Psychophysics: Detection

A detection threshold is the minimum stimulus strength that can be perceived. Even with such a humble definition, the measurement of detection thresholds has become a golden, powerful tool for psychophysicists, but it is elusive to pinpoint. Some examples of visual detection tasks discussed in this entry are:

• Detection of a tumor. In order to diagnose cancerous tumors, radiologists must learn to detect their presence in a complex X-ray.
• Detection of a feature. The ability to detect a gap in the letter C can indicate whether new eyeglasses are necessary, or whether, even with new eyeglasses, one is fit to drive a car.
• Detection of a needle in a haystack. Many real-life detection tasks are similar to finding Waldo in a crowd or a needle in a haystack. Careful analysis of how observers detect small targets in a cluttered environment can provide insight into whether brains can process many stimuli simultaneously and into how uncertainty limits visibility.
• Detection of a thin line. The timing of certain Jewish and Islamic religious holidays is based on the earliest detection of a new crescent moon by religious authorities using unaided vision.
• Detection of a star. A ritual requiring the bride and groom to detect the double star pair Mizar and Alcor in the handle of the Big Dipper is a part of many Hindu wedding ceremonies. Successful sighting portends a successful marriage.
• Detecting a minimal shift in the position of a target feature. This can be understood in terms of detecting a thin line, like the fourth item in this list, and as discussed in the following paragraphs.
• Detection of a glitch in a movie. Twenty years ago, video engineers were convinced that high resolution digital TV was impossible because the bandwidth they believed was necessary was hundreds of times larger than was feasible. However, vision scientists showed that limits on detection of tiny features, especially in clutter, enables dramatic image compression, making high definition digital TV possible.

We will come back to some of these examples later. But first we must take a detour into how detection thresholds can be measured.

A common method for measuring detection thresholds is the “Yes-No” method. In this method, [Page 846]the observer is shown, in random order, repetitions of a variety of stimulus intensities, ranging in strength from zero to clearly detectable. The task is to say yes or no according to whether a target was seen. The outcome of the experiment can be presented as an S-shaped plot (Figure 1a) showing the percent of times the person said yes, as a function of signal strength (s). The vertical axis is often called the hit rate (phit). This plot of the psychological response to a physical stimulus, called the psychometric function, is fundamental to the field of study called psychophysics.

The lowest point on the curve, called the false alarm rate (pfalsealarm) represents the probability of the observer saying “yes” even when no stimulus is present. The value of pfalsealarm (= 6.7% in the Figure 1a example) can be easily shifted by the observer's intentions. We now consider three possible definitions of detection threshold, shown as dots on the plot.

...