Discussion

2.4.1

Acoustic differences between stimuli

The word-medial /S/ segments were found to be shorter in duration and to have faster rise-times compared to the word-initial /S/ segments. This could reflect the effects of co-articulation in which adjacent phonemes in running speech may be shorter and may exhibit blurred phonemic boundaries, compared to phonemes articulated in isolation (Kent & Read, 2002). The onset is more gradual when the phoneme is spoken at the beginning of the utterance where the transition is from silence to speech as in the case of /Si/. The initiation of articulator motion could contribute to the slow onset. Spectrally, the abrupt /S/ contains more energy at low and mid frequencies compared to slow /S/. This could also be due to co-articulation (Kent & Read, 2002; Pickett, 1999). With respect to /S/, articulation of /S/ in /aSIl/ is not only influenced by the following vowel but also /l/ at the end. In contrast, /S/ in /Si/ is influenced by the following vowel only. These spectral differences evident in the stimuli chosen in this study are bound to vary with

different neighboring phonemes. Another factor that may contribute to the broader spectrum in the case of the abrupt /S/ is a faster rise-time (Burkard, 1984).

2.4.2

Effect of stimulus on CAEP peak latencies and

amplitude

The abrupt /S/ elicited CAEPs with significantly larger N1-P2 amplitude and shorter peak latencies compared to the slow /S/. This can be explained by the known effects of stimulus rise-time on CAEP. To our knowledge, studies that have evaluated the effects of rise-time on CAEP have only used tone bursts. This is probably because their rise-time can be precisely manipulated.

The general consensus among studies that investigated the effect of tone burst rise-time on CAEPs is that shorter rise-times lead to larger amplitudes and shorter latencies (Cody & Klass, 1968; Lamb & Graham, 1967; Onishi & Davis, 1968; Prasher, 1980; Skinner & Jones, 1968; Thomson, Goswami, & Baldeweg, 2009). This can be explained by the effect of stimulus rise-time on neural synchrony. Longer rise-time leads to increased jitter in neuron firing. The increased jitter may reflect inconsistent trigger points along the onset envelope of the tone burst (Picton, 2011, pages 335–343). Increased jitter leads to reduced neural synchrony, which results in broader peaks with lower amplitude and longer latency (Goldstein & Nelson, 1958; Hyde, 1997; Onishi & Davis, 1968). The effect of longer stimulus rise-time is more detrimental for earlier potentials, such as the compound action potential and auditory brainstem response (Arnold, 2007; Goldstein & Nelson, 1958). Even small increases in rise-time (e.g., from 2.5 ms to 5 ms) can abolish the compound action potential due to inadequate synchronous firing. In contrast, CAEPs can be elicited by longer stimulus rise-times that abolish the earlier potentials.

N1 latency based on behavioral detection threshold along the time course of tone burst onset ramp found that the N1 latency remained unchanged up to increases in rise-time of 100 ms. This implies that any stimulus-related activity is triggered only after the onset ramp is above the individuals threshold (Ruhm & Jansen, 1969). The effects of rise-time have also been explained by the theory of temporal

integration, where detection thresholds vary as a function of stimulus duration and intensity. As rise-time is increased beyond 30 ms, the effective intensity at the onset of the stimulus is reduced (Onishi & Davis, 1968).

The results of the present study are in agreement with the literature discussed above. This implies that the use of phonemes from different sources could potentially lead to differences in the interpretation of CAEPs. For example, it is possible that stimulus choice may also affect CAEP procedures such as threshold estimation, by facilitating detection of larger amplitude CAEPs. Although larger amplitude responses can be looked at as a stimulus advantage, word-medial

phonemes may have a spectral disadvantage. Co-articulatory effects and onset rate appear to contribute low frequency energy, reducing the frequency specificity of the stimulus relative to the word-initial position. This may be especially important to consider when using natural speech stimuli for the purposes of hearing aid

validation. These effects were studied for the phoneme /S/ in this experiment, but whether such effects generalize to other phonemes remains to be studied.

As explained above, the differences in waveforms elicited by the two variants of /S/ may be attributed to the sensitivity of CAEP to onset characteristics of the stimuli. Differences in the CAEP waveforms may not be interpreted as representation of other processes (e.g., discrimination) based on the present study. Hearing aid signal processing (e.g., compression and compression attack time) may interact with the temporal properties of natural speech stimuli, but such effects have not yet been

evaluated in detail. The present study evaluated the effects of rise-time in

individuals with audiometric thresholds within the normal range, and provides some evidence that stimuli with a more abrupt onset may provide larger amplitude

CAEPs. However, the effects of rise-time on CAEP in individuals with hearing impairment, specifically cochlear hearing loss may differ from that in individuals with normal audiometric thresholds due to recruitment (e.g., Moore, 2007).

2.4.3

Effect of stimulus on test-retest reliability

In the present study, mean ICCs of 0.77 (SE = ±0.03) and 0.72 (SE = ±0.04) were obtained for CAEPs elicited to abrupt and slow /S/, respectively. Test-retest

reliability of speech-evoked CAEPs assessed in terms of ICC has been previously reported in Tremblay et al. (2003). Although direct comparisons between the two studies cannot be made due to differences in stimuli (consonant vowel combinations were used in Tremblay et al., 2003), it is interesting to note that mean ICCs

reported by Tremblay et al. (2003) across stimuli were higher, ranging between 0.8 and 0.9. Factors such as number of sweeps in an average and an exclusive channel for eye blink rejection used in Tremblay et al.’s study could possibly explain this difference. Specifically, the use of more sweeps and rejection of epochs with eye blinks in the Tremblay et al. study may have allowed better quality responses and more repeatable waveforms. The present study illustrates that reliable

speech-evoked CAEPs can be obtained in normal hearing participants using a shorter protocol and a single-channel clinical system without eye blink rejection while maintaining reasonable test-retest reliability.