- No Recommendations yet - go add some!
- A lot of people have left what sound to me to be very plausible explanations for this "phenomenon". A good meta analysis of randomly selected topics would show whether this "decline" is to be expected statistically or is a signal in its own right. After all, this podcast really focused on anecdote and was highly biased in looking for similar examples.
Still, I actually wouldn't be surprised if the effect was real. My theory, which is only slightly different than one stated several times in the comments, starts off with the publication of a strong result because weak results with small N aren't significant and won't be published. Then, while a few others may seek to disprove the results, most (at least in my experience in neuroscience) will seek to confirm the results in order to explore how they happened or expounding upon the theory in some way. Since they are looking for confirmation, if they get a weak result relative to the original publication, they may figure they did something wrong and so they go back and try again. They either just don't publish the weak data/result (possibly doing a new experiment with some small methodological changes) or increase their N until the result is stronger. A few might publish results that are unbiased. However of all of the results published in the first wave following the original publication, the results will be more modest. On the second wave, it will not seem as unexpected to get an even more modest result, leading to more data/results being published that are closer to the true mean than before. This would continue until the mean has been reached.
So ultimately what we'd have would be regression to the mean, but filtered through the process of human implementation of the scientific method that produces an *apparent* decline.
The nice thing about this is, even though it takes a few years, the truth of the matter is ultimately found through the process. One could possibly confirm this via meta-analysis by choosing some random selection of new phenomena, getting in touch with any lab that might follow up on the experiment, and then somehow (and this would be the tricky part) get them to donate all data (including data that isn't published for whatever reason). If this trend were not happening, then you'd expect the data that gets thrown out to be both high and low relative to the mean. If the trend were happening, you'd see a decline in difference between the mean of the discarded data and the reported data, and you'd see a pure regression to the mean in all of the data provided for the meta-analysis. Any sociologist and philosophy of science takers?
- With regards to some of the comments about quantum mechanics (specifically the double-slit experiment), and in case this comment hasn't already been made, it's not necessarily about whether you observe a particular particle/wave, but whether you know enough about the *system* to know if it took one path or another. If you had a photon (or electron or other particle) that had a partner, and if determining the path of the partner allowed you to determine the path of the original, then you could force (or not) the original to take a single path by measuring the path of the partner. This makes the issue of observation a bit more general, and makes me wonder about whether there is some coupling effect between systems that is responsible for "collapse". Is it possible that we only notice collapse because we are now joined as a system, and that any other system NOT coupled to ours would somehow see a different result or collapse of the system? Now I'm speculating heavily :)
