An Explanation of Each of the Experiments Within Bem’s ESP Study

A scientific paper by Daryl Bem, a social psychologist, will imminently be published in the prestigious Journal of Personality and Social Psychology. What is significant about this paper is that it supports the case for extra sensory perception (ESP.) The paper also presents data that suggests that humans may be able to predict things before they happen. The paper has passed the peer-review process and has been accepted for publication.                                                                                                                                                                         I am personally a sceptic but also someone who embraces the scientific method. Until these fascinating and perplexing results are consistently replicated, I will still be a sceptic concerning all matters ESP. Despite my own position, I have attempted to present the data from Bem’s paper in an objective manner. It should be emphasised that each of the experiments involved the participants predicting a future event. Below I have explained each individual experiment of Bem’s.

Before the experiments

Prior to the experiments, participants were asked to self evaluate themselves based on two statements. 1. ‘I am easily bored’ and 2. ‘I often enjoy seeing movies I’ve seen before.’ Respondents were asked to rate, based on a 5 point scale, these statements across the spectrum from ‘very untrue’ to ‘very true.’ These questions categorised the participant into either a stimulus seeking group or a non-stimulus seeking group. Several past psi studies have concluded that there is a correlation between extraversion and psi performance. It has been speculated that the extrovert searches for stimulation due to a susceptibility to boredom.

Experiment 1

This first experiment was titled ‘Precognitive Detection of Erotic Stimuli’ and involved 100 Cornell undergraduate students. Prior to the test, each participant undertook relaxation for a 3 minute period. The test commenced with two curtains appearing on the computer screen- one on the left side and one on the right side. At this stage, neither curtain had an image behind it. The participants were instructed to predict which curtain would have an image behind it. This image would appear behind the curtain after the participant recorded their guess. The specific curtain with the image behind it would be selected randomly by the testing apparatus and would be independent of the participants’ choice. In essence this test involved using erotic images in posterity as reinforcement to the participants to correctly perceive the future position of the erotic images themselves. Non-erotic, neutral, positive, negative and romantic (non-erotic) pictures were inter dispersed amongst the erotic pictures.

This first experiment involved a total of 36 trials. The psi hypothesis of this experiment was that the future position of the erotic image would be selected significantly more often than by chance (50%.) It is important to emphasise that this was indeed a test of precognition as the position of the image and the nature of the image were not determined by the computer until after the participant made their guess. This was a test of predicting a future event.

Results

The participants in this first experiment correctly selected the future location of the erotic images ‘significantly’ more frequently’ than the chance 50% rate. Overall, the future position of the erotic images was selected with a 53.1% success rate. This was in contrast to the future location of the non erotic images being chosen with 49.8% accuracy. Those ‘stimulus seeking’ individuals selected the future position of the erotic picture at a rate of 57.6%.This same group selected the future location of non-erotic images 49.9% of the time. The participants who were low in stimulus seeking selected the future location of the erotic image at a 49.9% success rate. They also selected the future position of the non-erotic image with 49.9% success.

Summary of Experiment 1

Step 1: A left and a right curtain are shown.

Step 2: The participants guess which curtain will have an image behind it.

Step 3: A computer randomly selects which curtain will have an image behind it.

Step 4: An image is shown to the participants from behind one of the curtains.

Result: The future position of the erotic images was selected significantly more than the future position of the non-erotic images

Experiment 2

The second experiment was titled ‘Precognitive Avoidance of Negative Stimuli’ and involved 150 undergraduate Cornell students. Before the experiment, each participant was involved in a 3 minute relaxation period. The experiment started with two identical yet mirror image pictures being shown on the computer screen. The participants were asked which picture (or rather which flipped orientation of the picture) they preferred. These pictures were not arousing pictures and were considered to be neutral. Following the participants’ preference selection, the computer randomly chose one of the two images to be the ‘target.’ When the participant chose the future target picture, a positive picture was subliminally flashed on the screen 3 times for an exposure period of 33ms. When the participant chose what the computer randomly determined to be the non-target picture, an arousing negative picture was subliminally flashed on the screen 3 times for a period of 33ms.

This experiment involved 36 trials. A ‘hit’ was considered to be selecting the future target which consequentially made the participant avoid the subliminal, negative picture. The hypothesis for this experiment was that the participants would select what would be the future target on significantly more than 50% of the trials. It must be emphasised that the computer determined the ‘target’ after the participant chose their picture and this was independent from the participants’ choice.

