Home

The Online staring experiment has now been running since 2003. Here is a summary of the results from October 14th, 2003 to January 12th, 2005.

Altogether 343 pairs of people have taken part, giving a total of 6,860 trials.

By chance people would be right 50% of the time. In fact the overall score is 60.6% correct.

Of the 343 subjects who took part:
232 scored 11 or more correct out of 20, in other words they were more right than wrong. 70 people were more wrong than right, scoring 9 or less out of 20
41 people got 10 out of 20, exactly at the chance level.


Figure 1. Success rates in looking, not-looking trials and the overall totals.
Above: The results are expressed as the percentage of guesses that were right or wrong.
Below: The number of people who were more right than wrong and more wrong than right.

The statistical significance of this result is astronomical, with odds against chance of quadrillions to one.

There was a difference in the results in looking and not looking trials, which agrees with the pattern shown in previous research on the sense of being stared at.
For previous research paper Click here
People were more successful in the looking trials then in the not looking trials. In the looking trials the average success rate was 65.6%, and in the not looking trials 55.9% (Figure 1).

I think this difference arises because if we have a sense of being stared at, it is likely to work when we are actually be stared at, in other words in the looking trials. In the not looking trials, people are being asked to detect the absence of a signal, which is much harder to do, and under those conditions many people are just guessing.

Because these experiments took place under uncontrolled and unsupervised conditions, we cannot eliminate the possibility that some people were cheating, or that some starers were inadvertently giving clues to the subjects by the way they gave the signal for the beginning of the trial or by unintentional sounds that were different in the staring and the not staring trials.

However, this positive result is very encouraging and suggests that this experiment is well worth doing under more rigorous conditions. All those who take part should make sure that they use a beeper or a mechanical signal for indicating when the trials begin. If people give the signal by saying start or by using their voice in any other way, this could give unintentional clues to the subject.

Some people did the trials without feedback (71 pairs) while most people did them with feedback (231 pairs) however, there was no significant difference between the results with and without feedback.

Some people used blindfolds (139 subjects) and some did not (163). Again, there was no significant difference in the results with and without blindfolds.

The fact that the results were practically the same with blindfolds and without feedback shows that the positive findings in this experiment cannot be explained either in terms of peripheral vision or in terms of learning to detect subtle sensory cues, because such learning would not be possible without feedback.

Do try this test yourself! Click here

I am grateful to Charles Overby for programming this experiment.