Beyond cochlear implants: awakening the deafened brain David R Moore1,2 & Robert V Shannon3 Cochlear implants have provided hearing to more than 120,000 deaf people. Recent surgical developments include direct electrical stimulation of the brain, bilateral implants and implantation in children less than 1 year old. However, research is beginning to refocus on the role of the brain in providing benefits to implant users. The auditory system is able to use the highly impoverished input provided by implants to interpret speech, but this only works well in those who have developed language before their deafness or in those who receive their implant at a very young age. We discuss recent evidence suggesting that developing the ability of the brain to learn how to use an implant may be as important as further improvements of the implant technology. Cochlear implants deliver the ability to recognize speech to the profoundly deaf and are arguably the most effective neural prostheses ever developed. Using direct electrical stimulation of the auditory nerve (Fig. 1), these devices provide the brain���s central auditory system with peripheral input that is highly unnatural and impoverished relative to the normally functioning cochlea. However, adults who have had some hearing ability before their deafness can understand speech well enough with a cochlear implant that most can converse easily by telephone, using only the sound from the implant and with no lip-reading cues. These clinical results indicate that the brain can function adequately with distorted and impoverished input once it has learned complex pattern recognition during normal language acquisition. However, if the pattern-recognition system has never been established or if the peripheral input is very badly distorted, then the central auditory system must learn to interpret an entirely new array of peripheral inputs. Although conventional cochlear implants continue to successfully advance1,2, new developments in the delivery of electrical stimulation to the brain raise fundamental questions about how the brain makes sense of both artificial and natural input. We review cochlear and brain implants in the context of central auditory system dynamics and behavioral training that promises to accelerate and enhance (re)habilitation after implantation. As we approach the physical limits of artificial systems to deliver useful spectro- temporal information to the nervous system, we ask whether the incredible capacity of the brain to utilize the input at its disposal can be pushed to improve further benefits for deaf people. ���Brain training��� has been injected as a concept into the public consciousness via Nintendo���s popular and heavily publicized DS console application. Less well known, but possibly of greater impor- tance in the longer term, is the growing number of perceptual learning applications that are emerging from neuroscience research labs3. In sensory rehabilitation, computer-based training for amblyopia treat- ment is now undergoing clinical trials4, and such training has also been adopted as an adjunct treatment for hearing aid users5. Training also benefits cochlear implant users6 and this application is also being translated into clinical practice. Cochlear and central auditory implants Cochlear implants began in the 1950s when Djourno and Eryies, a French surgeon and engineer, collaborated to place a coil of wire in the inner ear of two deaf people7. Although the device failed after a short time, it did produce useful auditory sensations and led to further experiments on electrical stimulation of hearing. The 1960s saw an explosion of basic research on the physics and physiology of the cochlea. Scientists were focused on the mechano-electrical complexities of cochlear function and were therefore highly skeptical that wires inserted into the cochlea could replace the natural complexities they were observing. However, in spite of the crude stimulation of the early cochlear implants, patients showed great acceptance and benefits. Modern cochlear implants allow the typical patient to understand more than 90% of the words in unfamiliar sentences when presented in quiet listening conditions (Fig. 2)1. Speech recognition with a cochlear implant is often not this good when the device is initially activated but improves rapidly over the first 3���6 months of daily use2. This implies that the patterns of excitation produced by the implant are different enough from those of normal auditory stimulation that a period of adjustment is required for the new sensory pattern to be functionally linked with an existing, developmentally specified template that was also refined by previous experience. This process has been described in the cochlear implant literature as ���accommodation���, ���acclimatization��� and ���adaptation���, but the label ���learning��� seems to be more appropriate, as the process appears to depend on an active engagement between the cochlear implant user and their environment. Some patients are unable to benefit from a cochlear implant because their auditory nerve is dysfunctional. For these patients, who mostly suffer from neurofibromatosis type 2 (NF2), a Schwann cell disorder, a potential solution appeared to be direct electrical stimulation of the brain. Encouraged by the success of cochlear implants, an auditory brainstem implant (ABI Fig. 1) was developed8,9. This device is similar PERSPECTIVE HEARING Published online 26 May 2009 doi:10.1038/nn.2326 1Medical Research Council Institute of Hearing Research, Nottingham, UK. 2National Biomedical Research Unit in Hearing, Nottingham, UK. 3House Ear Institute, Los Angeles, California, USA. Correspondence should be addressed to D.R.M. (david.moore@ihr.mrc.ac.uk). 686 VOLUME 12 [ NUMBER 6 [ JUNE 2009 NATURE NEUROSCIENCE �� 2009 Nature America, Inc. All rights reserved.

