User manual

84 Data Files

4.3.1 Parser Operation

The parser uses velocity and acceleration-based saccade detection methods.

Because of the EyeLink 1000 tracker’s exceptionally low noise levels and high

spatial resolution, very little data filtering is needed and thus delay is kept

small. The 250, 500, 1000, or 2000 Hz sampling rate gives a high temporal

resolution of 4, 2, 1, or 0.5 millisecond (Note: Availability of some sampling

rates and options depends on the system model).

For each data sample, the parser computes instantaneous velocity and

acceleration and compares these to the velocity and acceleration thresholds. If

either is above threshold, a saccade signal is generated. The parser will check

that the saccade signal is on or off for a critical time before deciding that a

saccade has begun or ended. This check does not affect the recorded time of the

saccade start or end, but adds some delay to the real-time events sent through

the link.

During each saccade or fixation, data is collected on velocity, position, and

pupil size. At the end of the saccade or fixation, this data is used to compute

starting, ending, and average position, pupil size and velocity, as well as peak

velocity. Velocity data is also converted into units of degrees per second using

real-time resolution information. This data is then used to create events which

are sent over the link and/or recorded in an EDF file. See the section 4.5.3 “Eye

Movement Events” for more information on events.

4.3.2 Parser Limitations

The EyeLink 1000 parser was designed for on-line, low delay identification of

saccades and blinks. Detection of very small saccades may require off-line

processing, as the special filtering and computation of global velocity cannot be

performed on-line. In smooth pursuit research, the parser is less sensitive to

small back-up saccades (opposite to the direction of pursuit) than forward

saccades, due to the low peak velocity of back-up saccades.

The parser only looks “ahead” in the data a short time to compute velocity and

acceleration. This limits the data checking the parser can do. Post-processing or

data cleanup may be needed to prepare data during analysis. For example,

short fixations may need to be discarded or merges with adjacent fixations, or

artifacts around blinks may have to be eliminated.

Nonetheless, the EyeLink 1000 parser does an excellent job in most recording

situations. Adjusting the configuration of the parser may help bias its

performance for specific applications such as smooth pursuit or reading

research. Its performance is easily checked: record eye movements using the

display of interest, with both sample and event data. Then view the EDF file