6.37 sigma replication of Dean Radin's double slit consciousness experiments

Yes, that was why I contacted him. He'd evidently had other feedback on the preprint before that, though. My impression is that he is very receptive to constructive criticism.

Can you share what these concerns were?

Can you share what these concerns were?

The problem is that there's a lot in the preprint to understand, and I don't think I understand enough to judge it fairly. The revisions he had in mind would address the essential concern, though, so I think the best thing to do will be to wait for the revised version.

The problem is that there's a lot in the preprint to understand, and I don't think I understand enough to judge it fairly. The revisions he had in mind would address the essential concern, though, so I think the best thing to do will be to wait for the revised version.

If the point is that it was too technical and perhaps even a bit convoluted, I agree. It took me a few hours to get the gist of enough to feel comfortable commenting on it, and there are still things there that seemed above my pay grade... But, then again, I am not a physicist and they are most likely the target public.

A revised version of the preprint is now up:
https://osf.io/zsgwp/

[Laird]

A revised version of the preprint is now up:
https://osf.io/zsgwp/

Does it address to your satisfaction the statistical issues you raised? (I haven't made the time to read it in either form just yet)

Does it address to your satisfaction the statistical issues you raised? (I haven't made the time to read it in either form just yet)

Well, I've been slowly reading through Radin's double-slit papers, and haven't quite got to the end yet. But I am a bit clearer about things now. After a quick look at the revised version, a lot has been done to address the previous issues, and it seems reasonable as far as the V1 variable (which produces a Z value of 3.43), but I'm still not convinced about the procedure by which the V2 variable is added (or subtracted), to produce Z=4.68.

(V1 is related to the ratio of the amounts of light coming through the two slits, whereas V2 is related to a phase difference between them.)

There is still the puzzling feature that the V1 effect is in the opposite direction from the one expected.

Well, I've been slowly reading through Radin's double-slit papers, and haven't quite got to the end yet. But I am a bit clearer about things now. After a quick look at the revised version, a lot has been done to address the previous issues, and it seems reasonable as far as the V1 variable (which produces a Z value of 3.43), but I'm still not convinced about the procedure by which the V2 variable is added (or subtracted), to produce Z=4.68.

(V1 is related to the ratio of the amounts of light coming through the two slits, whereas V2 is related to a phase difference between them.)

There is still the puzzling feature that the V1 effect is in the opposite direction from the one expected.

Can you explain why Guerrer's replication is now 4ish sigma and no longer 6 for me Chris?

Just to explain very briefly what the concerns were -

In the previous experiments, Radin et al. used a number of different measures to characterise the effect of attention/concentration on the interference pattern. They were always partly based on the magnitude of the oscillations in the wavelike pattern of interference fringes. In some studies there was also a contribution from the phase of the oscillations (which relates to shifts in the location of the fringes).

Gabriel Guerrer's analysis is based entirely on the phase. The interference pattern is Fourier transformed, to break it down into a combination of a lot of sine waves with different wavelengths, each with a magnitude and a phase. Guerrer defines several variables called V, by adding up the change in the phase between concentration and relax conditions, over certain ranges of wavelengths. (In the original version the analysis used three ranges of wavelengths, and three resulting variables - V1, V2 and V3. In the revised version there are only two - V1 and V2.)

One problem with the original version was that the ranges of wavelengths were determined by an optimisation procedure. The ranges were chosen so as to maximise the difference between the concentration and relax conditions. But that really amounts to trying multiple hypotheses and picking the one that gives the most significant result, so it will tend to inflate the significance. In the revised version, the ranges are determined from the data for the first of the five experiments, and those data are excluded from further analysis, which solves that problem.

But another problem relates to how V1 and V2 should be combined together to form a single measure of the differences between the pattern in the concentration and relax conditions. The snag is that there are five different experiments, and the feedback given to the participants is differently defined in each. Sometimes the feedback tells the participants they're doing well if the amount of light going through the left-hand slit increases, but sometimes it encourages them to boost the amount going through the right-hand slit. That will tend to have opposite effects on the phase shifts. So, for example, V1 can be expected to be positive in some experiments, but negative in others.

Now, in the paper, there is a table in which the expected sign of V1 is worked out for each of the five experiments from the form of the feedback, using a simple model. So on that basis there is rational way of working out how to combine together the values of V1 for the different experiments. That seems reasonable (assuming those predicted signs of V1 have been worked out correctly). But then V2 is added in, and it still seems to me that there's a problem with this step, because there isn't a rational way of choosing the sign of V2. 

But as I say, I have only had a very quick look at the paper so far, so I may well be missing something.

Can you explain why Guerrer's replication is now 4ish sigma and no longer 6 for me Chris?

Yes - continuing from my last post, I can see three reasons why the Z value will be smaller. First, the contribution from the variable V3 (which was also problematical) has been removed. Second, for the remaining variables V1 and V2, their defining ranges are no longer being optimised in the way they were originally, and that also reduces the significance of the results. Third, the data for Experiment 1, which are now being used to define those ranges, are excluded from further analysis, so the overall result is based on a smaller amount of data, which also tends to reduce the significance.

(V1 is related to the ratio of the amounts of light coming through the two slits, whereas V2 is related to a phase difference between them.)

Sorry, if anyone is trying to make sense of my comments, I realise that probably looks very confusing.

What I mean is that - why V1 and V2 are both made up from changes in phase of the components of the interference pattern - V1 comes from a range of wavelengths for which the changes in phase are mainly supposed to reflect a change in the ratio of amounts of light coming through the two slits, while V2 comes from a different range of wavelengths, and is supposed to reflect a phase difference between the light coming through the two slits.