Background: Loudness is typically reduced in the presence of other sounds. When both target and masker are presented diotically or from the same loudspeaker, partial loudness can be estimated well by considering energetic masking. For speech intelligibility, spatial separation of sounds improves intelligibility, even though the degree of spectral masking remains largely unchanged. This is known as the cocktail party effect. In the present study we investigated this effect on partial loudness by varying the spatial separation between a female target speaker and a competing male speaker.

Method: The target speech was positioned at 0° and 45°. In the respective reference conditions, the competing speech was positioned at the same angle and had the same sound pressure level as the target. Comparisons were made by twenty participants for spatial separations of 22.5°, 45°, and 90° using a 1-up/1-down two-alternatives forced-choice procedure. When the target was at 45°, the comparison stimulus was positioned towards or crossing the centre. Sounds were recorded with an artificial head and presented via headphones. After combining the target and masker, the signals to the left and right ear were switched randomly to reduce a-priori focus on the target. All stimuli had a duration of 2 seconds.

Results: For the 0° target, level differences required for equal loudness were close to 0 dB, implying that the target was perceived as equally loud when the masker coincided in direction and when it was positioned on the side. For the 45° target, the effect of spatial separation was small, less than 1.5 dB, but statistically significant, and increased with separation.

Conclusions: The level difference necessary for equal loudness (LDEL) was small compared to previous speech intelligibility studies that used similar separations and found differences in speech reception thresholds larger than 5 dB. The lack of any effect for the frontal reference was surprising. It may be that participants always knew where the target stimulus came from, which could have facilitated attention. Another explanation for the small, if any, effect in both conditions could be that participants based their judgment on segments when the target was loud and between syllables or words of the masker. This would suggest participants effectively judged partial loudness at a high signal-to-noise ratio, similar to loudness in silence, and thus is less dependent on the masker.

Acknowledgments: This study was supported by the Royal Society, grant 232164.