User`s guide

E-Prime User’s Guide
Chapter 3: Critical Timing
Page 113
third at 7000, etc. The Onset Sync property for the displays is still set to vertical blank so that
each display will delay until the beginning of the refresh cycle. In Cumulative timing mode, E-
Prime will compensate for any onset delay by reducing the duration of the display, so that the
next display will occur at the next refresh cycle. This reduces cumulative timing error at the
expense of increasing between stimulus variability slightly.
The table below shows the durations for each event (left) and the cumulative times (right). The
goal of Cumulative timing mode is to minimize the cumulative drift. The last column shows the
cumulative error in milliseconds. Note that the cumulative error ranges from –9 to +6ms.
Because of the refresh synchronization, we expect to see a variability of up to one refresh
duration for onset events.
TimingParadigm4 Time Audit Data (ms)
Trial Fix-Probe
Probe-
Mask
Mask-
Feedback
Feedback-
Fix
Fixation Onset
relative to start
of sequence
Target
Onset
Cumulative
Error
1 499 100 1895 997 0 0 0
2 499 116 1895 997 3491 3500 -9
3 499 100 1895 1013 6998 7000 -2
4 499 100 1895 997 10505 10500 5
5 499 99 1912 997 13996 14000 -4
6 499 100 1895 997 17503 17500 3
7 499 99 1912 997 20994 21000 -6
8 499 99 1895 998 24501 24500 1
9 498 100 1912 997 27992 28000 -8
10 499 99 1895 1014 31499 31500 -1
11 499 100 1894 998 35006 35000 6
12 498 100 1895 1014 38497 38500 -3
13 499 99 1895 998 42004 42000 4
14 498 100 1911 998 45495 45500 -5
15 498 100 1895 997 49002 49000 2
16 499 100 1911 52492 52500 -8
Mean 498.75 100.69 1900.13 1000.60 -1.56
S.D. 0.43 3.98 7.74 6.55 4.69
Figure 21 below illustrates the cumulative error of this experiment when run with Cumulative
timing mode versus Event timing mode. The Event timing mode adds small delays waiting for
refresh events. In the current example, about 6.3ms were added per trial, causing an increase in
the cumulative error with each trial. In this case, within one minute the cumulative error
amounted to over 100ms for Event timing mode . In Cumulative timing mode, the experiment
was allowed to alter the duration slightly to remain in alignment with the initial start event. In
Cumulative timing mode, (as long as the sequence is not terminated by a subject response) all
stimulus onset times are maintained relative to the start of the sequence (i.e., the last object run
in Event timing mode), not the termination of the previous object. Thus, there is no drift in the
cumulative delay from the target/expected delay.