Wednesday, 26 September 2012

The Bem Experiment.

Blogger Reference Link

ESP evidence fails key test of repeatability

Peter Aldhous, San Francisco bureau chief
We cannot see the future. That's the conclusion from a meta-analysis of studies attempting to replicate astonishing findings on "precognition"  - which suggested that people's behaviour is influenced by events that haven't yet happened.
The original findings, revealed in 2010 by Daryl Bem of Cornell University in Ithaca, New York, would have turned established views of causality and human perception on their head. New Scientist's bet at the time was that most attempts to repeat the experiments would fail to replicate Bem's results.
Bem's experiments were disarmingly simple: he ran well-established psychological experiments - for instance those showing that typing selected words from a presented list facilitates their later recall - in the reverse order. In this case, student volunteers were better at remembering words that they would later type.

Several groups have reported previously on their attempts to repeat Bem's work, with most failing to replicate his findings. In the new paper, a team led by Jeff Galak of Carnegie Mellon University in Pittsburgh not only conducted their own experiments on the word-recall tests, but also ran a statistical analysis combining results from their work, Bem's studies and 10 other experiments.

That's important, because such meta-analyses are considered the best approach when assessing a body of scientific evidence - used, for example, to combine the results of multiple clinical trials to determine whether a medical treatment works.

The Journal of Personality and Social Psychology, in which both Bem's research and the new paper appeared, had earlier been criticised for failing to publish a failed attempt to repeat Bem's work. At the time, journal editor Eliot Smith, a psychologist at Indiana University in Bloomington, said he was open to publishing a meta-analysis.

"It seems that the normal practices of scientific research and journal publication can effectively correct claims that turn out to be incorrect or overstated," wrote Smith in an email to alert reporters to the new paper.

"An effect is not an effect unless it is replicable," conclude Galak and his colleagues, who nevertheless praise Bem for encouraging other researchers to repeat his work.

Facebook iconDigg iconDelicious iconStumbleUpon iconTwitter icon

No comments:

Post a Comment