Timing drift vs. missed keystrokes?

[Ignacio Valdes]

Hello all,

I have a paradigm which flashes numbers sequentially in a n-back
fashion and the user is supposed to click if there is a match. Each
number flashes for 500 msecs with 500 msecs blank in between. There
are some filler filler number trials that also last 500 msec with 500
msec blank screen. In all, there are 84 trials/Block, with a 10 second
rest in between for 6 blocks that is then repeated once. Total time
for the paradigm is 10:34.

There is a problem with timing drift throughout the paradigm if
Pre-release is set to zero. The timing drifts about 5 seconds
throughout the paradigm so that the last 5 trials do not get
displayed. So somehow the trials are slightly longer than they should
be. To fix this, Pre-release must be increased to at least 9 msec for
the number display and 9 msec for the blank screen in between. AFter
doing this, it no longer drifts and stops at the correct moment.
However, this impacts keystroke recording as in: it doesn't record
keystrokes reliably. This is grossly apparent at Pre-releases of 100
msec or more. I'm testing this at 9msecs x 2 = 18msec to see how many
keystrokes it misses. This leaves a possible 0.018 sec window of time
in which a subject response may be missed. E-prime support has been
notified of this last week and they are working the problem, so far
with no success.

In a nutshell, increase pre-release fixes timing drift problem and
introduces keystroke recording problem. Decrease pre-release to zero
fixes keystroke recording problem but introduces timing drift problem.

Has anyone encountered this or have a solution?

-- IV