Introduction
Many cochlear implant (CI) listeners understand speech well, at least in quiet. However, there remains substantial across-listener variability, with some struggling even in favorable listening conditions (Holden et al.
2013). In recent years, there has been considerable interest in identifying the reasons for poor speech perception, and in identifying the relationship between performance and the global and local pattern of neural survival in each individual patient. A potentially useful approach comes from single-electrode psychophysical measures, which have revealed substantial across-listener and across-electrode variability in a number of tasks. These include signal detection in quiet and supra-threshold measures. Authors have investigated the variation in not only the absolute level of thresholds (Pfingst and Xu
2004; Bierer et al.
2015) but also how they are influenced by pulse rate (“multi-pulse integration, MPI”; Zhou and Pfingst
2014; Zhou et al.
2015; Zhou and Pfingst
2016), pulse polarity (Macherey et al.
2017), and stimulation mode (Bierer
2007; Bierer and Faulkner
2010). Supra-threshold tasks have included modulation detection (Garadat et al.
2012; Garadat et al.
2013), gap detection (Bierer et al.
2015), and rate discrimination both at low and high rates (Cosentino et al.
2016). It is worth noting that variation in all of these measures has been observed even when stimulating in monopolar mode, which is believed to produce a wide spread of excitation within the cochlea. Indeed, even though the variation in thresholds across the electrode array is smaller in dB for monopolar than for tripolar stimulation, this is not true when the across-electrode standard deviation (s.d.) is normalized by the within-electrode s.d. (Bierer et al.
2015). In other words, monopolar stimulation may reveal across-electrode variation as reliably as tripolar stimulation, even though the size of this variation is smaller in monopolar than in tripolar mode when expressed in dB.
A potentially important application of single-electrode measures is that they may guide the clinician in choosing which, if any, electrodes to deactivate when optimizing patient maps. Indeed, significant improvements in speech perception scores have been obtained by deactivating electrodes based on highmodulation detection thresholds (Garadat et al.
2012; Garadat et al.
2013) and high thresholds for low-rate pulse trains (Zhou
2017). In order to provide a principled approach to channel selection, it would be useful to know how the various different single-electrode measures correlate with each other. This could then either reveal clusters of tests, each of which taps a particular consequence of neural degeneration, or reveal a single test factor that could be used to guide channel selection algorithms. This information may also provide basic insights into the limitations of hearing by CI users. For example, Cosentino et al. (
2016) found that the “upper limit” of temporal pitch—defined as the highest pulse rate on a single electrode above which pitch no longer increased—correlated significantly with gap detection thresholds (GDTs), but not with the smallest difference in the rate of a low-rate pulse train that could be discriminated. The significant difference between these two correlations led them to suggest that there is a limitation specific to tasks that require sustained temporally accurate firing to high pulse rates, and which is separate from that which limits low-rate discrimination. Zhou and Pfingst (
2016) reported that MPI correlated significantly with the degree of spatial selectivity for a given electrode, and concluded that integration of multiple pulses is most efficient when conveyed by neurons that innervate a wide region of the cochlea.
The present study forms part of a series that compares performance on different single-electrode psychophysical measures in a group of CI users (Bierer et al.
2015; Cosentino et al.
2016). Here, we measure polarity sensitivity, defined as the difference between thresholds for 99-pps trains of triphasic pulses in which the short high-amplitude portion is either anodic or cathodic. As with other types of asymmetric pulse, triphasic stimulation allows one to study polarity sensitivity by concentrating charge of one polarity into a short time period, while maintaining the charge balancing necessary for patient safety. All stimulation is in monopolar mode. The motivation stems from the finding that, although animal studies usually reveal greater sensitivity to cathodic than to anodic stimulation (Hartmann et al.
1984; Miller et al.
1999; Miller et al.
2004), the reverse is true for human CI users when presented with stimuli at or close to their most comfortable listening level (“MCL”; Macherey et al.
2006; Macherey et al.
2008; van Wieringen et al.
2008; Undurraga et al.
2010; Macherey et al.
2011). A possible reason for this discrepancy, consistent with computational models (Rattay
1999; Rattay et al.
2001), is that cathodic stimulation depolarizes the peripheral processes of the auditory nerve. These processes are likely to be intact in the recently deafened animals used in most physiological experiments. However, there is evidence that peripheral processes are more susceptible than central axons to auditory deprivation, and so may have deteriorated in human CI users who have been deaf for months or years prior to implantation (Johnsson et al.
1981).
Stimulus polarity can affect not only MCLs but also thresholds. Unlike MCL measures, the direction of the polarity sensitivity at threshold varies consistently across listeners and electrodes, and some electrode-listener combinations reveal lower thresholds for cathodic than for anodic stimulation (Macherey et al.
2017; Mesnildrey et al.
2017). These combinations may reflect local regions of good neural survival in which a relatively high proportion of peripheral processes remain. Here, we investigate whether this putative measure of neural survival correlates with measures of gap detection, low-rate discrimination, and the upper limit of temporal pitch obtained in our previous studies (Bierer et al.
2015; Cosentino et al.
2016). The hypothesis is that lower thresholds for cathodic than for anodic stimulation will correlate with tasks that depend on good local neural survival. Note, however, that this does not require that the variation in performance on those tasks is limited by the pattern of activity in the auditory nerve; rather, poor auditory nerve survival may lead to more central degeneration which in turn could limit performance on perceptual tasks. For example, Carlyon and Deeks (
2015) found that the “alternating-amplitude” pattern of auditory nerve-evoked responses (Wilson
1997) to high-rate pulse trains correlated across subjects with poor rate discrimination at high rates, but also demonstrated that the correlation was not causal, and that manipulating the stimulus so as to reduce the alternating-amplitude pattern did not improve performance.
A second prediction is that polarity sensitivity will correlate with the average of the thresholds in the two polarities, on the assumption that better sensitivity to cathodic stimulation will reflect better neural survival and hence lower overall thresholds. Specifically, we assume that the average thresholds will also depend on the distance of the electrodes from the modiolus (electrode-modiolus distance, “EMD”) but that the effects of EMD and polarity are independent. In the “
Discussion” section, we describe a recent study that provides evidence for this assumption.
When comparing performance on different tasks, two types of measure are possible. One of these is to correlate performance across subjects (e.g., Fu
2002; Won et al.
2011; Cosentino et al.
2016). This can harness the often substantial across-listener variability in performance, but is potentially susceptible to non-specific effects such as attention span and cognitive ability. Such effects could lead to a correlation that does not reflect any common processing of the two tasks, except at very central levels. A more rigorous approach is to partial out between-subject effects, and to correlate the relative pattern of scores across electrodes (Bierer
2007; Cosentino et al.
2015; Zhou and Pfingst
2016). This approach is immune to between-listener cognitive differences. Across-electrode differences are also of more clinical relevance because, as mentioned above, they may guide channel selection methods that aim to optimize performance on a listener-by-listener basis. However, because this type of analysis excludes the substantial variation in neural survival that occurs across listeners, for example due to differences in pathology (e.g., Zimmermann et al.
1995; e.g., Nadol
1997), it risks “throwing the baby out with the bath water”. We perform both types of analysis here. One advantage of studying polarity sensitivity is that, being a difference between two thresholds, it is unlikely to be affected by between-listener variation in cognition. Hence, it may exploit the benefits of measuring the substantial across-listener variation in neural survival without being strongly influenced by cognitive effects.