Cochlear implants can preserve the timing signal the brain needs for localization—and still make spatial hearing worse. This paper points to an unexpected culprit: neural timing that is almost too precise. Sound localization is essential for communication and navigation, and spatial hearing relies predominantly on comparing differences in sound arrival time between the two ears, called interaural time differences, or ITDs.

With bilateral cochlear implants, many hearing capabilities can be restored, but sound localization through ITD detection remains poor, and the underlying reasons are not well understood. The unresolved question is whether the deficit begins in binaural comparator neurons or already in their inputs.

The study compares electric and acoustic stimulation in gerbils using extracellular single-neuron recordings. Those recordings found severe hyper-precision and moderate hyper-entrainment in both excitatory and inhibitory brainstem inputs during electrical pulse-train stimulation, so the model then estimated the consequences for ITD sensitivity.

The model predicted maintained ITD sensitivity for electrical pulses, but severely altered spatial sensitivity compared with acoustic stimulation. The location of a sound source is neuronally determined by binaural comparison of interaural time and level differences, or ITDs and ILDs.

Cochlear implants extract the amplitude envelope of sounds in multiple spectral channels and pass that information to auditory-nerve fibers by modulating an electrical pulse train. Those auditory-nerve fibers provide input to brainstem nuclei involved in ITD and ILD detection, but bilateral CI users generally retain ILD sensitivity while ITD sensitivity remains very coarse and mostly resembles lateralization.

CI-based ITD sensitivity is limited to electrical pulse rates below 500 pulses per second, and the physiological reasons for this limitation remain poorly understood. Acoustic ITDs are primarily detected by neurons in the medial and lateral superior olives, with the medial superior olive regarded as primarily detecting fine-structure ITDs at low frequencies.

Lateral superior olive neurons receive excitatory input from the ipsilateral antero-ventral cochlear nucleus and inhibitory input from the medial nucleus of the trapezoid body, which is innervated by the contralateral antero-ventral cochlear nucleus. Both antero-ventral cochlear nucleus and medial nucleus of the trapezoid body neurons lock action-potential timing to stimulus-waveform phases, enabling lateral superior olive neurons to detect differences in input timing.

Auditory-nerve phase locking is severely heightened during electrical stimulation compared with acoustic stimulation, by up to approximately a factor of ten in synchronization index at frequencies below three kilohertz. Before this study, it was not known whether auditory-nerve hyper-precision was maintained in the antero-ventral cochlear nucleus and medial nucleus of the trapezoid body or instead caused degraded processing.

Good lateralization of electrical pulse ITDs in rats suggested that the lateral superior olive could provide behaviorally relevant information during bilateral cochlear-implant stimulation. Still, the effects of electrical stimulation on information carried by lateral superior olive inputs, and on ITD sensitivity in the lateral superior olive itself, were unclear.

An investigation of the lateral superior olive pathway during electrical stimulation was missing. The study uses a bottom-up approach to compare response features between electrical and acoustic stimulation within the initial stages of the lateral superior olive circuit.

Extracellular single-neuron recordings compare acoustic clicks and electrical pulses in the antero-ventral cochlear nucleus and medial nucleus of the trapezoid body of gerbils. Neurons in both monaural nuclei exhibit severe hyper-precision and moderate hyper-entrainment during electrical stimulation, meaning ITD detection must cope with highly altered input statistics.

A previously published lateral superior olive model was adapted to match the physiological recordings, and it reproduced observed ITD sensitivity and its dependence on interaural level difference. The model maintained ITD sensitivity for electrical pulses but narrowed the dynamic range: resolution increased at small ITDs, while larger ITDs became inseparable.

The recordings targeted bushy cells in the antero-ventral cochlear nucleus and principal cells in the medial nucleus of the trapezoid body, using two groups of gerbils. The acoustic group received six-click trains with inter-click intervals from five to one milliseconds, corresponding to click frequencies from 200 to 1,000 pulses per second.

The electrical group received six-pulse trains with the same inter-click intervals, but stimulation artifacts prevented analysis of the one-millisecond interval. The recordings included 22 antero-ventral cochlear nucleus and 18 medial nucleus of the trapezoid body neurons during acoustic stimulation, plus 11 and 9 neurons during electrical stimulation.

The analysis focused on response reliability and timing accuracy, two factors relevant to generating ITD sensitivity. Spike probability was the average percentage of clicks that elicited action potentials; because each train had six clicks, six action potentials corresponded to 100 percent spike probability.

