Timothy C. Hain, MD

Page last modified: March 12, 2021

There are many (rare) abnormalities of eye movements that can be diagnosed through observation of saccades - -rapid eye movements.

In this much too long web page, we discuss and illustrate the following:

  1. Methodology of Saccade Testing
  2. Disorders of Saccade velocity
    1. Too slow
    2. Too fast
  3. Disorders of latency (timing)
  4. Accuracy (discussion of these is on another page)
    1. Overshoot
    2. Undershoot
    3. Staircase
    4. Blindness
    5. Glissades
    6. Pulsion
  5. Square Wave Jerks (saccadic oscillations with a latency between each saccade)

There is a separate discussion of saccadic nystagmus.

1. Methodology of the calibration test:

The main take-home message is that commercial ENG software is terrible at saccades, and one must look at the traces, and the patient's eye, and use one's clinical judgement. There are many many rare disorders of saccades -- one needs to have a lot of knowledge to make much of them, and it doesn't pay off that often as there are not that many of these patients.

The technical problem is that saccades are very fast, and ENG manufacturers generally have optimized their equipment for slow things (like caloric nystagmus). Sampling rates of 15-30/second are fine for calorics, but just don't do much good for something that moves at 600 degrees/second. Consider -- at 600 deg/sec, the eye could go 30 degrees in 0.05 second. That is one sample at 20/sec. It is generally agreed that one needs to sample at least at 100/sec to have much of a chance at measuring saccades. Video is just not up to this - - because the usual frame rate is only 30/second. It can't be done (with standard cameras).

Also, ENG manufacturers haven't done a very good job at building reliable software that quantifies saccadic measurements. We routinely ignore the summary graphics on our own ENG system concerning velocity and latency, because they make so many errors. We just look at the traces (which may be misleading too), and the monitor.

So, what it boils down to is that one must generally ignore the ENG machine's summary graphs, examine the patient and watch their eye movements on the monitor. The VNG equipment just doesn't do the job.


A typical calibration test is shown below.



The calibration test measures rapid eye movements and calibrates the eyes for the remainder of the test. Horizontal rapid eye movements are always assessed. Vertical eye movements can also be measured, especially with newer commercial ENG systems. Only rarely are torsional eye movements are measured. Devices that measure eye movements are discussed here.

Table 1:

Methodology of eye movement Recording

Measures Doesn't measure
EOG (Electrooculography) Horizontal well, Vertical poorly Torsion
IR (Infrared) Small horizontal very well, Vertically poorly Torsion
Video Horizontal and vertical moderately well Torsion only by inspection or by using unusual, non-commercially available equipment (as of 2004)
Scleral eye coil Horizonal and vertical very well With special measures can measure torsion moderately well


Saccadic testing is sometimes useful as cerebellar disorders and degenerative disorders of the central nervous system can sometimes be revealed through saccadic testing. While the clinical examination by an experienced examiner is the more efficient method of making these diagnosis, such expertise is not always available. The three saccadic parameters most relevant to clinicians are peak velocity, latency, and accuracy.

2. Disorders of Saccadic Velocity

Table 2: Peak velocity of 20 deg saccades in normal subjects
METHOD Peak Velocity   Lower limit of normal
Infrared (300 hz bandwidth) 657   491
Eye coil (60 hz bandwidth) 650 325
EOG   (15 h bandwidth) 336 252

Normal values for the velocity of 20 degree saccades are given in table 2. Note that velocity is very sensitive to the method by which saccades are recorded. Normal saccadic velocity values obtained via infra-red methods or scleral search coil recordings are usually higher than those obtained via EOG recordings. EOG and video methods presently predominate oculomotor recordings. These two methods have similar bandwidths and thus similar limits for normal. These are just not very good at all for saccadic testing.

Saccadic velocity is approximately proportional to saccadic amplitude for sizes between 5 and 20 degrees. After amplitude reaches 20 degrees, saccadic velocity undergoes a soft saturation with respect to further increases in amplitude. This pattern is seen on main sequences, which plot peak velocity against saccade size. The usual upper limit for saccadic velocity, no matter how big the saccade, is about 750 deg/sec. The author uses the function given in equation 1 (see below) for his limits of normal velocity. For the lower limit, the asymptote is set at 350 deg/sec. For the upper limit, the asymptote is 750 deg/sec. Saccade amplitude is designated by E and saccadic velocity, E dot. Saccadic velocity cannot be altered voluntarily and is not affected substantially by age or gender.

Eqn 1


There are several pitfalls to be aware of in measuring saccadic velocities. Variability is appreciable and one is advised to acquire about 40 saccades varying in size between 10 and 40 degrees to develop a reasonable main sequence. Calibration error is another common problem. The calibration error may be related to subtle factors, which are not evident when the oculomotor test is read. For example, patients with ocular motor palsies are unable to get one or both eyes to the target. Patients with strabismus may alternate the eye that they view from, depending on the direction of gaze, and allow one eye to drift out away from the target. In these instances, measured saccadic velocities are wrong, because the calibration is inappropriate. Monocular recording and single eye viewing are essential to avoid error in these sorts of patients. In patients without problems of ocular alignment, evidence that the calibration is stable over several trials must be available before diagnosing abnormalities of saccades.

