Sound source identification
To establish a baseline of performance on the sound-source identification task, a group of typically-developing, 5-year-old children with normal acoustic hearing (NH group) was evaluated. Individual scatter plots of their localization accuracy and RMS errors are plotted in and , respectively. All children had RMS errors that were less than 30° (range: 8.9°–29.2°).
Figure 1 Individual performance on the sound source identification task in the NH group (A) and the BICI group (B). The size of the dots represents the number of responses for a given target location. Larger dots reflect a greater number of responses. The diagonal (more ...)
Figure 2 Individual performance on the sound source identification task (as quantified by RMS error) for the NH group (A) and the BICI group when using both implants (B). In both panels, data are plotted along the x-axis from small RMS errors (better performance) (more ...)
In contrast to the NH group, there was a larger range of RMS errors (19°–56°) among children who use BICIs in the bilateral condition under similar experimental settings ( and ). Visual inspection of the individual scatter plots () revealed large variability in sound source identification skills within the BICI group. Despite the variability, all but three children (CIAT, CIAG, CIAB) performed at least one standard error unit above chance levels on this task. To better quantify performance of the children in the BICI group, the following statistics are listed in : Percentage of responses in the correct hemifield (chance performance is 50%; two standard deviations above chance is 58% for the 15-AFC task and 62% for the 7-AFC task), RMS error for each hemifield, and correlation of target and responses within each hemifield.
Although there was a wide range of bilateral RMS errors among children in the BICI group, the individual scatter plots and additional analyses revealed three primary groups of children based on their performance. Group A included six children who performed similarly to the NH group according to the following criteria: First, the percentage of correct responses in each hemifield was within two standard deviations of the NH group average (Left: 92% ± 5.4%, Right: 89% ± 7.6%, mean ± SD, N=7). Second, correlations between target locations and responses were significant. Third, bilateral RMS errors ranged from 19.1° to 27.9° which fell within two standard deviations of the NH group average (18.3° ± 6.9°, N=7). Group B included 8 children who identified the correct hemifield of the target at better than chance performance, but varied in their ability to identify the target location within each hemifield which resulted in a lack of significant correlation between the targets and responses. Their bilateral RMS errors ranged from 32.8° to 42.5° which were larger by more than two standard deviations of the NH group average. Finally, Group C included 7 children who showed little ability to perform the sound source identification task. Responses were randomly distributed among the correct/incorrect hemifield in at least one hemifield for 3 children and in both hemifields for 4 children. In addition, few of their responses approximated the diagonal line as evidenced by both a lack of significant correlation between target locations and responses within each hemifield and bilateral RMS errors ranging from 43.5° to 66.8°.
To compare performance between the BICI and NH groups, an unequal N, between-subjects analysis was performed. CIAB and CIAC were removed from this analysis since they localized a different auditory stimulus than the NH group (see Methods, Stimuli). On average, the BICI group had significantly poorer localization accuracy (37.4° ± 11.0°, N=19) than the NH group [18.3° ± 6.9°, t(24)=4.2, p<0.001]. It is interesting to note, however, that six (6) of the 19 children in the BICI group had RMS errors that were within the range of those of the NH group (). While RMS error is a good tool for condensing performance down to a single metric, it is clearly not reflective of the various trends in the raw data. A careful inspection of individual subjects’ responses in suggests that RMS errors of similar values can be obtained for response profiles that are somewhat different. For example, the RMS of approximately 30° seen in several CI users resulted from different error types than that seen in the worse-performing NH subject (NH7, ). While the NH participant generally responded near the correct loudspeaker location, this child had a few large errors, which brought up the average error calculated. In contrast, the CI users with a similar or lower RMS (, top row) all had more scatter in their data. The source of this scatter is unclear; however, possibilities include spatial hearing abilities that are less well established, localization blur or uncertainty on the part of the participant.
In an attempt to identify predictors of localization accuracy in the BICI group on conditions with both CIs worn, a multivariate linear regression analysis was completed. Initially, six variables were used in the regression model: the children’s age at visit, age at first implant activation, age at second implant activation, history of acoustic hearing, duration of unilateral implant use, and duration of bilateral implant use were included in the regression model. Because of high intercorrelations among the three age variables, two of those variables were removed from the regression model. The analysis produced a significant result [F(4,14)=3.1, p=0.05] and revealed a significant effect of age at second implant activation [t(14)= −3.4, p<0.01; ). As noted above, however, age of second implant activation was highly intercorrelated with the children’s age at visit and age at first implant activation. This observation suggests that, although the age at second implant activation may be a predictor of localization accuracy, its effects cannot be separated from the possible effects of chronological age and/or age at first implant activation.
Coefficients of the multivariate regression analysis with bilateral RMS error as the dependent variable.
