User`s guide

E-Prime User’s Guide

Chapter 3: Critical Timing

Page 71

3.1 Executive Summary of E-Prime Timing

Precision and Implementation Methods

Psychology Software Tools, Inc. has tested E-Prime on a wide range of single processor desktop and

laptop computers using Pentium and Celeron series processors from Intel, and Athlon series

processors from AMD. Test machines varied in speed from 60MHz through current generation

processors in the GHz range. Results show that E-Prime can produce millisecond precision timing

(see operational definition of millisecond timing, section 3.2) on machines running faster than

120MHz. Internally, E-Prime uses a microsecond clock for all timing assessment. This crystal clock

maintains its precision relative to external testing hardware, and has been stable in all tests.

Although a wide range of processor types and speeds were used in testing E-Prime, much of the data

collected and presented in this section was obtained when running the system on mid to low-end

machines in order to more effectively demonstrate the various types of timing anomalies that can

occur during the run of an experiment. The most serious problem with maintaining millisecond

precision using common desktop computers and operating systems is that the operating system will

take control of the computer, occasionally blocking E-Prime (and any other program) from reading the

clock. E-Prime minimizes the quantity and duration of such events by running at high priority, and by

structuring experiments to minimize interruptions and virtual memory swaps. With a well-configured

system, the probability of missing a millisecond tick of the clock is small (typically less than 0.05%),

but in a preemptive multi-tasking environment (e.g., all versions of the Windows family of operating

systems released after Windows 3.1), it can never be zero. Rare, short duration (<5ms) missed clock

ticks have negligible impact in most experiments (e.g., equal to running 1 extra trial per 22,500 trials).

E-Prime provides time audio data logging facilities to identify errors and, if desired, the researcher

can exclude those trials from analysis. E-Prime has been extensively tested with external hardware,

and the timing is millisecond accurate (average error <= 0.5ms) over a wide range of delays,

input/output events, and objects used in E-Prime experiment development. The crystal clock has an

accuracy of 0.15ms. The photocell/simulated keyboard test uses external hardware to sense when a

stimulus is presented, record any delays that occur, and collect timing information in relation to

keyboard responses. The typical keyboard system delay was a stable 7+/- 0.5ms, and the correlation

between observed and expected was r=0.9999975. Refer to Appendix A for a complete description

of timing tests and results.

It is critical that the researcher test and configure a system to minimize concurrent processes

operating while the experiment is collecting time critical data. E-Prime provides a diagnostic program

(see User’s Guide Appendix A –Timing Test Results) to assess timing accuracy and precision and the

ability of the computer hardware/software to detect the onset of the video systems vertical

blanking/refresh events, which allows the synchronizing of stimulus displays with the video refresh of

the monitor. Some hardware/software configurations do not support millisecond timing or reliable

refresh detection, necessitating the running of the diagnostic test