Sound Localization in Bilateral Users of MED-EL COMBI 40/40 Cochlear Implants

Transcription

1 Sound Localization in Bilateral Users of MED-EL COMBI 40/40 Cochlear Implants P. Nopp, P. Schleich, and P. D Haese Objective: The purpose of the study was to investigate sound localization with bilateral and unilateral cochlear implants. Design: Sound localization tests were performed on 20 bilaterally implanted MED-EL COMBI 40/40 users. All subjects were bilaterally implanted during adolescence or later. Sound localization was tested in the frontal horizontal plane by using 9 equally spaced loudspeakers and speech-shaped noise bursts at randomized levels. Results: The group of subjects who were bilaterally deafened after 5 to 6 yr of age (18 subjects) showed a statistically significant improvement in sound localization when using both implants, compared with when using only one. The mean deviation between the presentation azimuth and the response azimuth was 16.6 when using both implants, which was on average 37.1 smaller than when using one implant only. When adjusted for the localization error that was constant across loudspeakers, the mean deviation was 15.9 for bilateral implant use, representing an improvement of 30.1 over unilateral implant use. Statistical analysis showed that in this group, performance measures were not correlated with subject details such as age at onset of deafness or duration of unilateral implant use. In contrast, subjects who were bilaterally deafened before 6 yr of age (2 subjects) did not show a benefit in sound localization from bilateral implants. Conclusions: Bilateral cochlear implants offer a substantial benefit in sound localization to latedeafened, late-implanted subjects. The very limited data from early-deafened subjects implanted at a later age could suggest that these subjects may not benefit in sound localization from bilateral cochlear implants. It is possible that early implantation for early deafened subjects might allow better acquisition of spatial hearing, thus leading to improved localization performance. (Ear & Hearing 2004;25; ) Unilateral cochlear implantation has become an effective method for the treatment of severe to profound deafness (e.g., Helms et al., 1997, Helms et al., 2001). Interest in bilateral cochlear implantation, however, is just beginning despite the fact that Medical Electronics, Innsbruck, Austria. DOI: /01.AUD listening with two ears allows subjects with normal hearing to understand speech better in background noise or in reverberant environments (MacKeith & Coles, 1971). Recently, research has shown that these advantages also hold true for bilaterally implanted cochlear implant users (Müller, Schön, & Helms, 2002, Schön, Müller, & Helms, 2002). Furthermore, binaural hearing is an essential requirement for spatial hearing and sound localization (Durlach & Colburn, 1978). In general, most noise reduction and acoustical orientation abilities of the human auditory system depend on the subject having crucial access to time, level, and spectral differences between the sound signals sensed by the two ears. Sound localization with bilateral hearing aids has been investigated for decades, and it is well accepted today that binaural hearing aid fitting can restore sound localization (Dermody & Byrne, 1975), at least in moderately to severely impaired listeners (Byrne, Noble, & LePage, 1992). In contrast, sound localization with bilateral cochlear implants has not been investigated in great depth so far. To our knowledge, the first results on sound localization with bilateral cochlear implants were published by Schoen, Müller, & Helms, (1999) who found that subjects could achieve a minimum audible angle in the frontal horizontal plane of approximately 5 azimuth. Later, Schoen, Mueller, & Helms (Reference Note 3) and Müller, Schön, & Helms (Reference Note 2) performed localization tests in the frontal horizontal plane on 9 adult MED-EL COMBI 40/40 users with a stimulus of fixed presentation level. Gantz et al. (2002) investigated side identification in the frontal horizontal plane on 10 adult Nucleus CI-24M users. van Hoesel, Ramsden, & O Driscoll (2002) reported on localization tests in the frontal horizontal plane in one adult Nucleus CI-24M user. Nevertheless, although rather sparse, these data suggest that bilateral cochlear implant users can localize sound sources, at least within certain limits. This principal finding is supported by studies that found good access to interaural level differences and a range from good to poor access to interaural time differences in bilateral cochlear implant users (Lawson, Wilson, Zerbi, & Finley, 1996, Lawson, Zerbi, & Wilson, 1998). 0196/0202/04/ /0 Ear & Hearing Copyright 2004 by Lippincott Williams & Wilkins Printed in the U.S.A. 205

