(2018-02-18, 09:54 AM)Chris Wrote: [ -> ]A third version of the preprint is now available:
https://osf.io/zsgwp/
I have read only the abstract so far, which says that based on the previous studies (now treated as exploratory), a further study of 80 sessions was pre-registered, to be tested using a directional hypothesis. But the results were not statistically significant. However, a post hoc bi-directional hypothesis gave significant results for the 80 sessions (2.75 sigma) and for all 240* sessions to date (4.73 sigma).
Clearly, if this is a real effect, there is a difficulty in predicting its direction. This was the case with previous versions of the preprint, where the effect appeared to operate in the direction opposite to the one expected on the basis of model calculations [though in some cases I'm not convinced the expected direction had been calculated correctly]. The abstract says that 240 control sessions, conducted without an observer present, continued to give non-significant results on both directional and bi-directional hypotheses. So despite the difficulty in predicting the direction, it's still possible there's a real effect there.
(* Edit: Apparently this figure actually relates to 180 sessions, because the first experiment of 60 sessions, which was used to optimise the variables V1 and V2, was excluded.)
Having read the new version, I'm not really any clearer about what's going on. The four new experiments are based on the final two in the previous version, which were the ones that gave the strongest results.
In these experiments, the feedback given to the participants would have encouraged them either to increase or decrease a variable based on the amplitudes of the first few terms in a Fourier series representation of the measured interference pattern (that is, the amplitudes of the wave-like components with the longest wavelengths, reflecting the large-scale shape of the interference pattern). But the response was analysed using two different variables, V1 and V2, which are defined in terms of -the phases (i.e. the sideways shifts of the wave-like components) for two ranges of smaller wavelengths (intermediate between the lengthscale of the overall interference pattern, and the wavelength of the interference fringes).
As the feedback variable is different from the variables used to analyse the response, it's not necessarily unexpected that the response could vary in direction. It might be that there were different ways in which the pattern could change, in which the signs of the phase shifts were different, but the changes in the amplitudes of the leading terms were in the same sense. (Guerrer used a theoretical model for motivation, which predicted the sign of V1 - though not of V2 - for given feedback. But as the predicted sign in the final two exploratory experiments was the opposite of that measured, the model doesn't seem to reflect what's happening.) Indeed, as the double-slit system is nearly left-right symmetrical, it's not hard to imagine that the pattern could change in such a way that a change in the sign of the phase shifts (and thus of V1 and V2) could leave the sign of the amplitudes (and thus of the feedback variable) unchanged.
But we're still left with a situation in which, for example, in experiment 5 (20 participants) both V1 and V2 show very significant decreases (Z=-3.15 and -2.87), whereas in experiment 9 (also 20 participants, with the same feedback), they both show significant increases (Z=2.23 and 3.00). Perhaps it could be that these significant results in opposite directions are produced by only a small number of participants, which might make them statistically reconcilable. Perhaps it could be some kind of experimental artefact occurring in only a small number of sessions, but it's not easy to imagine how it would work.
One other interesting finding is that pooling experiments 5, 7 and 9 (which share the same feedback), there is a very significant correlation between the observed changes in V1 and V2 (p=0.01). But for experiments 4, 6 and 8 (in which the feedback acts in the opposite sense), the changes in V1 and V2 are significantly negatively correlated (also p=0.01). Again, given the approximate left-right symmetry of the system, it's not too hard to imagine that the directions of steepest (1) increase and (2) decrease of the feedback variable could correspond to the changes in V1 and V2 having (1) the same and (2) opposite signs. But it's hard to see why this change in the correlation according to the sense of the feedback should occur if this is an experimental artefact rather than a psi effect.
Perhaps more could be gleaned from a closer look at the experimental data. It's good that the data can be downloaded from the same OSF website where the preprint is hosted.