Category Archives: eye movements

Saccade Gain Adaptation

(left:me right:Jordan)

I’m a graduate student. I study Neuroscience The City College of New York as a student of the CUNY Graduate Center. I work in the Biology Department of CCNY in the lab of Josh Wallman. I study a process known as Saccade Gain Adaptation.

Saccades (as I’ve described before in this very blog) are rapid point to point displacements of gaze. Unless you have a target which is moving that you can follow with your gaze, you make saccades. This is unique to eye movements. That is to say that there are no such constraints on arm or leg or any other kind of movements. If you want to move your arm from one location to the other, there are a tremendous number of paths to follow and speeds to employ. With eye movements, however, you have little to no control over the path taken or the speed of the action. If you’ve never heard of this, it’s useful to try and trace a line (like a corner where walls meet) slowly with your eyes. You’ll rapidly see that this is impossible, the best you can do is make small steps along the line. This is in contrast to say, holding out your hand and moving it across your field of vision while fixating on one of your finger-tips. In this case, you can make a smooth pursuit movement.

Saccade gain adaptation is a process through which the size of eye movements elicited by an abrupt change in position of a target is either increased or reduced. Below is a little flash movie that I made illustrating this procedure.

Why might the eye want to behave in this way? Let us suppose that over time the muscles in your eye become weak as a result of aging. It makes sense that the commands sent to your eye muscles must also change in order to accurately move your eyes to desired targets in the world. Josh (and I) think that this is not quite the whole story, but accept for the moment that it is possible and sensible for the brain to be able to change its saccadic gain.

Experimentally, we can induce gain adaptation in the following way: we start with “no step back” trials (runs) in which the subject fixate a small target. At some unpredictable (to the subject) time the target abruptly changes position, and the subject’s instructions are simply to follow the target. After some of these “baseline” trials, we move on to the “step back” trials. In this phase, the target steps to a new position, but when the subject moves their eye to the new target position, we move the target back slightly. Interestingly, if the step is small, the subject will not even notice the second target movement, they will simply follow it with their eyes. After many hundreds of trials (“Step Back (late)”), instead of making two saccades to reach the final position of the target (after its two steps), the subject will simply make one saccade to the final target location.

(Click on the black play button to see what a single trial looks like, and the trial type buttons to change the trial type)

In the above animation, there is a graph representing the type of data that we gather. We are only interested in the horizontal (or x) position of the gaze. All of the stimuli we’re using in the experiment are presented on a monitor. We use a computer to control the presentation of the stimuli, and we use a camera and a calibration procedure to record the direction of gaze from a subjects right eye. In the graph, the vertical axis represents horizontal displacement from some arbitrary zero point (I’ve purposefully ignored details such as this). So when the trace representing the target jumps up, that means it has moved to some new position to the right of where it formerly was. The same is true of the trace for the gaze position. When a trace steps down, it means the correspding thing (target or gze) has moved to the left. The set of gray dots merely represents the passage of time. I’m not sure if this description is complete enough, but hit the “play” button a bunch of times and puzzle over it if you’re still confused and it’ll make sense, or email me and I’ll explain ad nauseum.

A final note that there is a rather large (~%10 of the size of the eye movement) error in most saccades, I’ve simply idealized the graphics to simplify the presentation.

Miniature Eye Movements

Your brain doesn’t care about brightness, it likes contrast. In fact, by the time signals generated by light impinging on your retina propagate through its 10 layers of cells, brightness information has largely been discarded in favor of contrast, both spatial and temporal (see paragraph two for a description). A very simple example of this is demonstrated below. Initially the contrast (in time) of the two dots is the same because they are surrounded by the same brightness. When you click on the thin or thick surrounds button, the contrasts are now inverted between the two as evidenced by the change in percept. I guarantee that nothing about the dots themselves change, only the surrounding area.

(Shapiro, A. G., D’Antona, A. D., Charles, J. P., Belano, L. A., Smith, J. B., & Shear-Heyman, M. (2004). Induced contrast asynchronies. Journal of Vision, 4(6):5, 459-468,

Now let me disambiguate a bit what is meant by temporal and spatial contrast. A painting, say Seurat’s Sunday Afternoon on the Island of La Grande Jatte, has plenty of spatial contrast, but because nothing changes in time, there is no temporal contrast. A movie screen filled with white which fades to black and back to white, oscillating, has plenty of temporal contrast and no spatial contrast. Now, if there is no temporal contrast at all in your visual field, the world will fade away. Your visual system needs temporal contrast. This is one of the purposes of so called fixational eye movements. These are small involuntary eye movements which you make between the large point to point movements called saccades that we use to change the direction of our gaze. So even if you were standing in front of a painting such that it filled your vision completely and only stared at one point, your eyes would move slightly, around your fixation target, to prevent the image from fading away. If somebody drugged your eye muscles so that there was no way to execute these small movements and filled your vision with an image that had no temporal contrast, the world would fade away.

The idea that brains only encode change and not static values of sensory data is pretty ubiquitous, and there are a wealth of examples. What I’d like to continue with, however, is another function of fixational eye movements that has been speculated about but only demonstrated of late. In a recent paper in Nature*, researchers have discovered that these small eye movements serve to enhance our fine scale spatial resolution. That is to say that without small eye movements, we are less able to detect the presence of and report the properties of fine spatial scale visual stimuli.

One very useful analogy is the way we run our fingers over something textured to better comprehend the shape of it. For example, suppose you were blindfolded and I put a piece of wood in your lap. I tell you that this piece of wood has some number of very small adjacent grooves cut into it at some particular position. If I asked you to find them and count them, I suspect that you would run your fingers across the wood until you found them and then rub your index finger over them a couple of times to determine the number. It seems a very natural way to do it, and this is exactly akin to making small eye movements to improve spatial resolution. Not making small eye movements like that would be akin to simply pressing your finger down straight on the grooves in an attempt to count them. Perhaps you could do alright at this if there were only one or two, or if they were very big, but as the task got more and more difficult you’d need to use the sliding technique in order to discriminate. The commonality here is that both your sense of touch and sense of sight are mediated by an array of detectors of fixed size and position, and some stimuli are simply too small and/or finely spaced to be accurately detected by the particular array of detectors you’ve got.

Here’s another example: suppose you were using a number of long same-diameter, same length rods to determine the topographical features of a small area of the bottom a pool of water. One way to do this would be to take many rods in a bundle and push them each down (still in a bundle) until they stopped, recording each of their heights individually. The problem with this method is that the resolution of your image of the bottom is limited to the diameter of the sticks. Assuming you can’t use ever thinner sticks (we can’t make the receptor size in our eyes or hands arbitrarily small), you can get a better resolution image by running a single rod (or many rods) over the area to be mapped, continuously recording the height. In this way you have more information than if you simply assign each rod to a single point on the bottom, increasing your resolution.

*Rucci, M., Iovin1, R., Poletti1, M. & Santini, F. Miniature eye movements enhance fine spatial detail Nature 447, 852-855 (14 June 2007)