2 206 EAR &HEARING /JUNE 2004 TABLE 1. Overview of subjects included in the study Age DOD (yr) Implant use (yr) Implant type Hearing aid usage Etiology Subject (yr) L R L R L R L R L R AS C40 C40 no no M M AP C40 C CMV U AE C40 C40 no 0.00 M M BW C40 C O O BE C40 C SHL SAS BG C40 C SHL SHL DW C40 C P P DF C40 C M M FM C40 C P P IB C40 C MM MM IH C40 C no P P IV C40 C40 no 0.00 P P KF C40 C no TBF TBF KR C40 C no P MEI LR C40 C40 no no P SHL MM C40 C40 no 4.00 SHL SHL PS C40 C40 no 0.05 TBF TBF PR C40 C40 no no M M VM C40 C40 no no TBF TBF WQ C40 C SF SF All times are given with respect to the time of tests in the study. x hearing aid was used until x years before cochlear implantation; no no hearing was ever used; CMV cytomegalovirus; DOD duration of deafness; M meningitis; SAS silvian aquaduct syndrome; SF scarlet fever; MM morbus meniere; P progressive; SHL sudden hearing loss; TBF temporal bone fracture; MEI middle ear infection; O otosclerosis; U unknown; L left ear; R right ear. In this report, data from a study investigating sound localization in a larger group of adult subjects is presented. Sound localization with bilateral implant use and unilateral implant use is compared to determine to what extent in a larger patient population sound localization is improved with bilateral implants. In addition, performance measures are correlated with data from the subject case histories and implications for bilateral implant candidacy are discussed. METHODS Subjects Twenty adults (10 men, 10 women) with an average age of 45.1 yr (range, 17 to 67 yr) were included in the study (Table 1). All subjects but one (AP) were postlingually deafened. The duration of deafness across all ears ranges from 3 to 48 yr (mean, 13.6 yr). The subjects used hearing aids before cochlear implantation to various degrees. All subjects used MED-EL COMBI 40 or COMBI 40 implants and had at least 1 mo of experience with their most recently implanted cochlear implant system. All of them used their standard TEMPO speech processor (Helms et al., 2001), using the CIS strategy (MED-EL, Reference Note 1). All subjects use their bilateral implants routinely. The subjects signed an informed consent form and were reimbursed for travel expenses. Implant Fitting and Speech Processor Settings All subjects were tested with their normal everyday processor settings. These settings have been obtained during clinical device fitting in which implants are normally first fitted on each side individually, and the patient subsequently adjusts the loudness on each side to a comfortable level so that no difference in loudness exists between sides. The only aspects in which processor settings were manipulated before the tests concern volume and the automatic gain control (AGC). To compensate for a possible effect of loudness summation, before each test, the subject was asked to adjust the volume to a comfortable level. To have equal microphone sensitivity across all subjects with respect to the sound pressure levels used in the test, the AGC was set to maximum sensitivity in each subject during all tests. With this, the onset of compression is at approximately 45 db SPL, and the compression ratio for levels exceeding this level is 1:3 (Stöbich, Zierhofer, & Hochmair, 1999). Setup and Procedure All tests were performed in an anechoic chamber (4 m 4m 12 m). The test setup consisted of 9 loudspeakers arranged at equal angles in a semicircle of 1 m radius between 90 azimuth (left) and 90 azimuth (right). Thus, the spacing between

3 EAR &HEARING, VOL. 25 NO loudspeakers was The loudspeakers had frequency characteristics within 1.0 db between 0.5 and 2 khz and were equalized in loudness. Bursts of speech-shaped noise (CCITT noise, as standardized in CCITT recommendation 227, Fastl, 1993) of 1 sec duration and 100 msec rise/fall time were used as stimuli. Presentation levels were randomly chosen to be 60 db, 70 db, or 80 db SPL, that is, all levels were above the AGC compression onset. The order in which the noise bursts were presented was randomized across loudspeakers and levels. Each level/loudspeaker combination was presented 5 times so that a test consisted of 135 presentations. Tests were performed in each subject by using both cochlear implants, the left cochlear implant only, and the right cochlear implant only. In each listening condition, one test was performed. The sequence of listening conditions was quasirandomized. To not bias any unilateral condition, care was taken that in the group both unilateral conditions followed the bilateral condition equally often. Tests were carried out in the form of a sourceidentification task (Hartmann, Rakerd, & Gaalaas, 1998). The loudspeakers