The multiple timescales of memory

A few years ago, Yoshiko Kojima, Yoshiki Iwamoto, and Kaoru Yoshida (2004) took their monkey to a darkened room and measured its eye movements (saccades) in response to display of visual targets. Usually, when a target is shown at 10 o with respect to current fixation, the animal makes a 10 o saccade to foveate the target. If we define saccadic gain as the ratio between the displacement of the eyes (about 10 o ) and the displacement of the target (also 10 o ), then in the healthy monkey (and healthy person) the saccadic gain is approximately one. However, Kojima et al. were interested in having the monkey learn to change this gain. To do this, they tested the monkey in a gain adaptation paradigm (McLaughlin, 1967). In the experiment, a target is shown at 10 o but as soon as the eyes begin moving toward it, the 10 o target is extinguished and a new target is shown at 13.5 o . Saccades are so fast (peak speed of greater than 400 deg/s) and so brief (movements complete within 60ms) that the brain cannot use any sensory feedback during the saccade to help control it. In fact, one is effectively blind during a saccade. So in response to a 10 o target, the monkey makes a 10 o saccade, observes the endpoint error, and follows this with a second saccade. As the trials continue, the brain learns to increase the saccadic gain so that in response to the 10 o target it makes a larger than 10 o saccade. This is called gain-up adaptation.