The probe-signal method (Greenberg & Larkin, 1968) was used to determine the percentage of trials in which unpracticed subjects detected (two-interval, forced-choice)a soft, expected sound as compared with an unexpected sound. Pure tones at or near an expected frequency were detected in about 90% of the trials. Tones more than one-half critical band away were detected near chance (50%). Complex sounds (a band of noise or a multitone complex) were detected better if they were inside the same critical band as the expected signal than if they were outside the band. A signal that differed spectrally from the expected sound was not detected even though it had the same low pitch, based on a common fundamental frequency. The results may mean that under some conditions focused attention alters sensitivity in the auditory system.

When attending to a tone at a given frequency, listeners are most sensitive to that tone and others within a restricted band of frequencies surrounding it. This region of enhanced sensitivity defines the attention band that was measured in two experiments using a modified version of the probe-signal method of Greenberg and Larkin [J. Acoust. Soc. Am. 44, 1513-1523 (1968)]. Experiment 1 showed that at five center frequencies, from 0.25 to 4.0 kHz, the shape of the attention band resembles that of the auditory filter as inferred from notched-noise masking experiments by other investigators. The width of the attention band is close to the critical band at higher frequencies, but only half as wide at 0.25 and 0.5 kHz. Experiment 2 produced psychometric functions for unattended probe tones at least 0.23 kHz away from a fully attended, 1-kHz target tone. From these functions, the effective attenuation, measured as the threshold difference between the 1-kHz target and the probes, was estimated to be 7 dB; the amount of attenuation appeared to be about the same regardless of how far the probe frequency was from the attended band. One interpretation of these results is that bands centered on the unattended tones contribute to the decision process with some small but measurable weight and are not entirely ignored.

A psychophysical model is presented that explains why loudness summates across frequency as it does and that permits the precise calculation of loudness from the physical spectrum. Loudness is represented by geometrical patterns derived from the masking of pure tones by narrow bands of noise. The masking patterns are converted to loudness patterns by means of the critical-band function that relates tonalness in Barks to frequency in cycles per second and a power function that relates specific loudness, loudness per Bark, to sound pressure level (SPL). Plotted on the coordinates of specific loudness and tonalness, the geometrical patterns are integrated to yield a value in soneso for the overall loudness. Calculated values are compared to experimental values obtained from loudness balances with 3 types of sound.

No abstract