Thursday, September 25, 2008
I do an interview for a documentary film or TV show typically once or twice a month. Occasionally the director wants to film a demonstration of a test in the lab, and one of the options I offer is a presentiment experiment using a skin conductance level (SCL) measurement. About half the time the director or producer decides to do the test him or herself.
I noticed some years ago when I started doing these demo experiments that they often resulted in pretty good results, which I found surprising because the set and setting of interviews are not nearly as well controlled as actual experiments. But starting this year, out of curiosity, I decided to keep a closer eye on the outcomes of those demos.
The image here shows the composite results after running four presentiment demos this year, so far. This is not based on a selection of results -- this is all the data from the four demos. There was one demo per interview, and each demo consisted of 40 randomly selected images out of a pool of about 700 images, for a total of 160 targets of widely varying emotionality. The trials were partitioned based on a median split of the emotionality ratings of the 160 trials, so the 80 targets rated as being most arousing (based on the international standard rating per image) were defined as emotional, and the bottom 80 target were defined as calm.
Each trial was normalized to allow the emotional and calm trials to be easily pooled across all four sessions, so the y axis is in terms of z scores (standard normal deviates). The difference in the two curves prior to the stimulus (i.e., the moment the target is selected and shown, at second 0) is not quite statistically significant, mainly because with 160 trials there is insufficient statistical power. But the differential effect is in alignment with the results that I and others have obtained in previous experiments.
What it shows is that about 3 seconds before the stimulus is randomly selected, the SCL measurement begins to differentiate according to the future target selection. A few people have misinterpreted what's going on here. What presentiment predicts is that prior to a randomly selected emotional target SCL should become more aroused than prior to a randomly selected calm target. This only means that the ensemble emotional curve is predicted to be higher than the ensemble calm curve, and not that the emotional curve will necessarily go into positive territory.