User manual
86 Data Files
© 2005-2008 SR Research Ltd.
detection of saccades as small as 0.3°, ideal for smooth pursuit and
psychophysical research. A conservative threshold of 30°/sec is better for
reading and cognitive research, shortening saccades and lengthening fixation
durations. The larger threshold also reduces the number of microsaccades
detected, decreasing the number of short fixations (less than 100 msec in
duration) in the data. Some short fixations (2% to 3% of total fixations) can be
expected, and most researchers simply discard these.
Use of eye-movement acceleration is important for detection of small saccades,
especially in smooth pursuit. Acceleration data has much more noise than
velocity data, and thresholds of 4000°/sec
2
for small saccade detection and
8000°/sec
2
for reading and cognitive research are recommended. Lower
acceleration thresholds will produce false saccade reports. Acceleration data
and thresholds for the EyeLink 1000 system may be larger than those reported
for analog eye trackers. These systems use multi-pole filters for noise reduction
that adds delay and smoothes the data, significantly reducing the measured
acceleration.
The saccadic motion threshold is used to delay the onset of a saccade until the
eye has moved significantly. A threshold of 0.1° to 0.2° is sufficient for
shortening saccades. Larger values may be used with caution to eliminate short
saccades: for example, a threshold of 0.4° will always merge fixations separated
by 0.5° or less, but may eliminate some 1° saccades as well. The threshold
should be set to zero for non-cognitive research, or where statistics such as
saccadic duration, amplitude and average velocity are required.
Examples of the commands to set these thresholds are:
saccade_velocity_threshold = 30
saccade_acceleration_threshold = 8000
saccade_motion_threshold = 0.15
4.3.6 Pursuit Thresholds
During smooth pursuit and nystagmus, saccades must be detected against a
background of smooth eye motion as fast as 70°/sec. While acceleration can be
used to detect these saccades, velocity data must also be used for reliable
detection of all saccades. The EyeLink 1000 parser raises the saccadic velocity
threshold during pursuit by the average velocity over the last 40 milliseconds.
This is reliable, and does not degrade parser performance during non-pursuit
eye movements.
During long saccades such as the return sweep in reading, this fix up causes
the saccadic velocity threshold to be raised. This is not a problem as long as the
adjustment is limited, as it helps to prevent prolongation of these saccades by
overshoots and glissades. The pursuit threshold limits the amount that the
saccadic threshold can be raised. A limit of 60°/sec works well for most pursuit