in principle to a cochlear implant but stimulates the cochlear nucleus, the first auditory relay nucleus in the brainstem, via surface mounted ���button��� electrodes. Unlike the auditory nerve, the cochlear nucleus contains many different types of neurons that perform specialized processing and feature extraction on the auditory signal, such as onset and offset times and modulation. The cochlear nucleus also contains more than one tonotopic dimension, so electrodes placed on the surface of the cochlear nucleus can activate more than one pitch. Given these anatomical differences, can the same coarse electrical activation that works in a cochlear implant also work in the cochlear nucleus to convey speech? Psychophysical measures have shown that ABI patients are similar to cochlear implant patients and normal hearing listeners in their ability to detect and discriminate basic sound patterns8 their pitch, loudness and temporal resolution were all relatively normal. However, although the basic sensations were similar to those of cochlear implant patients and speech recognition improved slowly over time, the best performance was still poor compared with those of patients with cochlear implants (Fig. 2)9. An alternate ABI uses penetrating microelectrodes (PABI Fig. 1) to achieve more direct neural stimulation, with some specificity along the tonotopic axes of the cochlear nucleus. These electrodes achieve the desired goal of improved stimulation specificity, but do not improve speech recognition performance10,11. Although the perceptual ���pieces��� were improved, PABI patients have no better outcomes for speech than those of patients with surface ABIs (Fig. 2). This raises the questions of whether intrinsic neural processing in the cochlear nucleus is essential for speech understanding and whether this processing is being bypassed or distorted by direct electrical activation. The answers come from individuals with a different (non-NF2) etiology achieving cochlear implant���like performance with an ABI (Fig. 2)12. The difference between the early results and these newer ones appears to be related to the patient etiology rather than to changes in the device or its surgical placement or to intrinsic cochlear nucleus processing. Apparently, coarse multichannel electrical stimulation of the cochlear nucleus does provide a signal that is adequate for good speech recognition. The etiological difference between the two patient groups suggests that a region of the cochlear nucleus, which is essential for speech recognition, is damaged during the surgery on NF2 patients. The only psychophy- sical measure that was correlated with the extent of speech recognition in the two ABI groups was detection of amplitude modulation. This suggests that there is a separate pathway, distinctive as early in the auditory system as the cochlear nucleus, that is essential for detecting modulation and recognizing speech patterns. Excellent speech recogni- tion is possible with cochlear implants and ABIs using only coarse spectral information modulated at relatively low rates. This suggests that spectral and temporal fine structure information, although poten- tially important for musical pitch and localization of sounds in space, are not essential for obtaining high levels of speech recognition, at least in quiet. Recently, auditory midbrain implants (AMI Fig. 1) have been developed and applied to six individuals. AMIs have an array of either 16 electrodes that are placed on the surface13 or 20 electrodes that PABI AMI High Cable DCN VCN Medium Low ABI Stump of cochlear nerve Cochlear implant 0 20 40 60 80 100 0���20 21���40 41���60 61���80 81���100 Correct words in sentences (%) ABI, NF2 (n = 220) PABI, NF2 (n = 10) Non-NF2 (n = 29) CI (n = 30) Percentage of implant users Figure 2 Word recognition performance of implant users. Data are the percentage of users of the various implant types achieving different levels of word-in-sentence performance on standardized tests in two laboratories. Most ABI users have the NF2 variety of Schwann cell pathology. Data for ABI, PABI and cochlear implant are from the House Ear Institute and those for non-NF2 are from the University of Verona (R.V.S., unpublished data)12. Figure 1 Electrical stimulation of the human auditory system. In addition to the cochlear implant, introduced into the cochlea through the middle ear and stimulating spiral ganglion neurons of the auditory nerve through the modiolus in the core of the cochlear spiral, other methods for stimulating the brain directly have been developed. ABIs come in two varieties. The more common is a flat electrode button array that is positioned on the dorsal surface of the cochlear nucleus. PABIs are also targeted at the cochlear nucleus, but consist of multiple, sharp microelectrodes that are advance through the pia, into the ventral cochlear nucleus. The most recent development has been the AMI, which consists of multiple stimulating sites along a single, penetrating electrode that is targeted at the central nucleus of the inferior colliculus. Multiple electrode cochlear implants are positioned adjacent to normally higher tone-frequency sites in the basal and middle turns of the cochlea, whereas PABIs and AMIs have stimulating sites that are aligned with the tonotopic axis of their target nuclei. NATURE NEUROSCIENCE VOLUME 12 [ NUMBER 6 [ JUNE 2009 687 PERSPECTIVE �� 2009 Nature America, Inc. All rights reserved.