were consecutively numbered from 90 to 90 azimuth, and the subject had to identify by number the loudspeaker he thought the stimulus was presented from. No feedback as to correct or incorrect responses was given. To prevent any head movements, the subject had to lean his head against a 2-point headrest. In addition, head movements were monitored by the experimenter by means of a video surveillance system. Data Analysis The accuracy with which a subject was able to locate the sound sources is assessed by calculating the mean deviation d between the judged azimuth and the azimuth from which a sound was presented. d is the mean value of the unsigned difference d k between the average k of the judged azimuth values in response to a stimulus from loudspeaker k, and the azimuth k for loudspeaker k d N 1 N k 1 d k N 1 N k k (1) k 1 with N the number of loudspeakers. The smaller d is, the more accurate a subject s judgments were on an average. The smallest possible larger-than-zero error in d k is to be off by one loudspeaker in response to one of the 15 presentations from a certain loudspeaker, and thus d k 22.5 / With all other responses in a test being correct, d would result to d d k / Thus, the resolution of the test is approximately The deviation d can be separated into the localization bias b and the bias-adjusted deviation d b. The bias b is the localization error, which is constant across loudspeakers and is calculated from the mean value of a subject s responses in a test b 1 N N M k 1 M i,k (2) i 1 with M the number of stimuli per loudspeaker. The deviation d b is calculated by bias-adjusting the responses (i.e., bias-adjusting mean responses k ) before calculating the deviation from k d b N 1 N k b k (3) k 1 A subject s responses are reflected in the above measures. For example, if a subject shows perfect localization (i.e., ), then d d b b 0. Ifall responses of a subject deviate from the stimulus by a constant shift of X (i.e., X ), then d X, d b 0, and b X (neglecting edge effects). If a subject always indicates the same loudspeaker irrespective of the stimulus (i.e., X for all ), then d depends on the indicated loudspeaker, d b 50, and b X. Thus, d b is representative for the directional relation between a subject s responses and the stimuli in a test, whereas d measures the total localization error experienced by a subject. The variability s of the responses was calculated from the RMS average of the standard deviation values s k of the responses for each loudspeaker k 1 s N N N M 1 s 2 k N M 1 i,k k 2 (4) k 1 k 1 i 1 with i,k the judged azimuth in response to the i-th presentation from loudspeaker k. The smaller s is, the more consistent the subject s judgments were, that is, the less scatter there was in the subject s responses. Note that all measures can be calculated over all loudspeakers (N 9) or over a subset of the loudspeakers only (N 9). Statistics The main factor of interest is whether or not sound source localization is better with bilateral implant use than with unilateral implant use. A 1-way repeated-measures analysis of variance was thus used to evaluate possible differences in d, d b, and b among listening conditions. For s, a 1-way, repeated-measures analysis of variance on ranks (Friedman) was used because s was not Gaussian distributed. In all cases, the Tukey test was used for

4 208 EAR &HEARING /JUNE 2004 post hoc comparisons between the three listening conditions. RESULTS Results for three subjects are shown in Figure 1. For each subject (row in Fig. 1) and listening condition (column), mean values (diamonds) and standard deviation values (error bars) for the response azimuth are shown as a function of the presentation azimuth. These subjects represent three major response patterns that were observed among the 20 subjects. These mainly differ in how the subject copes with the task in the unilateral listening conditions. With both cochlear implants, there is a clear relation between the response azimuth and the presentation azimuth. On an average, mean values for align fairly well with the diagonal which marks perfect agreement between and. For the unilateral listening conditions, type 1 (subject VM, top row) resembled a pattern of guessing where responses are scattered over all presentation azimuths, so that the mean values for are close to 0 and standard deviation values are large. Type 2 (subject WQ, middle row) showed a pattern of guessing where responses were scattered over loud- Fig. 1. Localization results for (a) subject VM (type 1), (b) subject WQ (type 2), and (c) subject DF (type 3). Left column shows results for left cochlear implant only; middle column, results for both cochlear implants; right column, results for right cochlear implant only.

