Abstract: Users of most video-based eye trackers apply proximity algorithms to identify fixations and assume that saccades are what happen in between. Most video-based eye trackers sample at 60 Hz., a rate which is too low to reliably find small saccades in an eye position record. We propose to call these slower eye trackers and their typical proximity analysis routines "fixation pickers." Systems such as dual-Purkinje-image (DPI) trackers, coil systems, and electro-oculography permit higher sampling rates, typically providing the 250 Hz or greater sampling frequency necessary to detect most saccades. Researchers using these types of eye trackers typically focus on identifying saccades using velocity based algorithms and assume that fixations are what happen in between these movements. We propose to call these faster eye trackers and their velocity-based analysis routines "saccade pickers." Given that fixation pickers and saccade pickers extract different things from an eye position record, it is no wonder that the two systems yield different results. This discrepancy has become a problem in eye movement research. A study of cognitive processing conducted with one eye tracking system is likely to give results which cannot be easily compared to a study conducted with another eye tracker. Imagine that two investigators are both interested in studying visual search. Both choose the number of saccades as one of their dependent variables to measure search performance. One investigator chooses a video-based fixation picker. The other investigator chooses a DPI-based saccade picker. Because the saccade picker is tuned to identify smaller saccades, the investigator with the DPI tracker reports more saccades and fixations during an equivalent visual search task compared to the investigator with the video-based tracker. Both investigations are likely to produce valid results, but results which are not comparable to the other investigator's.