Relationship between right-left discrimination and sound source localization
Previous work from our research program has suggested that right-left discrimination abilities (another measure of spatial hearing that is quantified with the minimum audible angle (MAA)), typically emerge within 12 or more months following activation of the second implant in children who have sequential BICIs (Litovsky et al., 2006a
; Litovsky et al., 2006b
). Based on the observation that the majority of children were able to identify the correct hemifield of the target location significantly above chance (i.e., good spatial acuity), but continued to have difficulty identifying the specific location (i.e., poor localization accuracy), the next objective of the study was to determine a relationship, if any, between the two measures.
illustrates the relationship between spatial acuity and localization accuracy for the 19 children in the BICI group who localized a speech stimulus. There was a moderate correlation between the two measures [R2=0.68; F(1,16)=33.3, p<0.001], suggesting that the best performers on the right-left discrimination task were the best performers on the sound source identification task. Closer inspection of the data, however, revealed a wide range of RMS errors (19.1°–44.1°) for children who had relatively small MAAs (less than 20°). This finding suggests that when spatial acuity is poor (as reflected by a large MAA), localization accuracy is expected to be poor as well (as reflected by a large RMS error). However, when spatial acuity is good, children may exhibit wide-ranging localization accuracy.
Figure 3 Relationship between sound localization accuracy (as quantified by RMS error) and spatial acuity (as quantified by minimum audible angle). CNT: MAA could not be calculated because CIAG did not perform at above chance levels on the maximum angle separation (more ...)
Effect of Unilateral experience
One of the objectives of the study was to determine if there is a bilateral benefit on the sound source identification task. To evaluate this, children completed the sound source identification task with their first CI alone, and their performance was compared to that from the bilateral condition (; , circles). illustrates the individual RMS errors for the unilateral CI condition (squares) in which all but one child (CIAP) performed significantly above chance levels. A repeated-measures, within subjects t-test of the entire BICI cohort (N=21) showed that localization accuracy in the unilateral listening condition was significantly poorer than the localization accuracy in the bilateral listening condition [t(20)=−3.3, p=0.003]. Consistent with this finding, an unequal-N between subjects analysis revealed the RMS errors of the children in the BICI group who localized a speech stimulus in the unilateral condition (45.6° ± 7.1°, N=19) to also be significantly poorer than the RMS errors reported above for the NH group [8.3° ± 6.9°, N=7, t(24)=8.7, p<0.001].
Figure 4 Individual performance of the BICI group on the sound source identification task (as quantified by RMS error) when using a single CI (squares). Data from the bilateral condition (circles, from ) are plotted for comparison. Data are plotted along (more ...)
To better quantify a functional benefit of using bilateral implants for each child, bilateral benefit was defined as achieving RMS errors in the bilateral condition that were greater than the RMS errors for the unilateral CI condition by two standard error units (i.e., 7.3° for the 15-AFC task and 9.3° for the 7-AFC task; Hartmann et al., 1998
). Using these criteria, closer inspection of individual performance revealed that eleven of the children exhibited significantly better performance when using both CIs compared with the single-CI condition; ten children did not perform significantly different in the two listening conditions. A linear regression analysis was completed in attempt to identify possible predictors of unilateral performance. The children’s age at visit, age at first implant activation, age at second implant activation, history of acoustic hearing, duration of unilateral implant use, and duration of bilateral implant use were included in the regression model. None of these variables were found to be significant predictors of RMS errors in the unilateral condition [F(1,14)=1.04, p=0.42].
Emergence of spatial hearing abilities over time
To determine if sound localization abilities mature with increasing bilateral experience, 11 children were re-tested 7–21 months after the first testing. At each visit, children participated in the right-left discrimination task when using their BICIs. In addition, they were retested in the sound source identification task both in the bilateral and unilateral (first-CI) listening conditions. Experimental conditions were matched between the two visits so that changes in performance would be free from protocol changes (e.g., target stimulus or number of loudspeakers) and presumably reflect changes in each child’s ability to perform the tasks. illustrates RMS errors from individual children using their BICIs (circles) and their first CI alone (squares) as well as MAAs from these same children using BICIs (triangles). Although there was large individual variability in performance with BICIs over time, preliminary observations revealed three groups of children: 1) children who had large RMS errors (e.g., >50°) at both visits (top row), 2) children who showed a reduction of RMS errors (i.e., improvement) by 10° or more between visits 1 and 2 (middle row), and 3) children who had relatively small RMS errors (e.g., ≤ 30°) at both visits (bottom row).
Figure 5 Changes in sound localization abilities in children after 7–21 months of bilateral experience. RMS errors under the bilateral listening condition (circles) and the unilateral listening condition with the first CI (squares), as well as minimum (more ...)
Grouping children by both initial performance and change in performance over time led to a number of notable preliminary observations. For example, children who had large RMS errors with their BICIs on both visits (N=3; , top row) tended to have large RMS errors with their first CI alone on both visits as well. Two out of three of these children, however, showed an improvement in bilateral MAA. Children who had improvements of 10° or more (i.e., >2 standard error units) in localization accuracy with their BICIs on visit 2 had similar improvements in spatial acuity (N=4; , middle row). Although there was a concomitant reduction in the RMS error with the first CI alone for two of the four children, RMS errors continued to be larger in the unilateral condition. Finally, children who had RMS errors between 20°–30° on visit 1 (N=4; , bottom row) all had small reductions in RMS errors on visit 2. Three out of four of these children had a concomitant improvement in bilateral MAA. Changes in performance were not observed, however, for three out of four of the children when they used their first CI alone.