Speech-on-Speech Masking: Effect of Maskers with Different Degrees of Linguistic Information

The current study measured speech recognition and subjective listening effort scores while systematically varying the amount of linguistic information in maskers. Linguistic information in the maskers was varied by (a) increasing the number of speakers in the speech babble maskers and (b) time-reversing them. In Experiment 1, we measured speech recognition performance (signal-to-noise ratios required for 50% accuracy of sentences) for 16 participants. The speech (sentences) recognition scores were obtained in 15 background conditions: speech babble maskers with 2 to 8 speakers (7 conditions), time-reversed babble maskers (7 conditions), and a speech-spectrum noise. For Experiment 2, another 15 participants rated the effort (7-point rating scale) required to understand sentences in the same maskers as Experiment 1. This was done at a signal-to-noise ratio of 0 dB. Results showed that fewer speakers in the babble maskers (a) caused the greatest masking effects and (b) required the greatest listening effort ratings. Speech babble maskers resulted in significantly higher masking effects than reverse babble maskers only for the 2-and 3-speaker babble conditions. However, the listening effort scores were substantially higher for the speech babble maskers than reverse babble maskers in most of the conditions. Results suggest that both magnitudes of masking and the listening effort scores are related to the linguistic information in the masker.