Our sensory systems are very good at adapting. For example, our eyesight rapidly adjusts to changes in the ambient light level, allowing us to see in both near darkness and bright daylight. This allows the sensory neurons to use their limited response range to the full, reassigning each specific response value to a different stimulus strength that is appropriate to the new conditions. But Fairhall et al. now show that there is much more to adaptation than meets the eye.

Fairhall and colleagues recorded from motion-sensitive visual neurons in the blowfly Calliphora vicina to investigate the dynamics of adaptation to a randomly fluctuating motion stimulus, the variance of which changed over time.

The average firing rate of the neurons increased or decreased abruptly when the stimulus switched suddenly to a higher or lower variance. It then gradually reached a steady-state level, adapting relatively slowly. The timescale of this adaptation was related to the variable timescales of changes in the stimulus.

The coding in terms of precise spike timing, however, adapted much more quickly — within tens of milliseconds. This adaptation time was limited only by how quickly the system could gather enough information to reliably recognize the change. In this way, the cell minimizes the amount of time for which its information transfer is suboptimal. So there are at least two independent timescales of adaptation, carried by different aspects of the response statistics.

To avoid ambiguity, the nervous system must also carry information about the context of the adapted signal — in this case, the stimulus variance. Here, this information is conveyed by the same neuron. Although the firing rate changes slowly in response to the variance, Fairhall et al. found that another firing statistic — the interspike interval distribution — changed to reflect the new variance on similarly rapid timescales to the input–output relation.

So the coding scheme of this neuron, at least, seems to use several channels simultaneously to convey information about the stimulus — and these channels adapt independently, on different timescales, both to optimize efficient information coding and to prevent ambiguity.