Some of the strategies which an observer might use in detecting a tone presented against a background of noise are discussed. In Experiment I thresholds for tones in noise were measured under a number of different conditions, using flat−spectrum and notch−filtered noises, in a two−interval forced−choice task. Threshold did not correspond to a constant difference in critical−band levels on the two halves of the trial. Performance at the highest frequency used (6.1 kHz) was worse than at lower frequencies (1055 and 4080 Hz), and, in contrast to them, was not independent of overall level. This is contrary to the classical data, and indicates that the critical ratio is not a constant fraction of the critical band. In Experiment II the intensity discrimination of 1/3−octave bands of noise was studied. The noises were presented alone or in a wide−band noise background. For noise bands presented alone, performance is roughly independent of overall level, and improves slightly with increasing center frequency. For noise bands presented against a wide−band noise background, performance is independent of overall level for frequencies up to 4 kHz. A change in level of about 2−3 dB is necessary for 75% correct detection. For frequencies above this, performance worsens with increasing overall level. These results may be explicable in terms of a saturation of neurons at high intensities. The saturation effects are not observed at low frequencies, possibly because the primary cue for detection is a change in the temporal pattern of neural firing, rather than a change in amount of neural firing.
Subject Classification: 65.58, 65,35.