Rather than having the students read each others' minds or wear sliced ping-pong balls over their eyes, Daryl Bem has taken the unusual, yet elegantly simple, approach of testing a raft of classic psychological phenomena, backwards.
Take priming - the effect whereby a subliminal (i.e. too fast for conscious detection) presentation of a word or concept speeds subsequent reaction times for recognition of a related stimulus. Bem turned this around by having participants categorise pictures as negative or positive and then presenting them subliminally with a negative or positive word. That is, the primes came afterwards. Students were quicker, by an average of 16.5ms, to categorise negative pictures as negative when they were followed by a negative subliminal word (e.g. 'threatening'), almost as if that word were acting as a prime working backwards in time.
If psi abilities have really evolved, it makes sense that they should confer survival advantages by helping us find mates and avoid danger. In another experiment Bem had dozens of undergrads guess which set of curtains in a pair on a computer screen was concealing an erotic picture. Participants were accurate on 53.1 per cent of trials, compared with the 50 per cent accuracy you'd expect if they were simply guessing. This accuracy was increased to 57 per cent among students who scored higher on a measure of thrill-seeking. By contrast, no such psi effects were observed for neutral stimuli.
In another experiment participants looked at successive pairs of neutral mirror images and chose their favourite - the left or right. After each pair, an unpleasant picture was flashed subliminally on one side or the other. You guessed it, participants tended to favour the mirror image on the side of the screen opposite to where an unpleasant picture was about to appear.
The examples keep coming. The mere exposure effect is when subliminal presentation of a particular object, word or symbol causes us to favour that target afterwards. Bem turned this backwards so that participants chose between pairs of negative pictures, and then just one of them was flashed subliminally several times. Female participants tended to favour the negative images that went on to be flashed subliminally, as if the mere exposure effect were working backwards through time.
This backward mere exposure effect didn't work for male undergrads, perhaps because the images weren't arousing enough, so Bem replicated the experiment using more extreme negative images and erotic images. This time a 'backwards' mere exposure effect was found with men for unpleasant images. For positive imagery, mere exposure traditionally has a negative effect, as the stimuli are made to become more boring. Bem showed this effect could also happen from the future. Presented with pairs of erotic images, male undergrads showed less favour for the images that went on to be flashed subliminally multiple times. It's as if the participants knew which images were going to become boring before they had.
Finally, we all know that practice improves learning. Bem tested students' memory for word lists and then had them engage in extensive practice (e.g. typing out) for some of the words but not others. His finding? That memory performance was superior for words that the students went on to practice afterwards - a kind of reverse learning effect whereby your memory is improved now based on study you do later.
These reverse effects seem bizarre but they are backed up by some rigorous methodology. For example, Bem used two types of randomisation for the stimuli - one that's based on computer algorithms, which produce a kind of pseudo-randomisation in the sense that a given distribution of stimuli is decided in advance. And another form of randomisation based on hardware that produces true randomisation that unfolds over time as an experiment plays out. Also throughout his paper, Bem uses multiple forms of simple statistical test and he reports results for each, thus demonstrating that he hasn't simply cherry picked the approach that produces the right result. Across all nine experiments the mean effect size for the psi effects was 0.22 - this is small, but noteworthy given the nature of the results.
So what's going on? Bem doesn't proffer too many answers although he argues that his psi phenomena vary with subject variables, just like mainstream psychological effects do. For example, the phenomena were nearly always exaggerated in the more extravert, thrill-seeking participants. From a physics perspective, he believes the explanations may lie in quantum effects. 'Those who follow contemporary developments in modern physics ... will be aware that several features of quantum phenomena are themselves incompatible with our everyday conception of physical reality,' Bem argues. 'Many psi researchers see sufficiently compelling parallels between these phenomena and characteristics of psi to warrant considering them as potential candidates for theories of psi.'
Daryl Bem (2010). Feeling the future: Experimental evidence for anomalous retroactive influences on cognition and affect. Journal of Personality and Social Psychology. In Press PDF.
UPDATE, plucked from the comments:
Failures to replicate this study: 1, 2, 3
A successful replication.
A criticism of the stats methods used.
A flaw in the methodology.
A registry of replication attempts.