Scientific misconduct is a big problem in India. Since 2006, the rate of Indian papers that are retracted for misconduct has been double that of the US. Misconduct gives students the wrong idea of what kind of work is acceptable work, diminishes trust in India’s research output, and tarnishes the reputation of our researchers worldwide.
At its core, misconduct arises when scientists are able to control the outcomes of their experiments, instead of letting the scientific method take its course. But there is one experimental design that works with scientists to bring their best selves out. This is the blind test.
What is a blind test?
My students Felicite K. Noubissi, Ashwin Bhat, and Prakash Arumugam performed a blind test to show that the Adiopodoumé strain of the fungus Neurospora crassa suppresses a biological process called RIP.
RIP is a mutational process that peppers multiple mutations into all copies of any DNA sequence that appears more than once in the genome. RIP typically occurs during a sexual mating between two fungal strains. Yet of the hundreds of Neurospora strains isolated from around the world, only the Adiopodoumé strain from Côte d’Ivoire, Africa, contains a DNA sequence that appears in several copies in the genome, without being disrupted by RIP’s effects.
The students asked whether the fungus’s DNA was this way because the Adiopodoumé strain suppressed RIP.
Ms. Noubissi grew Neurospora strains on 19 petri dishes. On five dishes, she grew the Adiopodoumé strain, and on the other 14, other wild strains. She then handed the dishes over to Mr. Bhat, who – unobserved by anyone else – picked three Adiopodoumé dishes and seven other dishes, and relabeled them 1,2, 3, and so on until 10. He then handed them to Mr. Arumugam. Again unobserved, Mr. Arumugam re-relabeled them A, B, C, and so on until J, and finally sent them back to Ms. Noubissi.
This way, there were two sets of labels for the 10 dishes: one with Mr. Bhat, indicating which strains were numbered 1,2,3, …., and the other with Mr. Arumugam, indicating which of the plates 1,2, 3, …. were relabeled A, B, C, ….. They wrote down which dishes had been labelled what on pieces of paper and sealed them in an envelope. (One group stuck the envelope to the lab’s ceiling.)
Ms. Noubissi knew only that at least one culture was Adiopodoumé, but no one knew which plate carried which culture. Her task was to determine which dishes labeled A, B, C, … contained the Adiopodoumé strain. The way to do this was to cross the strains in the dishes with each other and test for RIP suppression.
Through a process of elimination, she correctly found that the dishes labelled B, C, and E held the Adiopodoumé strain.
(The students have since achieved their PhDs.)
How do blind tests help
Some years later, my group published a paper reporting that Adiopodoumé’s RIP-suppressing behaviour was the result of a gene called upr-1. We found this based on unblind experiments. We had to replicate the results using a blind study, to be doubly sure.
In this experiment, the investigator asked whether deleting Adiopodoumé’s upr-1 gene caused the loss of RIP suppression. In the unblinded experiment, the investigator knew which strains had the deletion and which ones the intact gene, but in the blind experiment she didn’t. And the blind experiment told us that the upr-1 gene was not responsible. Our previous finding was wrong and we had to issue a correction to our paper.
The advantage of conducting a blind experiment is that it makes the knowledge you have incomplete, forcing you to work around the resulting gaps with more rigour in order to arrive at the (experiment’s) truth. The moment you have a key – i.e. a way of labelling a set of things on which you have no control – you also have an individual disinterested in the result, and therefore not invested in biasing it. If you create more keys, you can also create more disinterested individuals.
The principal investigator of the experiment performs all the necessary tests, but now they don’t know the provenance of the samples they are testing. This information lies in a sealed envelope that can be opened only once she has finished all measurements. If the principal investigator was corrupt, it would have been hard to corrupt the others as well, at least without outing them as corrupt. In this way, blind experiments can deter the ‘bad eggs’ from taking short cuts while giving the ‘good eggs’ the benefit of additional oversight of their efforts.
Conducting blind experiments can be fun – with the opening of the envelope reminiscent of the moment a Nobel Prize or an Academy Award is announced. Such ‘rituals’ and the resulting celebrations foster camaraderie and trust. Perhaps the University Grants Commission should encourage all PhD candidates to include at least one significant blind experiment in their thesis research.
D.P. Kasbekar is a retired scientist.
