User`s guide

E-Prime User’s Guide
Chapter 3: Critical Timing
Page 122
The second error is a mismatch between the assumed refresh rate of the display card and the
actual refresh rate. Starting at images 241, there are a large number of spikes in the delay.
These indicate timing errors, occurring because the program was specified to run using a 60.1Hz
refresh rate but was actually run at 73.1Hz without correction for this difference. The experiment
was waiting for a delay at 66.6ms for 4 refreshes and began waiting for the vertical blanking
signal at 56ms. Sometimes the sync signal was detected at delays of 55ms or 68ms (4 or 5
refreshes at 73.1Hz). In comparison, the intended duration and the measured duration lines on
the graph also show the problem caused by the mismatch between the intended and actual
refresh rates. For images 1-120, the Intended duration was 100ms and the measured was 96ms.
E-Prime allows the logging of time related variables for all stimulus objects (refer to Technique 4,
section 3.3.1.4). The OnsetDelay shows delays from the specified times. This is usually due to
waiting for the vertical refresh. The OnsetTime is the time in milliseconds since the start of the
experiment when the stimulus presentation was actually initiated. Using E-DataAid the
experiment data may be loaded and these timing variables can be selected (e.g., use the Arrange
Columns dialog to specify a wildcard for variable selection “*time;*delay”, and click the Select
button), copied to the clipboard, and pasted into a spreadsheet such as Excel to produce plots.
This is how the timing graphs in this chapter were created.
The time between any events in the timing log can be calculated with Excel. For most display
events, the actual duration is not the actual duration of the stimulus object (i.e., object.Duration in
the data file), but rather must be computed as the difference between the stimulus onset and the
onset of the stimulus that removed/replaced the initial stimulus (e.g., due to added delays in
waiting for the vertical blank signal). The OnsetTimes can be exported and actual display
durations can be calculated easily in Excel. For example, if presenting a Fixation, a Probe, and a
Mask (as in Timing Paradigm 2 and 3 above), the following column-based calculations can be
performed in Excel to compute the Onset-to-Onset time of each event in the presentation
sequence.
Fixation_To_Mask
= Mask.OnsetTime – Fixation.OnsetTime
Fixation_To_Probe = Probe.OnsetTime – Fixation.OnsetTime
Probe_To_Mask = Mask.OnsetTime – Probe.OnsetTime
You should analyze your timing data as a function of your independent variables. For example, in
Timing Paradigm 3, we varied the duration of the display. In that example, we analyzed the
onset-to-onset time as a function of the ProbeDuration attribute. Checking those numbers allows
verification that the timing of each condition is as expected. Calculating the actual durations and
reporting the mean and standard deviation of the timing provides a method to communicate the
timing accuracy of the experiment (e.g., displays were presented for 16.88 (S.D. 0.33ms), 33.25
(S.D. 0.43ms) and 49.75 (S.D. 0.43ms)). Also, looking at those numbers can expose aberrant
values that may need to be reported or suggest the need for additional experiment debugging.