User`s guide

E-Prime User’s Guide
Chapter 3: Critical Timing
Page 114
Cumulative Timing Error
Event versus Cumulative Mode
-20
0
20
40
60
80
100
120
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Trial
Onset Error (ms)
Event Timing
Cumulative Timing
Figure 21. Timing onset errors for the same presentation sequence run in Cumulative versus Event
timing mode
Cumulative timing mode has a cost in terms of increased variability in stimulus duration. The
standard deviation for the Probe to Mask duration was 3.98ms for Cumulative mode and 0.48ms
for Event mode. In Cumulative mode, durations were expected to vary as much as a duration of
a refresh. The table above shows that the duration on trial 2 was 116ms rather than the intended
100ms. This occurs when the refresh for the previous event occurred very shortly after the target
onset time, and the mask had its display very late after its onset time, increasing the duration by
as much as a full refresh. In experiments that are presenting stimuli at a fixed rate, this should
not be a concern. It is not possible to have the condition of no cumulative drift and no duration
variation, unless the refresh duration is an exact multiple of the refresh rate and there are no
timing delays in the experiment. This is not possible in practice due to the nature of video cards
(see Problem 3, section 3.2.3).
Rescaling the E-Prime clock to match external hardware. When external recording hardware
is used during a behavioral experiment to save biological data (e.g., ERP, EEG, fMRI, eye
tracking) it is often necessary to relate the timestamps of critical events in the experiment to
specific temporal points in the biological data stream. This is commonly done by merging the two
data streams in some manner, and in the process, aligning all stimulus and biological events
sequentially in time. Ideally, a shared real-time hardware clock would be used to time stamp data
in each unique data stream so that the temporal alignment process would be straightforward.
However, when mixing and matching hardware and software from multiple 3
rd
party vendors, an
arrangement such as this is typically not technically possible. Thus, because multiple distinct
hardware clocks are used during the recording, the resulting timestamps must be adjusted
relative to one of the clocks, and must be made to share a common start point in the time base in
order to relate and align the data accurately within the merged data streams.
A shared reference or “zero point” can be established by sending a signal from E-Prime (e.g.,
using the WritePort command) to the external monitoring equipment at the beginning of the
experiment. At the same time the E-Prime clock is read and the timestamp of the synchronization
signal is saved. Post processing of the data streams can then subtract out the difference
between the two clock offsets to get both data streams aligned to a common zero point.