Natasha Demkina

Natasha Demkina came to notice in Russia in 2004 for saying that she could look at a living person and see the state of the organs inside them, a claim that received considerable coverage and led to invitations to be tested. She agreed to a test in New York, arranged with a sceptics organisation and a research body and recorded for television. The design was settled with her before it began, which is the feature that makes the case worth recording.

Seven volunteers took part, each with a documented medical condition, together with one who had none, and the task was to match each condition to the right person. Because a person can be matched by chance, a threshold was fixed in advance for what would count as a result supporting the claim, and the number set was five correct matches out of seven. She made four. The atlas records the arrangement in that order, the claim first, the design second, the threshold third and the outcome last, because the sequence is what makes the episode interpretable at all, and it is the sequence that is usually lost when the case is retold.

Two things about the arrangement are worth separating from the outcome. She agreed to be tested, which claimants of this kind frequently decline to do, and the terms were negotiated with her rather than imposed. And the task chosen was one her claim entailed: if a person can see the condition of the organs, then matching a documented condition to the body it belongs to is exactly what the ability would permit. A test that asks for something the claim does not assert proves nothing whichever way it comes out, and this one did not make that mistake.

Four out of seven is a number that supports two readings and it received both. For the investigators the answer was simple: a threshold had been agreed by everyone before the test, it was not reached, and a test whose criterion is set in advance is worth nothing if the criterion is renegotiated afterwards. For her supporters, four correct matches out of seven is well above what guessing alone would be expected to produce, the task was harder than a straightforward yes or no judgement, and the conditions, including a long session and an unfamiliar setting, were not those in which she ordinarily worked.

Both observations are true, and the case is a clean illustration of a problem that is not peculiar to it. A single test with a small number of trials has little power to distinguish a modest real effect from chance, so a threshold set high enough to be convincing if met is also high enough that failing it proves less than it appears to. What was not agreed in advance was the meaning of a near miss, and that is the gap both sides then occupied. The atlas records the design, the number and the two readings, and does not resolve them.

The episode is also a reminder of what a single test can and cannot do. One session with one person establishes nothing durable in either direction, and the standard remedy, repeating the test with more trials under the same agreed conditions, was not carried out. The case therefore sits in the literature as an unrepeated result that each side describes accurately and reads oppositely, which is a familiar and unsatisfactory position for a claim of this kind to occupy.