Response jitter was the standard deviation of action-potential latency relative to the click that elicited it. Figure one connects sound location with the binaural cues of interaural time and level differences, then shows where recordings were made in the auditory brainstem: the AVCN, MNTB, and LSO.

The authors compared acoustic and cochlear-implant stimulation using six-click trains with inter-click intervals from five to one milliseconds. Dot rasters from MNTB neurons, repeated twenty times per interval, reveal identifiable click responses and changes in spike timing variability and response probability as clicks become more closely spaced.

The largest difference from acoustic stimulation was temporal precision: jitter during electrical stimulation was approximately ten-fold smaller, with all inter-click intervals for medial nucleus of the trapezoid body and bushy cells producing p values below zero point zero zero one.

The data show that electrically induced hyper-precision found in auditory-nerve fibers is conserved, or even increased, in downstream brainstem nuclei. That conserved hyper-precision is likely to influence binaural spatial processing in the medial and lateral superior olives.

Figure two compares AVCN and MNTB responses as click intervals shorten from five to two milliseconds, for acoustic stimulation in panel A and electrical stimulation in panel B. Thin lines show individual neurons, while bold blue and orange lines show sample medians; the upper row measures spike probability and the lower row measures jitter, or timing variability.

The visual matters because electrical stimulation produces markedly tighter timing, while response probability changes across nuclei and stimulation conditions—differences relevant to downstream processing of interaural time differences. Table one summarizes median spike probability and timing jitter across inter-click intervals from five to one milliseconds.

For acoustic stimulation, MNTB probabilities range from ninety point eighty-three percent to thirty-five point eighty-eight percent, while BC probabilities range from eighty-four point fifty-eight percent to twenty-seven point ninety-two percent; corresponding jitter values are reported in milliseconds.

For electrical stimulation, the table reports MNTB and AVCN probabilities and jitter, providing the quantitative basis for the paper’s comparison of temporal coding under the two stimulation modes. For ITD detection, the lateral superior olive and medial superior olive integrate ipsilateral bushy-cell inputs and contralateral medial nucleus of the trapezoid body inputs for each click.

Earlier recordings showed high lateral superior olive sensitivity to the ITD of each click, with response modulation depending on the relative strength and timing of inhibition compared with excitation. That makes the unusually high precision in both excitatory bushy-cell and inhibitory medial nucleus of the trapezoid body inputs a direct candidate for changing temporal integration in the lateral superior olive.

The study recorded 15 lateral superior olive neurons during binaural acoustic click-train stimulation at various ITDs, and response rates were strongly modulated across a wide range of tested ITDs. At zero interaural level difference, the dynamic ITD range covered or exceeded the gerbil physiological range of approximately 300 microseconds for every inter-click interval.

The median dynamic ranges were 400 microseconds for five-, four-, three-, and two-millisecond intervals, and 600 microseconds for the one-millisecond interval. The wide range supports hemispheric rate-difference coding, in which individual ITDs are encoded through relative activity levels between the two lateral superior olive populations.

Changing interaural level difference had a small to modest average effect across all inter-click intervals, suggesting that ITD sensitivity can be maintained over a wide range of binaural conditions. Figure three shows that gerbil LSO neurons produce strongly ITD-dependent firing patterns across click trains, with panel A plotting mean spikes per repetition and standard errors for several inter-click intervals.

Panel B summarizes the dynamic ITD ranges across fifteen neurons, while panel C quantifies how the tuning slope changes with interaural level difference for neurons and the model. Panel D shows that the model reproduces monotonic ITD rate modulation during acoustic stimulation at zero decibels ILD, supporting the model’s use for studying ITD coding across level differences.

Direct recording from a single lateral superior olive neuron during bilateral electrical cochlear-implant stimulation proved exceedingly difficult. To approximate physiological data, the study therefore used a functional count-comparison model of the lateral superior olive as a surrogate during electrical stimulation.

The model can replicate typical lateral superior olive responses and accompanying spatial perceptions across a wide range of stimulus classes, and it allows acoustic and electrical properties to be read out at stages of the pathway. This made it possible to benchmark the model against the recorded antero-ventral cochlear nucleus and medial nucleus of the trapezoid body data before predicting changes in ITD sensitivity.

