Path: utzoo!attcan!uunet!cs.utexas.edu!yale!quasi-eli!cs.yale.edu!tarr-michael
From: tarr-michael@cs.yale.edu (michael tarr)
Newsgroups: comp.sys.mac.programmer
Subject: Re: Timer
Message-ID: <26834@cs.yale.edu>
Date: 18 Oct 90 14:53:13 GMT
References: <13261@sdcc6.ucsd.edu> <34967@cup.portal.com> <1990Oct18.032201.27217@ux1.cso.uiuc.edu>
Sender: news@cs.yale.edu
Organization: Department of Psychology, Yale University
Lines: 43
Nntp-Posting-Host: algeria.ai.cs.yale.edu
Originator: tarr@algeria.CS.Yale.Edu

Here for all that are interested are the two references I mentioned:


Westall, R. F., Perkey, M. N., & Chute, D. L. (1989).  Millisecond
	timing on the Apple Macintosh: updating Drexel's millitimer.  Behavior
	Research Methods, Instruments, & Computers, 21 (5), 540-547.

Ulrich, R., & Giray, M. (1989).  Time resolution of clocks: effects on
	reaction time measurement -- good news for bad clocks.  British Journal
	of Mathematical and Statistical Psychology, 42, 1-12.

Note that if you are using an ADB input device it is unlikely you are
getting millisecond precision anyway. For my own software I have two
strategies for measuring reaction times:

1. For ADB input devices: Run w/o multifinder and use an eventmask that
only looks for events from the device subjects are responding on. Use
the Event.when field to find out when the subject responded in ticks.
Note that there are 60.15 ticks per second (see IM V). This number is
fixed and is no longer tied to the screen refresh rate. If you need the
videosync to draw flicker free images, you need to use a VBL task which
is specific to your monitor (I still don't have this part working --
help please!).

2. To get millisecond accuracy: Buy a digital I/O card like that
available from Strawberry Tree which has counter/timers which can be
latched to the i/o channels -- this way you get REAL millisecond
precision since the timer stops when the channel goes high (e.g. your
subject presses a key connected to the channel).

Many psychologists are too hyper about millisecond precision -- given
the distributions of most RT's this is unnecessary. Plus few if any are
actually coaxing msec precision out of the computers -- but then they
report it as msecs cause the reviewers are saying "I don't trust your
data unless it is msec precise." Remember millisecond precision is just
as arbitrary as p < .05!!! It is doubtful brain function has little to
do with msecs. (Although it might be 50hz ticks if you believe Francis
Crick and Kristof Koch!!!).
-- 
 * Mike Tarr                                    The Human Neuron Project  *
 * tarr@cs.yale.edu                             Department of Psychology  *
 * "My opinions are always my own."             Yale University           *
 **************************************************************************