50%.  If you are interested in analyzing your own results  statistically, I recommend the use of the chi-squared test.  This can be applied to the total numbers of correct and incorrect guesses in the "looking", "not looking" and "total" columns, taking the result expected on the null hypothesis to be that the number of correct and incorrect guesses should be equal.  In addition, the data can be analyzed by another simple but effective method that also employs the chi-squared test.  For the looking trials, not-looking trials and totals, each subject is scored + if the number of correct guesses exceeds the number of incorrect guesses, = if the number of correct and incorrect guesses is the same, and - if the number of incorrect guesses is more than the number of correct guesses.  An example of this system of scoring is given in the specimen Table below,  using the same data as in the Table above:
Looker/subject Looking right Looking wrong Looking +/-/= Not looking right Not looking wrong Not looking +/-/= Total looking right Total looking wrong Total looking +/-/= NL/PJ      7     3   +     5      5    =     12       8    + PJ/NL      5     4   +     5      6 -     10     10    = JR/SW      8     2   +     6      4      +     14       6    + SW/JR      6     4   +     5      5     =     11       9    + AD/HK      5     6    -     4      5   -       9     11    - Totals    31    19 4+1-0=    25    25 1+2-2=     56     44 3+1-1=
The total numbers of + and - scores are then compared using the chi-squared test, taking the null hypothesis to be that there is no difference between the + and - scores.  For the purpose of this analysis the = scores can be omitted.  By carrying out one of the experiments described above you will be able to find out whether you get results that agree or disagree with this pattern.  Either way, please let me know.
Questions for further research Several important questions need to be investigated further, and you might like to consider research on one or more of the following: 1.  Given feedback, can subjects improve their performance over a number of sessions?  2.  Do looker-subject pairs who know each other well (e.g. twins or best friends) tend to do better than those who do not know each other so well? 3.  Through repeated testing of members of pairs with relatively high scores, is it possible to identify people who are unusually sensitive as subjects or unusually effective as lookers? 4.  How are the responses affected by the distance between the lookers and subjects? 5.   Are several lookers more effective than one?
Sending in the results I would be very grateful if you could send the results of these experiments to me, so that I can combine the results of experiments in different locations and analyze them statistically.  Please include the original score sheets for each subject (or copies of them) so that I can analyze them in detail.  Please also include a description of the methods you used, including details of the randomization procedure if you used your own; whether or not you gave feedback and how you signalled the beginnings of the trials; how far apart the subjects and lookers were, and whether the subjects were wearing blindfolds or not. Also note any other details that might be relevant.  I will send a report and analysis to all those who contribute data in this way. Rupert Sheldrake, 20 Willow Road. London NW3 1TJ. England.