5 EAR &HEARING, VOL. 25 NO speaker positions ipsilateral to the implant only. This leads to a marked bias in the mean values toward the side where the implant is used and standard deviation values of, on average, about half the magnitude of type 1. Type 3 (subject DF, bottom row) more or less showed a relation between and even for unilateral implant use, although this relation was less consistent and standard deviation values where higher than for bilateral implant use. It should be stressed that among these types, all forms of hybrids exist where, for example, a subject shows one behavior on the one side and another behavior on the other side or shows a hybrid of types within one ear. Results for the mean deviation d, the bias-adjusted mean deviation d b, the bias b, and the variability s for all subjects are shown in Table 2 and Figure 2a. Because of the large amount of data, no subject-specific symbols are provided in Figure 2a, but it is possible to identify a subject in Figure 2a by using the respective data in Table 2. Open markers symbolize expected values for unilateral implant use from type 1 (calculated from a guessing pattern with responses uniformly scattered over the frontal horizontal plane) and type 2 (calculated from a guessing pattern with responses uniformly scattered over the frontal horizontal half-plane ipsilateral to the implant). Note that expected values for d b are identical for type 1 and type 2. For all subjects but two (subjects AP and KR), both d and d b decrease when moving from any unilateral condition to the bilateral condition. When calculated over all loudspeakers, the smallest improvement found for any subject in this subgroup was 15 for d and 10 for d b. Subject AP showed even higher values for d and d b for the bilateral condition than for the unilateral conditions, and subject KR showed very similar values for the bilateral and the right cochlear implant only condition. These two subjects are indicated by dashed lines in Figure 2a. In general, s does not show such a uniform behavior. For the majority of subjects, s decreases from the unilateral conditions to the bilateral condition; however, for some subjects (IH, LR, MM, WQ) s increases from one or both unilateral conditions to the bilateral condition. All subjects show little or no bias b in the bilateral condition. In the unilateral conditions, most subjects have a large bias toward the side where the implant is used. However, some subjects show a bias in the unilateral conditions, which is as small as in the bilateral condition. Group data (mean and standard deviation values for d, d b, and b, RMS values for s) are shown in Figure 2b and in Table 3 (first row, other rows follow in the Discussion section). On an average, all measures are smallest for bilateral implant use. The statistical analysis showed that the differences between the bilateral listening condition and the unilateral listening conditions are statistically significant for all measures (Table 4). Significant differences between the unilateral conditions exist for b only. Taken together, Table 3 and Table 4 indicate that when comparing bilateral implant use with unilateral implant use (averaged across the two ears), on average d TABLE 2. Localization test results for all subjects d d b b s Subject Left Both Right Left Both Right Left Both Right Left Both Right AS AP AE BW BE BG DW DF FM IB IH IV KF KR LR MM PS PR VM WQ See Methods section for details on d, d b, b, and s. Left results for left CI only; both results for both CIs; right results for right CI only; all entries given in degrees.

6 210 EAR &HEARING /JUNE 2004 Fig. 2. (a) Deviation d, bias-corrected deviation d b, bias b, and variability s for all subjects. (b) Mean values (bars) and standard deviation values (error bars) for d, d b, and b, root mean square values (bars) for s. Left indicates results for left cochlear implant only; both, results for both cochlear implants; right, results for right cochlear implant only;, expected values for type 1;, expected values for type 2; dashed lines, data from early-deafened subjects AP and KR. All measures given in degrees. decreases significantly by 33.2, d b decreases significantly by 26.9, b changes significantly by 26.5, and s decreases significantly by DISCUSSION The results show a substantial and significant benefit of bilateral implants. Indeed, with two exceptions, all subjects had a substantial improvement in their sound localization ability when using both implants instead of one. To summarize, the results not only show that with bilateral cochlear implants, the accuracy with which subjects can localize sounds improves by more than 30 on average (i.e., d and d b both decrease by more than 30, Fig. 2), but the fact that the variability s of the responses decreases by more than 16 also indicates that subjects are more consistent in their judgments. In addition, percepts in the frontal horizontal plane are well centered around 0 azimuth as indicated by the fact that subjects experience small localization biases b only. For bilateral implant use, d and d b are practically TABLE 3. Mean values (standard deviation values) for d, d b, and b, RMS values for s d d b b s Left Both Right Left Both Right Left Both Right Left Both Right All subjects All loudspeakers (14.8) (9.6) (16.6) (9.6) (9.6) (9.6) (27.0) (5.7) (28.6) All subjects Central 7 loudspeakers (15.0) (8.5) (18.0) (7.9) (8.2) (9.0) (25.7) (6.5) (28.3) Subjects with OOD 6 yrs All loudspeakers (15.2) (4.4) (16.6) (10.0) (4.4) (9.9) (26.6) (5.9) (28.2) Subjects with OOD 6 yrs Central 7 loudspeakers (15.4) (4.9) (17.8) (8.2) (4.1) (8.9) (24.8) (6.8) (27.6) See Methods section for details on d, d b, b, and s. LS loudspeakers; OOD onset of deafness; left results for left CI only; both results for both CIs; right results for right CI only; all entries given in degrees.

