>> If I understand correctly, the participants rated their experience of these three factors on a scale of 0 (strongly disagree that I had such an experience) to 5 (strongly agree). Now if we look to see the scores they gave for how much dream insight, dissociation and control they had, we find that the averages for the gamma stimulation condition are around 0.6, 1.3, and 0.5 respectively.
>
> That’s a pretty low score, even if it’s higher than it would’ve been without the brain stimulation.
Maybe, maybe not. Averages work for some things, not so much for others. Suppose 4-5 are no-fooling-definitely-lucid dreams, and 1-3 are questionable. If the average is being driven by a few dreams falling into '5' and otherwise scores are unchanged, you'd get a worthwhile effect but only a small shift in average on your 1-5 scale. I'd want to read http://www.blogs.uni-mainz.de/fb05philosophie/files/2013/04/... more carefully looking at how the overall rating distribution looks split by condition and whether the rating scale is masking a binary effect.
Awfully, the standard deviations of almost every item in their validation study [1, Table 1] is greater than the means --in some cases even twice as big--, which themselves are very low (the majority are below 2, and several are below 1). This is, at a minimum, strongly indicative of bimodality --which would make sense since, among other things, participants were drawn mostly from Bonn University's lucid dreaming student club [p.11]. As a matter of fact, they report they threw away two data sets (!) precisely because of "extreme answering style (all items were scored as 0 or 5)" [ibid].
Why they didn't heed these warning signs (and at the very least use medians instead of means from then on) is beyond me.
Maybe, maybe not. Averages work for some things, not so much for others. Suppose 4-5 are no-fooling-definitely-lucid dreams, and 1-3 are questionable. If the average is being driven by a few dreams falling into '5' and otherwise scores are unchanged, you'd get a worthwhile effect but only a small shift in average on your 1-5 scale. I'd want to read http://www.blogs.uni-mainz.de/fb05philosophie/files/2013/04/... more carefully looking at how the overall rating distribution looks split by condition and whether the rating scale is masking a binary effect.