On the detection of stepwise changes in a constant signal level.

Abstract

A stepwise change in the frequency of a continuous pure tone is taken as an element signal. The detectability of a single step is compared with the detectability of a pattern composed of two such steps in order to determine the ways in which the steps interact and how these inter­ actions depend upon the time interval between the steps. It is shown that two positive steps separated by D msec interact by summation (d' for two steps being greater than d' for a single step) and that this summation is 100% even when D=100 msec. The extent of summation de­ creases as D is increased beyond 100msec, levelingoff at about 40% when D is near 1,000msec. A positive step followed by a negative step (an increment) presents a more complex picture. When D=O msec, the pattern is the null signal, and, of course, the two steps subtract completely. As D is increased, the extent of subtraction decreases, becomingzero when D is about 400 msec. Increasing D beyond 400 msec results in summation, reaching about 40% when D=2,000 msec. For D less than about 1,000 msec, the sign of the second step in a two-step signal is important: two steps of like sign summate, and two steps of unlike sign subtract. However, when the two steps are separated by about 2,000 msec, the sign of the second step is irrelevant: summation occurs,and the extent, about 40%,is the same for two steps of like sign and two steps of unlike sign. A brief theoretical discussion stresses the extant need for an information-processing theory of signal detection. The stimulus increment that serves as the signal in many detection experiments is a somewhat complex event that probably produces a total sensory effect that is both multidimensional and extended in space and time. An increment having a duration of D msec consists of a positive step as its beginning, followed by a negative step D msec later. The two steps are equal in magnitude but opposite in sign, and they are separated by a time interval during which the signal energy differs from the prevailing energy level (either that prevailing prior to the increment or in other 'regions simultaneously with the increment, or both). As an example, consider a signal that is a small spot of light presented briefly upon a larger background that is continually present. In the region of the target, there is a temporal gradient of luminance at signal onset and a second temporal gradient at signal offset. There is also a spatial gradient at the edge of the spot while it is present. These three stimulus gradients may produce qualitatively different excitatory ef­ fects. A similar example from audition is a brief sinu­