A wise man named Karl Popper once disagreed with the orthodox view that scientific activity starts with observation noting that “observation is always selective”. It is often possible to propose a theory and find results that support or verify it but Popper proposed that it is the negative results that are of crucial value. He used the following simple thought experiment to prove it.

A black swan disproves the theory that all swans are white

Europeans for thousands of years had observed millions of white swans. Using inductive evidence, we could come up with the theory that all swans are white.

However exploration of Australasia introduced Europeans to black swans. No matter how many observations are made which confirm the theory that swans are white there is always the possibility that a future observation could refute it. Induction cannot yield certainty.

This simple principle revolutionised science and the world lived happily ever after.
…Of course that was not how the story ended. Today there is still massive pressure on scientists by industry, funding bodies, the media and universities to chase after verifying positive findings and in effect supporting established knowledge. This is all very well but when a negative finding is discovered it is all too often brushed under the carpet, almost as if it is an embarrassed (an ugly duckling perhaps). For example consider for a second how truly horrifying the following statistic actually is:

Only 5.9% of industry sponsored cancer trials are ever published

Of that 5.9%, an astounding 75% give positive results – suggesting negative findings are simply not published. This is not the only problem for academics searching for negative results to support a proposition. Because of the way boolean search algorithms work you have to have a little bit of boolean-know-how to actually search for a negative result. Simply adding “not” in to an english language proposition will still yield positive findings. Now a group of scientists have created a database called BioNot that uses data mining and intelligent machine learning methods to systematically search for negative results in Pubmed and Elsevier.

You may wish to write down the URL http://snake.ims.uwm.edu/bionot/ because in an ironic twist the boolean wizards who created this programme have made the URL a little trixy for google to find, presumably because this website is google’s arch nemesis. Perhaps not, but either way if you hit “BioNot” in to google you end up with something to do with nuts.

Anyway, I’m off to have a play with it, if you need some search ideas, check out the fabulous results of my very first search!

Full Text (Open Access PDF) on Pubmed

Via Neuroskeptic (where you can pop on over and read a little more explanation of the potential applications of this tool)

Agarwal S, Yu H, & Kohane I (2011). BioNOT: A searchable database of biomedical negated sentences. BMC bioinformatics, 12 (1) PMID: 22032181