User`s guide

E-Prime User’s Guide
Chapter 3: Critical Timing
Page 76
The reporting of the refresh rate by Windows is "invalid". On this particular hardware, if you set
the Refresh Rate at 60Hz for the 1280x1024 resolution on the Settings tab, Windows will always
display 60Hz while the actual refresh rate is 75.3Hz. This is because Windows views the settings
as a "request" and allows the display device driver to provide its best approximation to the
request. The company that implemented the display driver code configured the software to
display an acceptable image for business applications, but did not ensure that it was accurately
reporting the refresh rate, or reporting back to Windows a close estimate of accuracy. This is a
dangerous problem. Windows reports only the requested time and not the actual time. A 20%
error means that if you are estimating refreshes for 60Hz, after 60 refreshes, which you expect to
be 1000 milliseconds, the duration was actually 797ms. This is a 203ms error. It should never be
assumed that the specifications provided by the hardware manufacturer provide the precision that
you need. For example, if the monitor specifications say 72Hz at 1024x768, that number might
be off by 50% depending on the interaction of the operating system, display driver software,
monitor, and hardware connection to the monitor (e.g., presence of a USB port). The point of the
above example is that the experimenter must measure the system and not depend on the
“manufacturer specs” to provide timing data with the validity and precision needed for
professional psychological research.
E-Prime provides routines to measure and report the refresh rate as well as other properties of
the display system. E-Prime automatically measures the refresh rate in Hz at the beginning of
each run of an experiment and logs the value under the variable name Display.RefreshRate in
the data file. As a secondary check on the refresh rate, current generation display monitors will
often provide a reasonably accurate measurement of the refresh rate using the monitor’s on-
screen display menus. However, the researcher is warned to be sure to measure the refresh rate
during the actual run of an experiment. Experiment generators that allow the user the ability to
specify the display resolution and color depth of the experiment to be different than that of the
default desktop display mode will force a display mode change to be performed in hardware at
runtime. When this mode shift occurs, the display hardware may be forced to alter the refresh
rate to accommodate the new display parameters.
3.2.2 Problem 2: Actual durations can deviate from
intended durations
The most serious problem for computerized psychological research is that the computer requires
substantial time to perform actions and may halt the experiment without notice, thereby grossly
distorting timing. You have probably experienced times when your word processor occasionally
pauses for a short time and then resumes. This same process can happen during an experiment.
Let us take a simple example. Assume you intend to display thirty-six bitmap images, each for
200ms (e.g., a rotating checkerboard). Use a program to read the bitmaps and display them then
check the timing with a stopwatch. When we do this using standard programming practices, the
displays seem to be taking about 300ms, with occasional displays taking nearly a second. The
specifications are clear: present each display for 200ms. In addition to the mean timing error,
when running the program 5 times, these long pauses happen at different times. Figure 1 below
illustrates the measured times from two such runs. The thick line at 200ms shows the intended
time. Session 1 (solid line) and Session 2 (dashed line) have a median display time of 305ms
with irregular spikes as high as 934ms. There are three glaring faults to be recognized from this
illustration. First, the median duration is about 105ms longer than was intended. Second, there
are severe spikes indicating long unintended pauses in the experiment. Third, the spikes happen
at random times.