7 EAR &HEARING, VOL. 25 NO TABLE 4. Post hoc analysis results of data for all subjects and loudspeakers Both vs. left Both vs. right Left vs. right d 0.001* 0.001* d b 0.001* 0.001* b 0.001* 0.012* 0.005* s 0.05* 0.05* 0.05 Left results for left CI only; both results for both CIs; right results for right CI only. * Result indicating significance (p 0.05, 2-tailed). identical, which again reflects the very small bias in this condition. When discussing results for unilateral implant use, it should first be noted that all subjects in the study use their bilateral implants routinely. With one implant, some subjects show results scattered around the hypothetical (guessing) values for type 1 and type 2 (Fig. 2), suggesting that the subjects responses are rather based on guessing than on a clear idea about the sound s origin. Few subjects show type 3 behavior (low d, d b, and b), that is, at least some ability to localize sound sources even in the unilateral conditions. Some subjects show a behavior that is a hybrid of type 1 or 2 and 3 (intermediate d, d b, and b). It should be stressed, however, that, for example, a d b of 50 can be achieved not only with scattering responses uniformly over the frontal horizontal plane (as typical for type 1) but also with other response patterns. Thus, the fact that a subject shows performance measures that are (approximately) equal to the hypothetical values for type 1 or type 2 does not absolutely reliably confirm that guessing (as characteristic for type 1 or 2) is actually that decisionmaking strategy that is used by the subject in the localization test. However, we might at least state that for unilateral implant use subjects with d b 50 do not perform better than guessing. Subjects with d b 37.4 perform significantly better then chance (p 0.01). From this and the data in Table 2, it follows that when using one implant only, four subjects perform significantly better than chance in one unilateral condition, and two subjects (DF, KF) perform significantly better than chance in both unilateral conditions. When using both implants, all subjects but one (AP) perform significantly better than chance. The significant difference in b between the unilateral listening conditions (Table 4) reflects the fact that subjects tend to lean toward different sides when using the left or right implant only. It should be noted, however, that the magnitude of b is almost identical for the unilateral conditions (Fig. 2 and Table 3). For unilateral implants, the difference between d and d b is highly dependent on the response pattern. For type 1 performers, d d b 50. Type 2 subjects show d values around the expected value of Asd b is independent of b, d b 50, as for type 1 subjects. For type 3 performers, d d b as b is small. The fact that d b is only slightly dependent on the response type is also reflected in smaller standard deviation values for d b than for d (Fig. 2 and Table 3). The data found here agree well with the few data published or presented elsewhere. In tests on 9 bilaterally implanted adults (COMBI 40/40, 7 loudspeakers between 90 and 90 azimuth, CCITT bursts at 80 db SPL), Müller et al. (Reference Note 2) found that 8 could localize sounds when applying a significance criterion based on the correlation between presentation azimuth and response azimuth. In these 8 subjects, for bilateral implant use, an average RMS deviation (calculated from the sum of the squared differences between the average responses and the loudspeaker azimuths) of 20.3 and an s of 25.3 was found, which corresponds well with our data that show an RMS error of 24.5 and an average s of 28.9 (Table 3). In tests on one bilaterally implanted subject (Nucleus CI-24M, 11 loudspeakers between 90 and 90 azimuth, quadruples of pink-noise bursts at 70 db SPL 3dB jitter), van Hoesel et al. (2002) found a d of 15.7 and an s of 17.6 for bilateral implant use. Use of the left implant only resulted in d 81.8 and s 9.6, and for the right implant only d 72.5 and s 9.6 was found. These results are well within the range found in the study here. In addition, the data show that the subject falls into the type 2 category. Gantz et al. (2002) used a setup with loudspeakers positioned at 45 azimuth and 45 azimuth to investigate right/ left localization. They found that 10 of 10 subjects (Nucleus CI-24M) could identify the correct loudspeaker with scores exceeding 95% correct. No further analysis of the data was provided. It should be stressed that using the term sound localization for unilateral implant use is somewhat ambiguous. When using one cochlear implant, the listener lacks those binaural cues that are thought to be most essential for sound localization in the horizontal plane, namely interaural time and level differences. Thus, responses for unilateral implant use vary, largely depending on how the subject copes with the task, and this most probably mainly determines in what category the subject falls (type 1, type 2, or type 3). A subject might always identify the most lateral loudspeaker ipsilateral to the implant as the sound source, since the sound is always heard in that one ear only (this could be called some form of guessing in the ipsilateral half-plane and might thus be identified as type 2); or a subject finds that always pointing to the same loudspeaker is not