2a. Slow saccades:

The discussion of slow saccades is very long and has been moved here.

2b. Too fast saccades

Abnormally fast saccades

Abnormally fast saccades can usually be traced to an error in calibration or a noisy eye movement recording. These can be very difficult to spot as commercial ENG equipment is usually able to produce much useful information concerning overly fast saccades (or overly slow ones either).

Calibration errors:

Calibration error in a patient with an INO that made saccades look more normal than they actually are. The top trace shows that the eye excursion is far greater than the target. The bottom trace shows that while adducting saccades are slower, nevertheless they are quite close to normal. If one were to reduce the velocities in proportion to the calibration error, nothing would be normal in this patient's recording. Thus, the laboratory test is misleading.

This recording was made with Micromedical Technology Equipment.

As an example of calibration error, consider the case when a patient makes a 5 degree saccade to a 15 degree target displacement. Then calibration factor will be three times too large, and saccades will appear to be abnormally fast. Noisy recordings, such as due to a poorly applied EOG electrode or blink artifact are another cause of what appears to be inappropriately fast saccades. Because saccadic velocity is calculated from the peak velocity, velocity noise adds to the real peak velocity, and results in incorrectly high velocities. This is particularly a problem with infra-red recordings, which combine high bandwidth with susceptibility to blink artifact.

A more subtle example of a calibration error is shown in the figure above. Here, a patient who was obviously abnormal in the clinic, looked nearly normal on oculomotor testing. This is due to an incorrect calibration that made all saccade velocities artifactually larger. Thus, a wrong calibration can make someone who is very abnormal, look almost normal !

Opsoclonus and flutter

In rare instances, abnormally fast saccades may real, and not due to a technical artifact. One cause is the opsoclonus syndrome or it's relative, ocular flutter. In these conditions, patients make unintended saccades without intersaccadic interval, which may be abnormally fast for their size (Bergenius, 1986). Much more about ocular flutter and opsoclonus can be found here.

A rare cause of saccades that are too fast for their size are ocular disorders in which eye movement is restricted. A large saccade may be programmed centrally, but because the eye is brought up short by muscular restriction or rapid fatigue, a small saccade is made with the velocity appropriate to a bigger saccade. A clue here is that these patients never make saccades faster than the upper limit of normal for large saccades (about 750 deg/sec for recordings made with a 40 Hz bandwidth)

Table 6: Asymmetrical saccadic velocity

Saccadic velocity, for a given amplitude, should be equal between eyes. Velocity should also be equal whether the eye is abducting or adducting. Saccadic velocity asymmetry then consists of significant inequality in velocity between eyes or directions. Of course, asymmetry between eyes can only be detected when monocular recordings are available. Unfortunately, the method by which velocity is measured can create an artifactual asymmetry. Studies done using EOG recordings suggest that adducting saccades are faster, while studies performed with infrared recordings suggest that abducting saccades are faster (9)(Becker, 1989). The difference between the peak velocities of abducting and adducting 20 degree saccades reported by Fricker and Sanders (1975)(13) in a population of 40 normals ranged from -70 to 200 deg/s (95% range, infrared method). We recommend establishing one's own fifth percentile lower limits of normal, for the method in use locally.



Internuclear ophthalmoplegia (INO) recorded in right eye (recording method: VNG, Micromedical Technologies). Note how the rightward going saccades are brisk, while the leftward going saccades (adducing) are slow.

Supplemental material : Video of Internuclear ophthalmoplegia due to multiple sclerosis (same patient)

Table 6 lists the most frequent causes of asymmetrical velocity. There are several potential asymmetry patterns, the most common of which is normal abduction with slowed adduction. This occurs mainly in internuclear ophthalmoplegia or "INO" . INO is due to a lesion in the median longitudinal fasciculus or "MLF", which connects the paramedian pontine reticular formation and the oculomotor nucleus. The MLF is located immediately adjacent to the cerebral aqueduct.

INO is most often found in patients with multiple sclerosis or cerebrovascular accidents involving the brainstem (Fischer, 1967; Wall and Wray, 1983). The hallmark of INO is slowing of adducting saccades, accompanied by an overshoot of the abducting eye. The condition can be unilateral or bilateral. A reduction of adducting velocity into the abnormal range, accompanied by normal abducting velocity, for medium size saccades (about 20 deg), should cause one to consider INO. In this case, one should also examine the position traces of each eye. The combination of an overshoot of the abducting eye, and significant slowing of the adducting eye occurring simultaneously, confirms INO.

Normal adduction with slowed abduction occurs most commonly in patients with palsies of the sixth cranial nerve. One should look for substantial slowing for a medium size saccade. Note that calibration error is common in this situation, as the patient with a sixth nerve palsy will often be unable to fixate the target with both eyes, when looking in the direction of paresis.

