Filling the blanks of checkerboard speech with noise: Evidence for phonemic restoration and maskinga)

Sixteen-band checkerboard speech (interrupted in time and frequency) is perfectly intelligible. Whereas two- and four-band checkerboard speech is usually less intelligible than speech interrupted only in time [Ueda et al., J. Acoust. Soc. Am. 154(4), 2010-2020 (2023)]. The intelligibility of speech interrupted only in time improves when the gaps are filled with intense noise. However, the effects of filling the blanks of checkerboard speech with noise on intelligibility are largely unknown. Here, we show that such filling results in benefits and losses in intelligibility (N=20). The intelligibility of two- and four-band stimuli improved by up to 40%, except for the 20 ms segment duration. Surprisingly, signal-to-noise ratios (SNRs) in the ±6-dB range had little or no effect on the benefits. By contrast, the intelligibility of eight- and 16-band stimuli was reduced by up to 70% with a -6-dB SNR and 80 ms segment duration. Phonemic restoration occurred for two- and four-band stimuli filled with noise. In contrast, energetic masking dominated the results for eight- and 16-band stimuli. These results suggest that the four frequency bands important for speech perception may explain the benefits and losses in intelligibility when checkerboard speech is filled with noise.