User`s guide

E-Prime User’s Guide
Chapter 3: Critical Timing
Page 96
environment. Each crystal, while very precise and accurate, will still have some very small
degree of error inherent within it (e.g., some number of parts per billion). The exact degree or
amount of error is unique to each individual crystal. Unfortunately, the amount of error in distinct
crystals varies from one crystal to the next such that two computers may have stable clock rates
but one is a tenth of a percent faster than the other and thus their times will “drift” out of
synchronization over time. This small degree of error or crystal imperfection is commonly
considered to be within the manufacturer's operating performance specifications of the clock. For
behavioral research on one computer, this is not a problem. However, this means that if there
are two computers trying to synchronize time, there is a possibility that they might lose a
millisecond per second. If trying to synchronize two computers for extended periods of time, this
drift (60ms per minute) can be a serious problem.
This clock drift scenario is commonly detectable when an experimental computer is required to
share its timing data with a recording computer that maintains its own independent time base, as
occurs in ERP (Evoked Response Potential brain wave) or fMRI (functional Magnetic Resonance
Imaging) . In these cases, we ideally want to scale or calibrate the two clocks such that they stay
synchronized over the critical timing period (e.g., if the recording period is 600 seconds, you
would not want to lose more than a millisecond per 600 seconds).
3.4 Implementing Time Critical
Experiments in E-Prime
This section will explain how to implement critical timing paradigms. Also included are sample
programs for each of these methods. There are six steps to implementing and verifying critical
timing.
Step Task Action
1. Test and tune the experimental computer for research timing. Run
test experiment to determine the refresh rate of the display and
general timing capabilities of the computer. Note this must be
done for the machine configuration and does not have to be
done for each experiment, but should be repeated if any
significant hardware or software change is made to the
machine after initial testing.
Record clock and refresh miss rates and
refresh frequency; set up the machine to
perform millisecond timing.
2. Select and implement a paradigm timing model that best matches
the critical timing requirements of the paradigm.
Identify and implement one of the
common paradigm models:
• single stimulus event to response
timing
• critical sequence of events
• critical sequence of events with
varying duration of probe
• cumulative timing of a repeating
sequence of events
• continuous sequence of events at high
rate with short stimulus times.
3. Cache or preload stimulus files being loaded from disk as needed to
minimize read times.
Preload stimuli to minimize generate time
and operating system delays.
4. Test and check the timing data of the paradigm. Record timing data and review or plot
stimulus durations, onset times, and onset
delays. Verify refresh rate reported.
5. Run pilot experiments to collect timing and behavioral data and
analyze the timing data.
Record subject and experiment timing data.
Check timing data and report critical timing
data.