^{1}and Brian C. J. Moore

^{1,a)}

### Abstract

The dominant region for pitch for complex tones with low fundamental frequency (F0) was investigated. Thresholds for detection of a change in F0 (F0DLs) were measured for a group of harmonics (group B) embedded in a group of fixed non-overlapping harmonics (group A) with the same mean F0. It was assumed that F0DLs would be smallest when the harmonics in group B fell in the dominant region. The rank of the lowest harmonic in group B, N, was varied from 1 to 15. When all components had the same level, F0DLs increased with increasing N, but the increase started at a lower value of N for F0 = 200 Hz than for F0 = 50 or 100 Hz, the opposite of what would be expected if the dominant region corresponds to resolved harmonics. When the component levels followed an equal-loudness contour, F0DLs for F0 = 50 Hz were lowest for N = 1, but overall performance was much worse than for equal-level components, suggesting that the lowest harmonics were masking information from the higher harmonics.

This work was supported by Deafness Research UK and by the Medical Research Council (Grant No. G0701870). We thank Brian Glasberg for help with signal processing and figures and Peter Cariani for comments on an earlier version of this paper.

I. INTRODUCTION

II. METHOD

A. Subjects and training

B. Stimuli

C. Procedure

D. Conditions for each experiment

III. RESULTS

A. Experiment 1

B. Experiment 2

C. Comparison of results for experiments 1 and 2

IV. SUPPLEMENTARY EXPERIMENTS

A. Subjects

B. Conditions

C. Results

V. DISCUSSION

VI. SUMMARY AND CONCLUSIONS

### Key Topics

- Sound discrimination
- 98.0
- Timbre
- 20.0
- Auditory system
- 7.0
- Pitch
- 7.0
- Sound pressure
- 5.0

## Figures

Schematic spectra for four example stimuli for F0 = 100 Hz. For these examples, the value of N was 3 (top) or 7 (bottom), and the value of nB was 4 (left) or 8 (right). In the figure, the harmonics in group B occur at integer multiples of 100 Hz. In the actual experiment, the F0 of the harmonics in group B was shifted upward by 0.5ΔF in one interval of a trial and downward by 0.5ΔF in the other interval. The value of ΔF was varied to estimate the F0DL.

Schematic spectra for four example stimuli for F0 = 100 Hz. For these examples, the value of N was 3 (top) or 7 (bottom), and the value of nB was 4 (left) or 8 (right). In the figure, the harmonics in group B occur at integer multiples of 100 Hz. In the actual experiment, the F0 of the harmonics in group B was shifted upward by 0.5ΔF in one interval of a trial and downward by 0.5ΔF in the other interval. The value of ΔF was varied to estimate the F0DL.

Geometric mean F0DLs for experiment 1 for F0 = 50 Hz (top) and 200 Hz (bottom). The F0DL is plotted as a percentage of F0 as a function of the harmonic rank, N, of the lowest harmonic in group B. The frequency of the lowest component in group B is shown at the top. Parameters are the starting phase of the components (filled symbols = random, open symbols = cosine) and the number of components, nB, in group B (squares = 4, triangles = 8). Error bars indicate ± one standard error of the mean. The vertical dashed line indicates the boundary between resolved and unresolved harmonics based on the assumption that a harmonic needs to be separated from neighboring harmonics by 1.25ERBN or more for it to be resolved.

Geometric mean F0DLs for experiment 1 for F0 = 50 Hz (top) and 200 Hz (bottom). The F0DL is plotted as a percentage of F0 as a function of the harmonic rank, N, of the lowest harmonic in group B. The frequency of the lowest component in group B is shown at the top. Parameters are the starting phase of the components (filled symbols = random, open symbols = cosine) and the number of components, nB, in group B (squares = 4, triangles = 8). Error bars indicate ± one standard error of the mean. The vertical dashed line indicates the boundary between resolved and unresolved harmonics based on the assumption that a harmonic needs to be separated from neighboring harmonics by 1.25ERBN or more for it to be resolved.

As Fig. 2 , but for experiment 2 and for F0 = 100 and 200 Hz.

Comparison of the data for experiments 1 and 2 for the cosine-phase stimuli with F0 = 200 Hz. Parameters are nB (circles = 4, diamonds = 8) and whether PPA cues were effective (experiment 1, filled symbols) or ineffective (experiment 2, open symbols). Otherwise as Fig. 2 .

Comparison of the data for experiments 1 and 2 for the cosine-phase stimuli with F0 = 200 Hz. Parameters are nB (circles = 4, diamonds = 8) and whether PPA cues were effective (experiment 1, filled symbols) or ineffective (experiment 2, open symbols). Otherwise as Fig. 2 .

F0DLs from the supplementary experiment for the following conditions: (1) With the levels of the lowest 15 components selected so that each would have a loudness level of 40 phons if presented in isolation, without any masker (filled triangles); (2) with the levels of the components in groups A and B selected as in (1), with narrowband noise in the spectral gaps (open triangles); (3) with the levels of the components in groups A and B equal, with narrowband noise in the spectral gaps (open squares). Data from experiment 1 for F0 = 50 Hz, nB = 4, random phase, equal-level components are shown by filled squares.

F0DLs from the supplementary experiment for the following conditions: (1) With the levels of the lowest 15 components selected so that each would have a loudness level of 40 phons if presented in isolation, without any masker (filled triangles); (2) with the levels of the components in groups A and B selected as in (1), with narrowband noise in the spectral gaps (open triangles); (3) with the levels of the components in groups A and B equal, with narrowband noise in the spectral gaps (open squares). Data from experiment 1 for F0 = 50 Hz, nB = 4, random phase, equal-level components are shown by filled squares.

F0DLs from the supplementary experiment for random- and cosine-phase conditions. In both cases, the levels of the lowest 15 components were selected so that each would have a loudness level of 40 phons with narrowband noise in the spectral gaps.

F0DLs from the supplementary experiment for random- and cosine-phase conditions. In both cases, the levels of the lowest 15 components were selected so that each would have a loudness level of 40 phons with narrowband noise in the spectral gaps.

F0DLs plotted against the width of group B expressed in ERBN-number, with unit Cams.

F0DLs plotted against the width of group B expressed in ERBN-number, with unit Cams.

Article metrics loading...

Full text loading...

Commenting has been disabled for this content