8 212 EAR &HEARING /JUNE 2004 appropriate and makes judgments based on guessing (type 1 or 2); or a subject makes judgments based on secondary cues such as a change in sound timbre and/or perceived loudness (although levels were randomized in this test) if the sound source changes side (type 3). In none of these cases is the subject locating the sound in the sense of perceptually attributing to the sound source a location in the 3-dimensional space. With bilateral implant use, subjects showed a marked central bias, that is, a shift toward the midline, in response to presentations lateral to 45 and 45, respectively, leading to a d k (Equation 1) of up to 45 for the most lateral loudspeakers (Fig. 1). In tests on normal-hearing subjects investigating sound localization in the (complete) horizontal plane, Blauert (1997) found an average inward shift of approximately 10 (standard deviation: 10 ). This indicates that the shift found here is at least in part a phenomenon that has also been found in normal listeners. There might be additional factors contributing to the central bias. One is the edge effect, that is, the fact that the two most lateral loudspeakers have one neighbor only. Another factor might lie in the map law used in the CIS strategy. A toocompressive map law could lead to a reduction in the interaural difference between stimulation current levels produced by the head shadow effect for high presentation levels and to an enhancement for low presentation levels. Similarly, a too-expansive (shallow) map law could lead to a reduction in the interaural difference between stimulation levels for low presentation levels and to an enhancement for high presentation levels. However, that part of the central bias that is an artifact caused by the map law could thus be corrected by adapting the map law accordingly. When calculating performance measures over the central seven loudspeakers (i.e., over azimuths from 67.5 to 67.5 ) only, d decreases to (Table 3), and d b decreases to , for bilateral implant use. s and b practically do not change. As mentioned above, subjects AP and KR did not show a bilateral benefit. In addition, these subjects showed the highest deviation d for the bilateral condition of all subjects (Table 2). Both subjects were deafened bilaterally in early childhood (Table 1), so that binaural hearing abilities could probably not (fully) mature. If these subjects are excluded from the data analysis, then d decreases to and d b decreases to when calculated over all loudspeakers (Table 3). The average improvement in d is 37.1 and in d b it is When calculated over the central seven loudspeakers only, d decreases to and d b decreases to Again, s and b practically do not change. However, it should be mentioned here that another subject (AE) has a history of auditory experience similar to that of subject KR but performed as well as the rest of the group. The fact that two of the three subjects who were deafened in early childhood did not show a bilateral benefit could indicate a critical period for the acquisition of binaural hearing. If so, then the fact that AE and KR behave differently, although they have a similar history of auditory experience, would indicate that this critical period is not equally long across subjects, that is, subjects acquire binaural hearing at different rates of learning. The literature indicates that spatial hearing improves with age (Ashmead, Davis, Whalen, & Odom, 1991, Litovsky, 1997). Indeed, for the bilateral condition, a negative correlation was found between d b and the age at the onset of deafness (OOD) (first deafened ear: r 0.615, p 0.005; second deafened ear: r 0.628, p 0.004; Fig. 3, dashed line). For d, the correlation with OOD was also significant (first deafened ear: r 0.582, p 0.007; second deafened ear: r 0.589, p 0.006). The outlier at an age of 35 yr is produced by subject BG, who showed a very strong central bias (i.e., shift toward the midline) leading to high d b and d. We assume that this was at least in part produced by an overcompressive or overexpansive map law so that the outlier can be considered an anomaly. However, applying the linear model underlying the Pearson correlation coefficient inherently assumes an equal importance of all life years for the acquisition of binaural hearing. To account for the fact that for the maturation of the binaural system, the early life years probably play a more crucial role than the years later in life, a linear correlation was calculated between the logarithm of d b and the logarithm of OOD (for subject AP, the OOD at birth was substituted by an OOD at 1 mo of age to allow calculation of the logarithm). An even