Results

Four different methods were used to analyse the results due to the nature of the experiment. Overall, the target was predicted by the participants on 51.7% of occasions. This was a ‘significant psi performance.’ The hit rate of low stimulus seekers was between 50.7% and 50.8%. This contrasted the hit rate of high stimulus seekers which was between 53.5% and 53.6%.

Summary of Experiment 2

Step 1: Two identical mirror image pictures are shown to the participants

Step 2: The participants select which image they prefer

Step 3: The computer randomly selects one of the two images

Step 4: If the computer selected image matches the participant selected image, then a positive picture is subliminally flashed on the screen

Step 5: If the computer selected image is different from the participant selected image, then a negative image is subliminally flashed on the screen.

Result: The future computer selected image was selected by the participants significantly more than the non-computer selected image

Experiment 3

In order to understand this experiment, it is important to be aware of psychological priming experiments. These experiments involve a positive or negative word flashing up on the screen. This is followed by a picture appearing on the screen. Participants are asked to judge as quickly as possible whether the image is a positive or negative image. Generally, respondents respond quicker if the original word on the screen is of the same nature as the image on the screen. For instance, if both the word and the image are positive, then participants will generally respond quicker than if the word is negative and the image is positive. If the word and the image are of the same nature, it is known as a ‘congruent trial.’ If the word and image are of different natures, then it is known as an incongruent trial.

The third experiment was titled ‘Retroactive Priming I’ and involved 100 Cornell undergraduates. Prior to the experiment, participants experienced a 3 minute relaxation period. The experiment was divided into two sections- the retroactive priming trial and the forward priming trial. The retroactive priming trial involved each participant being shown an image on their computer screen. The participants then had to indicate as quickly as possible whether the image was ‘pleasant’ or ‘unpleasant.’ Following their selection, a word would flash up on the screen. This word would be randomly selected by the computer after the participant had indicated the nature of the image. This word would either be a positive word such as ‘beautiful’ or a negative word such as ‘ugly.’ This first part of the experiment (retroactive priming) involved 32 trials.

The second part of the experiment featured the forward priming aspect. This involved the standard priming procedure during which a positive or negative word was flashed on the screen and then the appearance of an image. The participants then answered as quickly as possible whether the image was pleasant or not. This forward priming part of the experiment also featured 32 trials. In both portions of this third experiment, the participants’ response times were being measured and were of merit.

Results

Several methods were used in this study to analyse the response times. These methods are standard priming analysis methods were not post hoc methods used by the study authors. This first analysis method involved a 1.5 second cut off criteria (this excluded those trials in which the participant took longer than 1.5 seconds to select the nature of the image.) Using this method on the forward primes, the participants were 23.6 ms faster at answering congruent trials than incongruent trials. This is a typical priming result. The same method used on the retroactive primes yielded the result that the participants were 15.0 ms faster at responding to future congruent trials than future incongruent trials.

Using the 1.5 second cut off method, 64.9% of participants were faster at congruent forward priming as opposed to incongruent priming. 60.8% of participants were faster at congruent retroactive priming as opposed to incongruent retroactive priming. The individuals’ stimulus seeking level did not correlate with higher priming scores in either the forward or retroactive experiments.

Summary of Experiment 3

First Part of the experiment

Step 1: A pleasant or an unpleasant image is shown to the participant

Step 2: The participant indicates as quickly as possible whether the image is pleasant or unpleasant

Step 3: The computer randomly selects a negative or positive word and flashes it on the screen.

Result: Participants were significantly faster at indicating whether an image was pleasant or unpleasant if a corresponding (negative or positive) word was later shown to them.

Second part of the experiment

Step 1: The computer randomly selects a negative or positive word and flashes it on the screen

Step 2: A pleasant or an unpleasant image is shown to the participant

Step 3: The participant indicates as quickly as possible whether the image is pleasant or unpleasant

Result: Participants were significantly faster at determining whether an image was pleasant or unpleasant if a corresponding word (positive or negative) was shown before the image.

