A tachistoscope is a device that has been used by visual psychologists for many years to present stimuli of differing durations to the subject. Its three essential elements are a constant (non-pulsed) light source, a slide containing the desired image such as a shape or word and a shutter mechanism to control the illumination of the slide by the light.  By adjusting the shutter, the image on the slide can be exposed to the subject for variable lengths of time. In a typical experiment, the tachistoscope is used to show the subject a word for increasing amounts of time until he can just read it. Because of the inflexibility of these devices inherent in their mechanical nature it would obviously be very useful to be able to reproduce the functionality of the tachistoscope using computer generated images which could be presented for computer controlled lengths of time. This would even allow more complicated experiments involving such techniques as masking to be designed.

At first sight it seems that a CRT based display would not be suitable for such a system as the frame rate inherently limits the time resolution to about 10ms which is probably ten times too low to be useful. As we shall see however, this is an erroneous analysis based on several misunderstandings and a computer system can be used as a very successful tachistoscope. To help the explanation it is useful to know a little about the response of the eye to pulses of light such as those emitted by a CRT display but there is no need to consider the brain or assume anything about higher order cognitive functions. Many people have described and modelled the impulse response of the eye, which technically is the response to a pulse of light of infinitesimal duration and unit area but is effectively the response to the short pulse  (of order 1ms) produced by a CRT display as each area of the screen is scanned. There is general agreement that it consists of a positive response followed by a slightly delayed negative response with the whole thing happening over a period of about 150ms which we shall call the natural time . In addition to the shape of the response it is generally accepted that the eye can be considered under most conditions as a linear system so that the response to an arbitrary input is simply the convolution of the input with the impulse response. In other words, to find out what happens when we expose the eye to two successive pulses of light from two consecutive frames from a display we can simply add together two impulse responses with the second being delayed by one frame time. The situation is shown in Figure 10 below. One curve shows the response to a single displayed frame while the second curve shows the response to two successive frames.

Figure 10 Visual response to short duration pulse stimuli

The most obvious feature about the two curves is not that one is twice as long as the other which might have been expected from the fact that it was created by displaying two frames rather than one but that the response to two pulses is nearly twice as large as that caused by the single pulse. In fact it can be shown that the response to twin pulses depends little on the width of the pulses and little on their separation (i.e. frame rate) and a lot on their aggregate area which is a measure of the stimulation energy. Even the small difference in phase between the two curves shown can be removed if time is measured from the centroid of the stimulation rather than the start. In situations like this in which the stimulation time is significantly shorter than the natural time of the eye all we achieve by increasing the stimulation time is to increase the magnitude of the response and it makes little difference if the stimulation is provided by a uniform light source such as found in a tachistoscope or a pulsed light source such as that provided by a CRT. If you still think that the stimulus duration is important in this situation then the energy input can be maintained at a constant level simply by adjusting the luminance downwards as the stimulus duration is increased.

Let us now consider a different condition where the stimulation time is fairly long compared to the natural time of the eye. This situation is illustrated in Figure 11 below for stimulations of 14 frames, 15 frames and 15 frames where the last frame is at half luminance. A frame rate of 100Hz was used  so this represents the response to stimuli of about 150ms. As can be clearly seen, the shapes of these curves are very different to those produced by short duration stimuli and show three distinct regions: the initial turn-on transient, a steady state regime and a turn-off transient. The turn- on and turn-off transients are the same for all three stimulations while only the length of the steady state region depends on the stimulation time. The ripple that can be seen in the steady state region is the only legacy of the pulsed input and this can be shown to be reduced as the frame rate is increased beyond 100Hz. The most interesting fact to notice is that the turn-off transient starts at a time determined by the length of the stimulation; increase the stimulation time by one frame or 10ms and the turn-off transient begins 10ms later. More interesting even than that perhaps is the fact that making the fifteenth and final frame half luminance produces a response that occurs half way in time between that caused by fourteen and fifteen frames at full luminance. This means that by controlling the luminance of the final frame in the sequence we can effectively change the stimulus duration smoothly between 140 and 150ms. If we use a monitor with 100Hz frame rate and have 256 linearly spaced luminance levels available our temporal resolution can be increased from 10ms to 40us and hey presto we have a high resolution tachistoscope.

One question that is often asked is How can I present a mask followed by my special image with intervening times of less than one frame on a CRT display? The answer is simple: adjust the luminance of the last frame of the mask before the onset of the experimental stimulus as described above. By varying the luminance from nothing to full, the last frame time can be traversed with high resolution. When doing this there may be one frame that has information from both the mask and the stimulus on it and these can just be summed but remember that the result must still be within the linear range of the display.

Figure 11 Visual response to long duration stimuli

^ Back to top