stronger correlation was found here (first deafened ear r 0,815, p 0.001; second deafened ear r 0,814, p ). For d, a similar correlation was found (first deafened ear r 0.790, p 0.001; second deafened ear r 0.784, p 0.001). The resulting curves for d b (first deafened ear: d 32.3 Age ; second deafened ear: d 39.6 Age ) and d (first deafened ear: d 32.6 Age ; second deafened ear: d 39.3 Age ; Fig. 3, continuous line) indicate a large improvement in a subject s localization accuracy within the first 5 to 6 yr of age. In line with these correlations, a positive correlation was found between the duration of deafness expressed as a fraction of age, and d b (first deafened ear: r 0.696, p 0.001; second deafened ear: r 0.705, p 0.001) and d (first deafened ear: r 0.684, p 0.001; second

9 EAR &HEARING, VOL. 25 NO Fig. 3. Bias-adjusted deviation d b for bilateral implant use as a function of the age at onset of deafness (OOD) for (a) the first deafened ear and (b) the second deafened ear. Dashed line indicates linear best-fit curve; continuous line, powerfunction best-fit curve. deafened ear: r 0.690, p 0.001), respectively. This again confirms that the longer in life one was deaf, in relation to his age, the poorer his localization ability. It should be mentioned that in contrast to the above no correlation was found between the localization accuracy measures (d, d b ) and the duration of deafness (DOD). This is because relating performance measures to DOD leaves out the aspect that it makes a difference if one who is (bilaterally) deafened for, for example, 18 yr, is 20 yr old or 30 yr old. The former subject had binaural hearing for 2 yr only, whereas the latter had binaural hearing for 12 yr. This aspect is taken into account by expressing DOD as a fraction of age (see above). In the subgroup of subjects who showed a bilateral benefit (i.e., the whole group minus AP and KR), no significant correlation was found between the localization accuracy measures (d, d b ) and age at OOD nor between these measures and duration of deafness as a fraction of age. This indicates that once binaural hearing is acquired, the ability to localize sounds is very robust and does not vanish or even deteriorate significantly over the time of deafness which ranges from 3 to 48 yr in this subgroup. Thus the significant correlation between localization accuracy and duration of deafness as a fraction of age for the whole group (see above) is obviously produced by subjects AP and KR. In addition, no correlation could be found between the localization accuracy measures and the duration of unilateral implant use (range, 0 to 85 mo in this subgroup), suggesting that a longer period of unilateral implantation neither improves nor impairs a subject s localization accuracy. Thus, a long duration of unilateral implant use seems to be no contraindication to bilateral cochlear implantation. Considering how heterogeneous this group is regarding hearing aid usage before cochlear implantation (Table 1), it seems that hearing aid usage is not an important factor either. In this context, it should be mentioned that some subjects noted that they wore their hearing aids although they did not receive (at least consciously) any useful input from them. This means that an eventual time gap between abandoning the usage of hearing aids and cochlear implantation might not be the only time a subject was without any auditory input or that even with no time gap between hearing aid usage and cochlear implantation, a subject might have been without any auditory input on a certain ear. In the context of the correlations discussed above, it should be stressed that all subjects received their first implant after the age of 11 yr and their second implant after the age of 14 yr. Thus we may conclude that as far as sound localization is concerned, subjects who are bilaterally implanted in adolescence or later are good candidates for bilateral implants regardless of their details (age at OOD, duration of deafness as a fraction of age, duration of unilateral implant use), provided their OOD is after 5 to 6 yr of age. Subjects who became deaf before 5 to 6 yr of age and are implanted in adolescence or later may have little or no benefit from bilateral implantation in sound localization. The situation is probably very different in a congenitally deaf or an early deafened subject who is bilaterally implanted in early childhood. Here, the early access to binaural hearing through bilateral cochlear implants might allow binaural hearing abilities to mature. Indeed, first experiences with bilaterally implanted children point to this direction (Müller et al., Reference Note 2). Finally, from the fact that early-deafened, late-implanted subjects may not benefit from bilateral implants in sound