Experiment 4

The fourth experiment was titled ‘Retroactive Priming II’ and included 100 undergraduates for Cornell. This experiment was almost identical to experiment 3 with only one major and two minor differences. The minor differences involved slight changes to the length of time certain parts of the experiment appeared on the participants’ computers. The major difference involved the pairings of the words with their appropriate picture. In experiment 3, the word that appeared before or after each picture (depending on the part of the experiment) was random and didn’t necessarily have a specific relation to the picture. For instance, in experiment 3, a basket of fruit could have been shown and the priming word accompanying the basket of fruit could have been a random word such as one of; beautiful, ugly, friendly, threatening etc. In experiment 4, the priming word was appropriately paired to the image. For instance, the basket of fruit could have had only the positive priming word ‘luscious’ or negative priming word ‘bitter’ appearing before or after it. In this experiment (number 4) the computer randomly selected either the one negative priming word or one positive priming word to accompany each image. Other than these changes, experiment 4 was a replication of experiment 3.

Results

The results of this experiment were very similar to the results of experiment 3. Using the 1.5 second cut off criteria, the forward priming experiment resulted in participants being on average 27.4 ms faster on congruent primes than on incongruent primes. Using the same method, the retroactive priming experiment showed that the participants were 16.5 ms faster at answering congruent primes as opposed to incongruent primes. Like experiment 3, there was no correlation between being a stimulus seeking individual and the priming effect for either the forward or retroactive experiments.

Summary of Experiment 4

First Part of the experiment

Step 1: A pleasant or an unpleasant image is shown to the participant

Step 2: The participant indicates as quickly as possible whether the image is pleasant or unpleasant

Step 3: The computer randomly selects a negative or positive word that matches the image and flashes it on the screen. For instance a basket of fruit would match either luscious (positive) or bitter (negative.)

Result: Participants were significantly faster at indicating whether an image was pleasant or unpleasant if a corresponding (negative or positive) word was later shown to them.

Second part of the experiment

Step 1: The computer randomly selects a negative or positive word that matches the image and flashes it on the screen

Step 2: A pleasant or an unpleasant image is shown to the participant

Step 3: The participant indicates as quickly as possible whether the image is pleasant or unpleasant

Result: Participants were significantly faster at determining whether an image was pleasant or unpleasant if a corresponding word (positive or negative) was shown before the image.

Experiment 5

In order to understand experiment 5, it is imperative to understand what a standard ‘mere-exposure experiment’ entails. A mere-exposure experiment begins with a participant being shown a picture subliminally by it appearing on a screen for a short period of time at regular intervals. This image is known as the ‘habituation target.’ Following the exposure to the subliminal habituation target, two very similar pictures are shown on the screen statically and side by side. One of these pictures is the habituation target and the other is a similar picture to the habituation target.

The fifth experiment was titled ‘Retroactive Habituation I’ and included 100 Cornell undergraduates. Prior to the experiment, the participants were subject to a 3 minute relaxation period. This fifth experiment was a retroactive version of the aforementioned mere-exposure experiment. The participants in experiment 5 were shown two similar and static images that appeared on their computer screen side by side. They were asked to indicate which picture they preferred. After the participant selected the preferred picture, the computer randomly selected one of the two pictures that was now the habitation target. This habituation target was then flashed subliminally on the screen several times for a 17ms period.

Experiment 5 featured only negative pairs of images as well as neutral control images. The retroactive hypothesis was that the study participants would prefer and hence select what would in the future be randomly deemed the target image on more than 50% of occurrences.

Results

When negative picture pairs were shown (as opposed to the neutral and control pairs) the participants selected and hence preferred what would be the future target 53.1% of the time, which constitutes a significant result. Women selected the future target picture 53.6% of the time while men achieved this feat on 52.4% of occasions. The neutral and control pair targets were selected 49.4% of the time.

Summary of Experiment 5    

Step 1: Two similar images appear on the computer screen next to each other (either 2 negative pictures or 2 neutral pictures)

Step 2: The participants select which image they prefer

Step 3: The computer randomly selects one of the two pictures and flashes this image on the screen subliminally

Result: The participants selected the same negative image (from a pair of negative images) that the computer would randomly select significantly more than the same neutral image (out of a pair of neutral images) that the computer would select.

Experiment 6

Experiment 6, titled ‘Retroactive Habituation II’ was identical to experiment 5 except for 1 major and 2 minor changes. The major change involved the introduction of an erotic image pair trial. This changed the retroactive hypothesis to predict that participants would prefer the target picture on less than 50% of erotic trials. This was due to an erotic positive image being involved as opposed to experiment 5’s negative image. The first of the 2 minor changes involved an improvement on experiment 5 by determining if any image, comprising a pair had been preferred frequently. Consequentially, frequently selected images as part of pairs from experiment 5 were replaced with new images. The second minor change involved the introduction of gender specific negative and erotic images.

