Statistical analysis

The latencies of the various ELAR responses were compared using 2-way ANOVAs (electrode site / duration of deafness; M levels / duration of deafness; audiogram type / duration of deafness). To perform these tests, the parameters cited above were separated into groups in such a way that each group contained the same number of data or subjects. For the M level factor, groups M1 and M2 comprised subjects with M levels below and above 122 CU respectively. The audiometric and duration of deafness groups presented in Table I were used. It is worth noting that groups 1 and 2 for the duration of auditory deprivation, audiometric performance, and M levels do not contain the same subjects and were shown to be statistically unrelated by the Chi-square test (p > 0.05 for all three combinations). Different aspects of deafness were therefore investigated. It is also worth noting that all subjects in deafness duration group D1 have better speech scores than subjects in group D2 (Table 1). This finding is in line with findings by Blamey et al. (1996) who showed that speech perception performance is related to the duration of deafness. The potential relationship of ELAR latency with deafness duration that may be found would therefore also be applicable to speech perception. However, the duration of deafness may be the dominant factor: Poorer benefit in long term deafened cochlear implants users may indeed be partly due to the effect of long duration of deafness on sound processing delays. Differences between M levels according to electrode sites were also analysed using 1-way ANOVA to check whether neural degeneration of auditory fibers results in a gradient that should be taken into account in interpretation of the potential effect of electrode site on ELAR latency. Given that the perception of high pitch sounds is generally lost first, M levels could indeed be higher at the base than at the apex. The possible finding of an effect of stimulation site on ELAR latency, which may reflect the increase of EABR latency with more basal stimulation sites (e.g., Allum et al., 2000; Firszt et al., 2002a; Guiraud et al., in press), should then be moderated. It could indeed simply be due to a baso-apical gradient of neural degeneration resulting from the deafness (Abbas and Brown, 1991) rather than from an anatomical effect (i.e., a different way of transmitting the neural information as a function of pitch). Finally, multiple linear regression analyses were completed for the factors, which ANOVAs found to have significant effect on ELAR latency. These investigated whether ELAR latency can be predicted from a linear combination of those factors.