There are several other patterns which occur frequently in patients with cerebrovascular disease or demyelinating disease involving the brainstem. Preserved abduction in one eye, combined with slowing of all other horizontal motion in both eyes, occurs in the one-and-a-half syndrome. Reduced speed of adduction in one eye combined with reduced abduction in the other eye, occurs in conjugate gaze palsies. Slowing of all horizontal saccades, combined with normal vertical saccades occurs in bilateral pontine lesions which affect the burst cells, such as pontine hemorrhage.

3. Saccade Latency (timing)

Table 7: Disorders of Saccadic Latency

Saccadic latencies are nearly useless in clinical ENG testing, and the pragmatic reader who merely wants to know about diagnostic information should skip this section.

Saccadic latencies are calculated from the difference in time between target displacement and the onset of the first saccade towards the new target position. Several paradigms to elicit saccades are available which differed mainly in their effects on saccadic latency. These included the random, express saccade, and anti-saccade paradigms. At this writing, of these, only the random paradigm is used clinically. In this simple procedure, the target changes position at unpredictable times, to unpredictable positions.

Normal saccadic latencies are independent of target amplitude and are insensitive to the method used to record eye movements, but vary according to target luminance, size, contrast, whether the target is visual, auditory or both, and the predictability of the target (Leigh and Zee, 1991). Thus it is best to obtain normal values specific to one's own laboratory unless one is using commercial equipment within an environmentally controlled booth.

Saccadic Latency pitfalls:

There are several pitfalls to consider when measuring latency. Usually, because there are so many pitfalls, it is best to simply ignore latency. But if you insist on knowing anyway -- here are the issues:

Poorly designed software: Commercial ENG software (such as that sold by Micromedical Technology), sets unrealistically high standards for saccadic latency. Many patients are scored as "abnormal", who actually have nothing wrong with them. The well informed interpreter mainly ignores "abnormal" saccadic latencies as indicated by this program.

Visual acuity -- the big problem ! Saccadic latency is strongly affected by visual acuity, and delayed latencies are common in persons with cataracts or other disorders which reduce vision. In clinical ENG testing, when spectacles are removed, latencies will naturally lengthen. As most ENG operators pay no attention at all to visual acuity, and blithely remove spectacles, this causes tremendous scatter in clinical latency measurements !

Brightness: Latency decreases about 15 ms per logarithmic unit of luminance above foveal threshold (Wheeless et al, 1967). Thus a bright target is essential. A small laser produces an extremely bright target which is ideal for this purpose. If one is using an light-emitting diode (LED) based stimulator such as a light bar, it may be helpful to test in dim lighting to improve contrast and minimize effects of visual acuity. Data projector devices vary in brightness, the rooms vary in their lighting characteristics, and generally no consideration is made of these factors.

Paradigms: Latencies are relevant only when the timing of target motion is unpredictable -- patients may anticipate predictable targets, producing an latency that is impossibly short or even negative. Latency may also be reduced by input from nonvisual senses, such as noises associated with target displacement (Konrad et al, 1989).

Prolonged and reduced saccadic latencies

Prolonged latency of saccades in a patient with mild Huntington's chorea. Blue-target, Red-horizontal. Dashed lines are 1 second apart.

A general prolongation of saccadic latency is an average latency greater than 400 msec. While general prolongation is associated certain disease processes as outlined in table 7, in most instances this finding has no diagnostic significance because saccadic latencies are sensitive to the mental state of the subject. Uncooperative patients may simply produce erratic or prolonged saccadic movements.

Malingering patients often produce saccades with variable and (sometimes) extremely long latencies. This can be used as a diagnostic point. Computer software does not pick this up -- the interpreter must make this judgement from experience.

There are no disease processes that cause a general shortening of latency and accordingly this finding is always related to technical error, anticipation, or lack of cooperation. Lack of cooperation can cause the appearance of a general shortening of latency if one is testing a subject who makes large numbers of extraneous saccades, because latency is calculated from the time between target displacement and eye movement. When many extraneous movements are being made, if one occurs by chance just after target displacement, an abnormally short latency may be registered.

Asymmetrical saccadic latencies

On the other hand, asymmetry in latency between saccades into one or the other hemifield is useful clinically as it may indicate the presence of a lesion involving the superior coliculus, or parietal or occipital cortex. What is helpful in this instance is that saccades in one direction provide a control for saccades in the opposite direction. This pattern is frequent in patients who have had cerebrovascular accidents. Patients with occipital lesions may not see targets in the blind parts of their fields, and may produce a "staircase" of searching saccades, the first of which has a prolonged latency (Troost, 1972). Patients with parietal lobe lesions have inattention to the side of their lesion, and may produce no saccade at all or make saccades with prolonged latency to that side (Meienberg, 1983). Patients with unilateral lesions of the superior colliculus may also show asymmetry of latency (Pierrot-Delseilligny et al, 1991).

References on latency:

4. Saccade Accuracy

The eye may go too far, too short, drift, make multiple steps. These are discussed here.

5. Square wave jerks

Square wave jerks -- abrupt saccades away from fixation, followed by a short delay -- and then back onto fixation, are common, and one must develop clinical judgement as to what is normal and what isn't. Follow this link for more about SWJ and a case of pathological SWJ.