Results

Both of the retroactive hypotheses held up. The negative image aspect of experiment 6 resulted in 51.8% of participants retroactively selecting and hence preferring the future target image. The erotic image aspect of the retroactive study had a hypothesis that less than 50% of target images would be preferred. This hypothesis was fulfilled as 48.2% of target images were selected. The neutral, control target images were selected on 49.3% of occasions.

Summary of Experiment 6

Step 1: Two similar images appear on the computer screen next to each other (either 2 erotic pictures, 2 negative pictures or 2 neutral pictures)

Step 2: The participants select which image they prefer out of the two similar images

Step 3: The computer randomly selects one of the two pictures and flashes this image on the screen subliminally

Result: The participants selected the same negative image (from a pair of negative images) that the computer would randomly select significantly more than the same neutral image (out of a pair of neutral images) that the computer would select. The participants selected the same erotic image (from a pair or erotic images) that the computer would randomly select significantly less than the same neutral image (out of a pair of neutral images) that the computer would select.

Experiment 7

The seventh experiment was titled ‘Retroactive Induction of Boredom’ and included 200 Cornell undergraduates. Prior to the test, participants undertook relaxation for a 3 minute period. This seventh experiment involved two similar pictures appearing side by side on the participants’ computer screens. These image pairs ranged from mildly negative to mildly positive. The participants were asked to specify which image they preferred. After the selection was made, the computer randomly chose one of the images as the ‘target.’ This image was then flashed on the computer screen. Unlike the previous experiments, the image was flashed in such a manner as to make the participant conscious of the image being flashed. This flashing process involved the image being visible (and filling the entire screen) for a period of 0.75 seconds followed by a blank screen lasting for 0.25 seconds. This flashing process was repeated 10 times. Each participant partook in 24 trials of experiment 7.

The hypothesis for this experiment was that those participants within the high stimulus seeking group (which was also the high boredom prone group) would get bored of the same mildly positive and mild negative images and hence show a significant degree of dislike for the retroactive target image. This test design was based on a peripheral result from experiment 6- a result that suggested retroactive boredom may be due to repeated neutral stimuli.

Results

Due to the nature of this test, the hypothesis was that a target image selection percentage would be significantly less than 50%. Overall, across all participants, the hit rate was below 50% but not significantly. 49.1% of sessions involved the selection of the eventual target image. The participants who were deemed as high in stimulus seeking achieved a hit rate of 47.9% which was a significant result. The remaining participants achieved a hit rate of 50.1%.

Summary of Experiment 7

Step 1: Two similar images appear side by side on the participants computer screen

Step 2: The participant selects which of the two images they prefer

Step 3: The computer randomly selects one of the two images and flashes it on the computer screen in a manner that the participant is conscious of the image

Result: There was no significant difference between participants selecting the same image that the computer would select or selecting the image that the computer wouldn’t select.

Experiment 8

The eighth experiment was called ‘Retroactive Facilitation of Recall I’ and included 100 Cornell undergraduates. Prior to beginning the test, the participants undertook the standard 3 minute relaxation process. The experiment began with participants being shown a common noun for a period of 3 seconds. This process was repeated 47 times making the number of words each participant saw total 48. The participants were asked to visualise the physical manifestation of each word when each word appeared on the screen. It should be noted that the words came from 4 different categories; foods, animals, occupations and clothes. After the 48 words were shown, the participants completed a recall test during which they had to type all of the words that they could remember (regardless of order.)

Following the recall aspect of the test, the retroactive aspect of the experiment began. This involved the computer randomly selecting 6 words from each of the 4 categories (this was a random selection made from the 48 words on the original list.) The 6 words from the first category then appeared on the screen and the participants had to retype these words in an empty slot. This task was repeated for all 4 categories. In total, the participant had to type each of the 24 words. These 24 words were known as practice words.

   Results

Due to the nature of experiment 8, the results constitute a different format to the previous 7 experiments. All of the previous experiments involved a result figure in the 0% to 100% range, with 50% being chance. Experiment 8 involved a range from       -100% to 100%, with 0% being the chance score. A positive percentage shows that more practice words (the 24 words shown to the participants which they had to type out after they took the test) were recalled than non-practice words. The average score for all trials was 2.27%. This score supports the hypothesis that practicing a set of words after completing the test improves the ability to recall words in the original test. Those in the high stimulus seeking group scored a mean of 6.46%. This was compared to those in the low stimulus seeking group scoring at the chance level of -0.90%.

