Speech in noise perception in childhood: Role of modulation filtering and processing efficiency
The present study explored the relationship between the capacity to detect amplitude modulation (AM) and speech-in-noise (SIN) identification during childhood. Auditory models suggest that AM detection is not only constrained by the filtering properties of sensory mechanisms in the modulation domain, but also by “processing efficiency”, the ability to make optimal use of the available sensory information. Behavioral tasks were designed to assess the development of modulation filtering and processing efficiency of AM cues and its relationship with SIN between 6 and 8 years of age.
Eighty-two children first completed a 2-alternative-forced choice task (AFC) using an adaptive procedure estimating AM detection thresholds for an 8-Hz sinusoidal AM. In this task, the AM carrier was varied in 2 conditions to assess: i) AM sensitivity using a 500-Hz sine tone (No Masking), and ii) AM masking using a 4-Hz wide narrowband noise centered at 500 Hz with small envelope fluctuations (Masking). Second, a “double-pass technique” evaluated the consistency of children’s responses for AM detection using a constant-stimuli procedure. Then, AM detection performance in the Masking condition was measured at threshold for 200 trials repeated twice (2 passes) using a 2-AFC task. Percentage of Correct AM detection in each pass (PC) and Percentage of Agreement between the 2 passes (PA) were used to estimate within-listener consistency, a proxy of AM processing efficiency related to internal noise. Finally, children completed an XAB adaptive task measuring consonant identification thresholds in noise using fricatives and stops contrasting over three phonetic features (voicing, place, and manner). Additionally, children completed two standardized tests assessing receptive vocabulary and non-verbal reasoning.
Results showed that AM detection thresholds obtained with both carriers did not significantly improve from 6 to 8 years (ps > .37) and all children were similarly affected by AM masking (p < .001). When children were tested at threshold, both PC and PA increased with age (ps = .03). Thus, AM filtering is not affected by age, but aspects of processing efficiency are. Regarding SIN, thresholds significantly improved with age (p = .02) and were affected by phonetic feature (p < .001). Backward regression analyses showed that AM masking associated with vocabulary scores significantly predicted data for Manner (8.9%), PC and PA predicted to a small extent SIN data for Voicing (adjusted R2 = 5.8%) and vocabulary predicted data for Place (5.1%). Overall, processing efficiency, modulation filtering and linguistic level determine SIN identification in childhood.