Cited research: Neuron 66, 610–18 (2010)

To be able to recognize the source of a sound — a flute or a train, say — we first have to learn that sound's acoustic characteristics. Scientists in France report that such learning is subconscious, rapid and remarkably robust.

Trevor Agus at the École normale supérieure in Paris and his colleagues asked 12 volunteers to listen to a series of short snatches of random noise, and to identify which of the snatches included repetitions. Their success rate was below 50%.

But when one specific snatch containing repetition was secretly and randomly interspersed at frequent intervals, the listeners quickly became near-perfect in detecting repetition — even when the snatch was compressed in time or reversed. Moreover, they retained the memory for several weeks. A.A.