10 214 EAR &HEARING /JUNE 2004 localization, we do not conclude that this situation is a contraindication to bilateral implantation per se, as these subjects might still benefit from their bilateral implants in speech perception. CONCLUSION Bilateral cochlear implantation offers a substantial benefit in sound localization to late-deafened subjects. Since the literature indicates that these subjects also benefit from bilateral implants in speech perception (Müller et al., 2002, Schön et al., 2002), they seem to be good candidates for bilateral cochlear implants in general, without restrictions imposed by subject details such as age at onset of deafness or duration of unilateral implant use. In contrast, the very limited data from early-deafened, late-implanted subjects could suggest that these subjects may not benefit in sound localization from bilateral cochlear implants. It is possible that early implantation for early-deafened subjects might allow better acquisition of spatial hearing, thus leading to improved localization performance. Address for correspondence: Dr. Peter Nopp, Medical Electronics, Fürstenweg 77a, A-6020 Innsbruck, Austria. p.nopp@medel.com Received November 13, 2002; accepted December 12, 2003 REFERENCES Ashmead, D. H., Davis, D. L., Whalen, T., & Odom, R. D. (1991). Sound localization and sensitivity to interaural time differences in human infants. Child Development, 62, Blauert, J. (1997). Spatial hearing. London: The MIT Press. Byrne, D., Noble, W., & LePage, B. (1992). Effects of long-term bilateral and unilateral fitting of different hearing aid types on the ability to locate sounds. Journal of the American Academy of Audiology, 3, Fastl, H. (1993). A masking noise for speech intelligibility tests. In: Proc. TC Hearing, Acoustical Society of Japan, H Dermody, P., & Byrne, D. (1975). Auditory localization by hearing-impaired persons using binaural in-the-ear hearing aids. British Journal of Audiology, 9, Durlach, N. I., & Colburn, H. S. (1978). Binaural phenomena. In: Carterette, E. C., & Friedman, M. P. (Ed.), Handbook of Perception, Volume IV (pp ). New York: Academic Press. Gantz, B. J., Tyler, R. S., Rubinstein, J. T., Wolaver, A., Lowder, M., Abbas, P., et al. (2002). Binaural cochlear implants placed during the same operation. Otology & Neurotology, 23, Hartmann, W. M., Rakerd, B., & Gaalaas, J. B. (1998). On the source-identification method. Journal of the Acoustical Society of America, 104, Helms, J., Müller, J., Schön, F., Moser, L., Arnold, W., Janssen, T., Ramsden, R., von Ilberg, C., Kiefer, J., Pfennigdorf, T., et al. (1997). Evaluation of performance with the COMBI 40 cochlear implant in adults: a multicentric clinical study. ORL; Journal of Oto-Rhino-Laryngology and its Related Specialties, 59, Helms, J., Müller, J., Schön, F., Winkler, F., Moser, L., Shehata- Dieler, W., Kastenbauer, E., Baumann, U., Rasp, G., Schorn, K., et al. (2001). Comparison of the TEMPO ear-level speech processor and the CIS PRO body-worn processor in adult MED-EL cochlear implant users. ORL; Journal of Oto-Rhino- Laryngology and its Related Specialties, 63, Lawson, D., Wilson, B., Zerbi, M., & Finley, C. (1996). Fifth Quarterly Progress Report. NIH Contract N01-DC Retrieved from SpeechProcessorsforAuditoryProstheses%20DC52103/qpr5/ qpr5.html Lawson, D., Zerbi, M., & Wilson, B. (1998). First Quarterly Progress Report. NIH Contract N01-DC Retrieved from npp.ninds.nih.gov/progressreports/speechprocessorsforauditory Prostheses%20DC82105/qpr1/qpr1a.html Litovsky, R. (1997). Developmental changes in the precedence effect: estimates of minimum audible angle. Journal of the Acoustical Society of America, 102, MacKeith, N. W., & Coles, R. R. A. (1971). Binaural advantages in hearing of speech. Journal of Laryngology and Otology, 85, Müller, J., Schön, F., & Helms, J. (2002). Speech understanding in quiet and noise in bilateral users of the MED-EL COMBI 40/40 cochlear implant system. Ear and Hearing, 23, Schoen, F., Müller, J., & Helms, J. (1999). Results of bilateral cochlear implantation. European Archives of Oto-Rhino-Laryngology, 156, 106. Schön, F., Müller, J., & Helms, J. (2002). Speech reception thresholds obtained in a symmetrical four-loudspeaker arrangement from bilateral users of MED-EL cochlear implants. Otology & Neurotology, 23, Stöbich, B., Zierhofer, C., & Hochmair, E. (1999). Influence of automatic gain control parameter settings on speech understanding of cochlear implant users employing the continuous interleaved sampling strategy. Ear and Hearing, 20, van Hoesel, R., Ramsden, R., & O Driscoll, M. (2002). Sounddirection identification, interaural time delay discrimination, and speech intelligibility advantages in noise for a bilateral cochlear implant user. Ear and Hearing, 23, REFERENCE NOTES 1 MED-EL (1999). Combi 40/40 system software manual. Innsbruck, Austria. 2 Müller, J., Schön, F., & Helms, J. (2001). Restoration of binaural hearing by means of bilateral cochlear implantation. Abstracts of the 2001 Conference on Implantable Auditory Prostheses. 3 Schoen, F., Mueller, J., & Helms, J. (2000). Directional hearing in bilaterally implanted cochlear implant patients. Abstracts of the 6 th International Cochlear Implant Conference.