To evaluate changes in ITD coding and spatial resolution, the analysis determined the informational content of the model response. The model had to distinguish adjacent ITDs separated by 20 microseconds across approximately plus or minus 600 microseconds, the full inter-ear-distance range generated by the human head.

The analysis used standard separation, called D, which quantifies adjacent-ITD separability from the ratio of differences in mean rate and response variability. Because the model is highly deterministic, Poisson noise was used as a conservative assumption, and D was based on the rate difference between lateral superior olives in opposite hemispheres.

Under acoustic stimulation, the model’s rate-ITD functions spanned almost the entire human physiological range of plus or minus 600 microseconds, with peaks centered on midline or slightly lateralized depending on inter-click interval. Under bilateral electrical stimulation, ITD sensitivity was maintained: the model showed steep rate modulation for every inter-click interval.

But the modulation extended over only a narrower range, approximately between minus 100 and zero microseconds, while responses became effectively identical for more lateralized ITDs. This produced high separability near midline, even higher than with acoustic stimulation, but no separability at more lateralized positions.

The pattern resembles lateralization: left can be distinguished from right near midline, but resolution is almost absent for larger ITDs within each hemisphere, unlike the graded acoustic distribution. Figure five compares model LSO rate responses to interaural time differences under acoustic stimulation, electrical stimulation, and electrical stimulation with added jitter.

The rate–ITD curves and the separability measure D show that electrical stimulation preserves ITD sensitivity, but concentrates its strongest separation near midline; adding jitter broadens and reshapes this pattern toward the acoustic case. This matters because it links stimulation variability to the model’s predicted range and precision of binaural timing coding.

The most drastic input change during electrical stimulation was increased response precision, meaning decreased jitter. To test its effect, jitter was reintroduced into the lateral superior olive model at acoustic-like values of zero point seventeen milliseconds for excitation and zero point sixteen milliseconds for inhibition.

This modification restored a wider dynamic ITD range and widened the range of separable ITDs. The model therefore suggests that increased spiking precision from artificial electrical stimulation directly alters lateral superior olive ITD coding by changing the range of ITDs that alter response rates.

The study directly assessed brainstem-neuron responses to cochlear-implant stimulation and found severe hyper-precision and moderate hyper-entrainment in both bushy cells and medial nucleus of the trapezoid body neurons. The response alterations previously reported in the auditory nerve during cochlear-implant stimulation were passed on to subsequent synaptic stages in the brainstem.

Our gerbil antero-ventral cochlear nucleus recordings showed no significant transmission failure or compensation below five hundred pulses per second during electrical stimulation. About half of the medial nucleus of the trapezoid body neurons showed high spike probabilities even at 500 pulses per second, suggesting slightly enhanced fidelity compared with bushy cells, although not in every cell.

The next binaural stages must therefore cope with highly altered input statistics, and diminished perceptual ITD sensitivity is not caused by a lack of temporal information reaching binaural integration. The increased separability measured by D decreases considerably as variability in the lateral superior olive population code increases.

Because the model is highly deterministic and actual in vivo population variability was unknown, the analysis had to assume a variability level and used Poisson noise. That choice might underestimate true in vivo variability, particularly during electrical stimulation, but the loss of spatial resolution at larger ITDs is mostly independent of this estimate.

The experiments used animals with fully developed hearing and no degenerative effects before implantation, leaving experience-dependent tuning of input strength and timing intact. By contrast, patients typically undergo prolonged deafness before receiving cochlear implants, and this prolonged inactivity can degrade the auditory system.

That difference matters because degradation during auditory inactivity could create complications not represented in the animal preparation. The authors propose that the lateral superior olive may be the main detector for CI-based ITDs because its circuits develop to functional maturity even without auditory experience, unlike medial superior olive ITD sensitivity.

However, electrophysiological recordings of cellular integration in the lateral and medial superior olives during CI stimulation have not yet been obtained. The implanted-rat rate-ITD functions also do not readily indicate excitatory-inhibitory interaction and cannot be reproduced by this model.

The summary identifies lateral superior olive processing as likely the main site of electrical CI-mediated ITD processing. It also identifies temporal hyper-precision in inputs to the binaural comparator stage as a key problem underlying diminished ITD sensitivity in cochlear-implant users.

The study links cochlear-implant stimulation to hyper-precise brainstem inputs and a narrowed range of ITDs that can be distinguished in the model. Restoring acoustic-like jitter widened that range again.