Summary of Experiment 8

Step 1: A word (a common noun) appears on the participant’s screen for 3 seconds

CAT

Step 2: This occurs with another 47 words

Step 3: The participant has to recall as many words as possible by writing them down (with no regard for their order.)

Step 4: The computer randomly selects 24 of the 48 words- these 24 words are known as practice words

Step 5: The 24 practice words appear on the participant’s screen and the participant types them out in an empty slot.

Result: The randomly selected practice words were recalled by the participant significantly more accurately than the non-practice words. This was despite the participant only dealing with the practice words after the recall test.

Experiment 9

The ninth and final experiment was titled ‘Retroactive Facilitation of Recall II’ and was almost identical to experiment 8. 50 Cornell undergraduates participated in this experiment. The only change to experiment 8 was the addition of a new practice exercise immediately following the recall test. This exercise involved the participants being shown a random selection of 24 of the 48 words. These 24 words comprised 6 from each of the 4 categories; food, animals, occupation and clothing. Each of these practice words was shown for a period of 3 seconds. The 6 food words were shown first (one by one) followed by the 6 animal words, the 6 occupation words and finally the 6 clothing words. It should be noted that this exercise of showing the participants the 24 practice words occurred after the participants tried to recall as many words as possible from the original list of 48 words that they were shown.

Results

Like experiment 8, the results from this experiment fall into the range from -100% to 100% with 0% being a change score. A significant positive score supports the hypothesis of retroactive recall. Overall, the mean score was 4.21% which was a significant result and supports the hypothesis that showing words after a recall test, enhances the recall of words in the original test itself. There was no significant difference between those participants who were high in stimulus seeking and low in stimulus seeking. High stimulus seeking participants achieved a mean score of 4.47% and those who were low in stimulus seeking scored a mean score of 4.09%.

Summary of Experiment 9

Step 1: A word (a common noun) appears on the participant’s screen for 3 seconds

DOG

Step 2: This occurs with another 47 words

Step 3: The participant has to recall as many words as possible by writing them down (with no regard for their order.)

Step 4: The computer randomly selects 24 of the 48 words- these 24 words are known as practice words

Step 5: The 24 practice words appear one by one on the participant’s screen for a period of 3 seconds

Step 6: The 24 practice words appear on the participant’s screen and the participant types them out in an empty slot.

Result: The randomly selected practice words were recalled by the participant significantly more accurately than the non-practice words. This was despite the participant only dealing with the practice words after the recall test.

Advertisements

4 comments

  1. Very fascinating. I like your presentation, although, I’d put the summary part first for each experiment.

    Like you, I’m skeptical of concrete claims of ESP, although ESP is just like UFOs: they certainly exist (via, say, skunkworks military), the public just doesn’t yet have well-accepted explanations of the mechanisms at play. As soon as we do, though, things cease to be “unidentified”. Similarly, “extra-sensory” becomes merely “sensory”.

    For all the experiments, important details are not presented; some of these details are likely missing from the original paper. What is the Left/Right (or whatnot) distribution of images? (The roll of a loaded die is still a random event.) Any bias or sequence “stickyness”? How many images were shown per trial? (Can’t state significance without this information.) Is the source truly uniform random? How well is the source isolated from experimenter/experimentee?

    If a hardware random number generator (based on radioactive decay, for example) is not used, the randomization is merely pseudo-randomization (google PRNG). It would then definitely NOT be a test of precognition. PRNG sequences, even the “cryptographically secure” ones, are deterministic for a given seed. Even with hardware generation, any cleaning/whitening techniques to correct bias become suspect, as does physical breakdown of the hardware. Saying that the “apparatus did not make its choice until after the user” cannot be asserted without real details of the hardware. The output of a HW generator at time T may be very much dependent (and deliberately so) on many prior states of its physical random process.

    The erotic/non-erotic (and positive/negative, preferred/non-preferred) differences seem compelling, however, because the results would (for most experiments) still stand even if there were weaknesses in the source of randomness. In fact, an interesting extension of these experiments, away from controversial claims of ESP, would be to use a variety of deliberately non-random sources for how the “apparatus” makes its choices, and then see how various factors affect the test subject’s performance at detecting a useful pattern in the apparatus. Why is it that sex–or positivity/negativity, or (particularly) $$$–has such a influence? Is the effect consistent across cultures/genders/educational-backgrounds/etc.? And so on.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s