Processing speed and age predict recognition of spectro-temporally degraded speech

Abstract

Cochlear implants (CIs) provide a listener with the temporal envelope of an acoustic signal, but not the spectral fine structure. Speech recognition performance is highly variable among individuals and is often assessed clinically in ideal conditions (over-articulated, clean speech) rather than real-world listening environments. This study examined the effects of age and processing speed on understanding of time-compressed speech by listeners with normal hearing (NH) presented with vocoded speech of varying spectral resolution and listeners with CIs. NH listeners repeated sentences that had been time-compressed to varying degrees (no compression, 20%, 40%, or 60% time compression ratios) and then noise vocoded with various channels (4, 8, or 16 channels or unprocessed), while CI participants repeated unprocessed time-compressed sentences. Performance of age-matched NH listeners on the 4-channel vocoded speech most closely matched CI listeners’ performance at each level of time-compression. For NH listeners, age effects were observed on conditions featuring time compression and reduced spectral resolution. Both NH and CI listener groups showed a significant interaction between age and time compression. As difficulty increased (greater time compression for all, fewer channels for NH), cognitive measures of processing speed were correlated with speech recognition performance across all age and listening groups.