Coherence masking protection for speech in children and adults

Abstract

In three experiments, we tested the hypothesis that children are more obliged than adults to fuse components of speech signals and asked whether the principle of harmonicity could explain the effect or whether it is, instead, due to children’s implementing speech-based mechanisms. Coherence masking protection (CMP) was used, which involves labeling a phonetically relevant formant (the target) presented in noise, either alone or in combination with a stable spectral band (the cosignal) that provides no additional information about phonetic identity and is well outside the critical band of the target. Adults and children (8 and 5 years old) heard stimuli that were either synthetic speech or hybrids consisting of sine wave targets and synthetic cosignals. The target and cosignal either shared a common harmonic structure or did not. An adaptive procedure located listeners’ thresholds for accurate labeling. Lower thresholds when the cosignal is present indicate CMP. Younger children demonstrated CMP effects that were both larger in magnitude and less susceptible to disruptions in harmonicity than those observed for adults. The conclusion was that children are obliged to integrate spectral components of speech signals, a perceptual strategy based on their recognition of when all components come from the same generator.

Explaining coherence in coherence masking protection for adults and children

“…An earlier study by the authors revealed greater CMP for children than adults, with more resistance to disruptions in harmonicity across spectral components [Nittrouer and Tarr (2011). Atten.…”

“…From a different perspective, however, that earlier outcome might be described as revealing how strongly children are obliged to process speech signals as integrated spectral patterns: It is not so much that children benefited more than adults from the high-frequency cosignal as it is that children were disadvantaged when it was not present. Cohen's d's computed on CMP scores for adults and 5-yr-olds in Nittrouer and Tarr (2011) were 2.49 for the F1-only condition and 1.40 for the F1 þ cosignal condition, meaning that age-related differences were greater for the F1-only condition than for the F1 þ cosignal condition. Thus, it seems fair to suggest that children required the broad spectral pattern for judging vowel quality, whereas adults were able to make do with F1 alone, to some extent.…”

Release from masking of low-frequency complex tones by high-frequency complex tone cue bands

“…This experimental setup is similar to experiments that demonstrated a phenomenon termed coherence masking protection (CMP; Gordon, 1997aGordon, ,b, 2000Tarr and Nittrouer, 2011;Nittrouer and Tarr, 2011). In those experiments, a masked LF target signal was presented that was either a single tone (Gordon, 2000), a narrowband noise (Gordon, 1997b), or a low-pass filtered synthesized vowel (Gordon, 1997a;Nittrouer and Tarr, 2011).…”

“…These basic conditions were compared with conditions in which an additional "cosignal" was presented in higher frequency regions. This cosignal was either a high-pass filtered synthesized vowel (Gordon, 1997a(Gordon, ,b, 2000Nittrouer and Tarr, 2011), a complex tone (Gordon, 2000), or a noise band (Gordon, 2000(Gordon, , 1997b. If the cosignals were presented with the same onand offset as the target signal, the identification thresholds were found to be 3-4 dB lower than in conditions without a cosignal or with a temporally separated cosignal.…”

“…In those experiments, a masked LF target signal was presented that was either a single tone (Gordon, 2000), a narrowband noise (Gordon, 1997b), or a low-pass filtered synthesized vowel (Gordon, 1997a;Nittrouer and Tarr, 2011). For each target signal, either a LF or a HF signal was presented in a one-interval task, and the subjects were asked to identify the presented signal as low or high pitched for the single tone or narrow-band noise target, or as one of two vowels for the low-pass filtered vowel target.…”