Can the brain learn a new sense?

In a 2012 study, seven congenitally blind adults trained on such a device, for 73 hours on average, learned to identify letters and other images from their soundscapes. Asked which of seven categories a soundscape showed, the six who were tested were right 78 per cent of the time on average, against 14 per cent by chance.

In the scanner, their visual word form area responded more to letter soundscapes than to soundscapes of faces, houses, bodies, objects or textures. That is the small region in the left occipitotemporal cortex that lights up when a sighted person reads print.

Those participants had never seen. The region had never received visual input, and it took up reading anyway. In one blind woman who had never learned the shapes of printed Hebrew letters, it began responding to them as soundscapes after less than two hours of training.

The standard picture assigns cortex to senses: this is visual cortex, that is auditory. The result above fits a different picture, in which a region is defined by the computation it performs and recruits whatever input can support it.

Amir Amedi’s group, which ran the study, argues that this is a principle of the whole brain and that rehabilitation after sensory loss should be built on it. Marina Bedny’s group reads the evidence from blindness differently. In people born blind they find the same patch responding to the grammar of spoken sentences as well as to Braille, and argue that cortex deprived of its usual input can take on a wide range of functions, reading among them.

Users describe the experience differently, and some report that it stops feeling like listening. Whether that is seeing, a changed kind of hearing or something closer to reading is argued over, and some who study it doubt that any measurement could settle it.

The behavioural question is cleaner: performance improves with training and carries over to new images of the kinds that were trained, though it stays slower and less accurate than sight.

The idea is decades older than this device. Work from the 1960s onwards delivered images through an array of stimulators pressed against the skin of the back, and later through a grid on the tongue, and reported that trained users came to describe objects as being out in front of them rather than as patterns on their skin, though that rests mostly on what users said.

Training is long, the bandwidth is low compared with vision, and the devices compete with a cane and with Braille, aids that are familiar and already proven.

A finding can be robust in the laboratory and marginal in a life, and the gap between those is not a failure of the science.