Various methods of implementing a computer-based tachistoscope are compared. Software considerations are emphasized, but hardward issues are also examined. Three types of systems are identified: file-driven systems, sets of callable subroutines, and special languages. All have advantages in some circumstances. A system of callable subroutines is favored, together with a file-driven system written in the host language using the subroutines.Computer-based tachistoscopes come in a wide variety of styles and sizes, as the preceding articles have demonstrated. Each system has advantages and limitations. Each style has advocates and skeptics. Each size has different capabilities, and a different price tag. Here, we provide a comparative evaluation of the types of systems and discuss some considerations involved in planning or improving a system.We should first note that doing a standard tachistoscopic experiment on a computer is logically trivial. The standard noncomputer experiment uses either cards viewed through an optical system or slides projected on a screen; the control processes involved in this experiment can be readily transferred to a computer and need relatively little of the computer's power. The main problem is creating the stimulus display. The stimuli in the mechanical and photographic tachistoscopes could be anything from a few capital letters to a page of text, or a complex black-and-white figure, or even a scene requiring several brightness levels or colors, that is, a picture. Computer-based tachistoscopes can usually generate text and complex figures, but typically do not generate either color displays or displays requiring gray-level differences. Color is possible but requires equipment not common on laboratory computers, and it is not discussed here. Many computer displays do permit different gray levels, but the systems discussed here make no provision for different brightnesses because there is no call for it. Nearly all of the experiments actually done use black-and-white figures, usually letters or numbers in a linear array, that can easily be provided by the computer.Actually, most experiments need nothing more than a "mechanical horse." (If the fledgling automobile makers at the turn of the century had duplicated mechanically the then-current form of transportation, they would have built a mechanical horse. TheWe thank the authors of the preceding papers in this section for their wise comments, which were always attended to, but not always heeded. Hence, the opinions that are expressed here can only be attributed to us. 789 mechanical horse syndrome-doing by computer exactly what was done without a computer-is rife in many computer applications.) The traditional experiment offers very little basis for choosing among computerbased tachistoscope systems; they are all quite adequate mechanical horses. Comparisons among systems are possible, and are made here, but we shall also be interested in what the systems have in reserve, for doing the experiments that only a computer can do. Sperling (N...