Computer Based Testing: A Question of Time - Andrew Linn

We used Macromedia Director to create these programs. For timing, we used Director's "the Milliseconds" command. When a computer is turned on, the internal clock starts counting the passage of time in milliseconds. The Milliseconds command reads the current number of milliseconds that have passed since the computer was turned on. Very accurate, (down to the millisecond) and available on both Macs and PCs.

When we wanted to begin timing, the program set a variable to the current number of milliseconds, something like "BeginTime = the Milliseconds." When we wanted to stop timing, the program would issue something like "EndTime = the Milliseconds." The it does the math: EndTime - BeginTime = TimePassed.

The problem is accurately getting the computer to recognize when the EndTime command should be issued and processed. We needed user input, and so required some way for the user to signal the computer that something needed to happen. We chose the keyboard as the most common, most readily available user interface, and specified various keys as indicators of correct/incorrect answers.

When a key is pressed, a signal, called an interrupt, is sent to the CPU. In a multitasking environment such as Windows or the Mac OS, the CUP might or might not be busy at any given time, and the interrupt may or may not be processed the instant it is sent. This variances in timing may be substantial enough to skew data, or if the computers are sufficiently powerful the variance might be too small to notice.

There are literally dozens if not hundreds of variables that might or might not affect this timing: CPU speed, amount of memory, other processes running concurrently, changes in the operating system, variances in keyboards, variances in the components that make up the computer itself, who knows what all, perhaps even ambient temperatures.

The long and the short of it is: I don't know. Tests could be devised to determine the variances, but only under certain circumstances,