Saturday, July 11, 2015

We cannot really smell a trillion odors

Slashdot reports:
Last year a paper in Science magazine reported that humans can distinguish a trillion different odors, a result that had already made its way into neuroscience and psychology textbooks. Two new papers just published in eLife overturn that result, pointing to fatal flaws in experimental design and data analysis.
I suspected that the trillion odors were bogus, because it seems unlikely and because it is hard to see how a test could confirm it.

Comparing to colors, I can easily imagine showing someone a bunch of random colors, confirming that he can distinguish them, and deducing that trillions of colors are distinguishable. Likewise with musical sounds. But if the data passes thru some low dimensional filter, then there will be trillions of different inputs that are perceived as identical. Unless the experiment is set up to produce these examples, they will be missed.

I did not read the papers, but it is amusing how such a completely bogus result can be reported in the most prestigious science journals, and widely reported as fact in the popular press.

No comments:

Post a Comment