Brain-Zapping Neuroscientist Experimenters Play "Keep Fiddling With Convoluted Models"

 A recent paper claims to have increased altruistic behavior by brain-zapping. It is a paper entitled "Augmentation of frontoparietal gamma-band phase coupling enhances human altruistic behavior" which you can read here. A BBC article describes the experiment:

"Scientists have discovered how to make people less selfish - at least temporarily - by stimulating two areas of their brain. In a new study carried out at the University of Zurich, 44 volunteers were asked to decide how to split an amount of money between themselves and an anonymous partner.

During the experiment, an electrical current was applied to the frontal and parietal areas of the brain - situated at the front and towards the back. When these areas were stimulated at the same time, the participants gave away more money."

Now, you might expect that at this point in the post I will chime in my very common complaint about neuroscientists using ridiculously too-small study group sizes. But unlike the great majority of cognitive neuroscience experiments, this experiment had a fairly decent size of 44 subjects. My main complaint is the experimenters fiddling with what they call "models" in a way that can be described as "keep torturing the data until it confesses."

We read of three analytic models used by the experimenters, which they describe as examples of "post hoc analysis." Presumably these were not models created before the experiments were done, but models of analysis created after gathering data, by scientists looking for evidence of some desired effect. We read of a Model 1 (M1) that is described in the screen shot below from the paper:

There is no clear reason stated why anyone should accept these weird equations being a correct analysis of whether the subjects were acting in an altruistic way. What's going on in this model is actually far more complicated, because we read other passages of statistical gobbledygook describing how the model is computed, such as this little section which sounds like "keep torturing the data until it confesses":

The experiments also came up with Model 2, and a much more convoluted Model 3 (M3), part of which is shown in the screen shot below from the paper:

The mention of "50 multiple starting  points" in one of the passages shown above makes it sound like really 50+ models were tried, rather than just 3. Just because we have a mention of a Model 1 (M1), a Model 2 (M2), and a Model 3 (M3), that's no reason to think that only three models were tried. 

We read that one of these models was declared to be the "winning model." This judgment of one model winning over the other apparently occurred because one model produced results with higher statistical significance. This "keep torturing the data until it confesses" way of doing science is profoundly dysfunctional. An intelligent way to proceed would be (1) do a pre-registered study or "registered report" in which experimenters commit themselves (before gathering any data) to using one and only one way of analyzing data -- a simple and straightforward method -- while justifying that method of data analysis; (2) publish whatever results are produced from that analysis method, regardless of whether a mere null result is produced. 

We read the following chilling statement in the paper: "As we stimulated each participant with an electric current intensity in line with his/her tolerance level for the stimulation currents tested before the experiment runs, the peak-to-peak current intensity for different participants varied between 2.4 and 4 mA." So this was apparently no minimal brain zapping that was occurring, but some level of zapping approaching the subject's "tolerance level," which apparently was somehow determined by some previous "push-them-to-near-their-limits" deal finding out how much brain zapping the subjects could take or would allow. For these absurd machinations in which the student's health may well have been endangered, each student was paid a "chump change" amount of only about 60 dollars, along with some "bonuses" that probably were just as measly. 

Nothing has been done here that produces any robust evidence that altruism can be altered by brain zapping.  What has mainly gone on is that students have had their health risked by brain zapping done in a "keep torturing the data until it confesses" methodological mess that fails to follow good standards of doing a neuroscience experiment.