The processing of which-questions in noise-vocoded speech
Whereas children with cochlear implants (CIs) score like their age matched peers on vocabulary, their understanding and use of grammatical aspects of spoken language lacks behind (Nikolopoulos, Dyar, Achbold & O’Donoghue, 2004; Friedmann & Szterman, 2011). In German, grammatical aspects such as case and verb-agreement are crucial for correct interpretation of object questions (in which the object precedes the subject). In previous research, we found that children with cochlear implants, when interpreting object questions make less use of case (e.g. Welchen Esel fängt der Tiger ‘Which donkey is the tiger catching?’) and/or verb-agreement cues (e.g. Welche Maus fangen die Frösche? ‘Which mouse are the frogs catching?’) than children with normal hearing (Schouwenaars et al., in prep). One explanation for these interpretation problems is the input quality (degraded speech input by the CI), whereas another explanation is the lack of input during a so-called sensitive period. The speech input by CIs can be simulated by noise-vocoded speech and its effect can be tested on normal hearing populations (e.g. Scott 2005). A first step and the aim of this study is to investigate whether adults with normal hearing show similar problems when interpreting which-questions in noise-vocoded speech as the children with CIs.
We tested 30 adults with noise-vocoded stimuli (simulated CI listeners) and 30 adults with the original stimuli (control group listeners). The noise-vocoded stimuli were generated by a software-implemented vocoder designed to simulate a 22 channel implant type (Langner and Jürgens, 2016). A picture selection task with eye-tracking was carried out to test comprehension of subject-, object- and passive questions. Two additional screening tasks tested comprehension of verb-agreement in canonical sentences and auditory discrimination of case (e.g. der vs den).
The simulated CI listeners (only those who performed well on the screening tasks) comprehended object questions worse than the control group listeners (GLMER: p<0.001; simulated CI listeners: 83%; control group listeners: 97%). The comprehension of subject- and passive questions was not decreased by the simulated CI input. At the moment, the gaze data is being analyzed to gain more insight into the online processing of which-questions by simulated CI input.
The offline results of the simulated CI listeners show a similar pattern compared to children with actual CIs, in that only the performance on object questions was decreased to a similar extent. This suggests that the degraded speech input of the CIs plays a big role in the explanation of children with CIs’ problems with object questions.