Assessing bottom-up and top-down effects on speech-in-noise recognition by adolescents with normal hearing or cochlear implants

The terms “bottom-up” and “top-down” effects are commonly used when measuring speech recognition in competing sounds to refer to listeners’ abilities to exploit details of the sensory input and to use linguistic constraints, respectively. The magnitude of each effect has been assessed for listeners with cochlear implants, but typically in separate studies. We varied the availability of bottom-up and top-down information in a single experiment by generating sentences with three kinds of clause structure, from less to more restrictive, in three different babbled backgrounds, with sex and number of talkers varied. Next we examined the factors that might explain abilities to use bottom-up and top-down information in speech-in-noise recognition. Predictor variables included spectral modulation detection abilities, lexical-syntactic knowledge, verbal working memory, and phonological sensitivity. Participants were 14-year-olds with normal hearing (NH; 46) or with cochlear implants (CIs; 46). Adolescents with cochlear implants showed deficits in their abilities both to segregate the target speech from background babble and to benefit from restrictive syntactic constructions. For adolescents with NH, only verbal working memory explained variability in abilities to use bottom-up or top-down information in speech-in-noise recognition. For adolescents with CIs, only spectral modulation detection abilities explained their abilities to apply these effects.