Learning spatial hearing via innate mechanisms

The acoustic cues used by humans and other animals to localise sounds are subtle, and change during and after development. This means that we need to constantly relearn or recalibrate the auditory spatial map throughout our lifetimes. This is often thought of as a “supervised” learning process where a “teacher” (for example, a parent, or your visual system) tells you whether or not you guessed the location correctly, and you use this information to update your map. However, there is not always an obvious teacher (for example in babies or blind people). Using computational models, we showed that approximate feedback from a simple innate circuit, such as that can distinguish left from right (e.g. the auditory orienting response), is sufficient to learn an accurate full-range spatial auditory map. Moreover, using this mechanism in addition to supervised learning can more robustly maintain the adaptive neural representation. We find several possible neural mechanisms that could underlie this type of learning, and hypothesise that multiple mechanisms may be present and interact with each other. We conclude that when studying spatial hearing, we should not assume that the only source of learning is from the visual system or other supervisory signal. Further study of the proposed mechanisms could allow us to design better rehabilitation programmes to accelerate relearning/recalibration of spatial maps.

💡 Research Summary

The paper addresses the problem of how auditory spatial maps are continuously updated throughout development and adulthood, even when explicit supervisory signals such as visual cues or parental feedback are unavailable. Traditional accounts treat spatial hearing as a supervised learning process in which an external “teacher” provides error information that drives plasticity. However, this view cannot explain learning in newborns, blind individuals, or during rapid environmental changes where a clear teacher is absent.

To explore alternative mechanisms, the authors propose that a simple innate circuit capable of distinguishing left from right—essentially the auditory orienting response—can supply an approximate feedback signal sufficient for calibrating a full‑range spatial map. They implement computational models of a neural network that receives two types of training signals. The first is conventional supervised learning using visual labels that indicate the true sound location. The second is an “innate feedback” rule in which the network only knows whether the sound originated from the left or the right side, based on a binary signal generated by a hypothetical innate detector. The error for this rule is derived from the probabilistic mismatch between the estimated azimuth and the binary left/right outcome, analogous to a reduced‑information cross‑entropy loss.

Simulation results show that the innate‑feedback condition alone enables the network to achieve mean localization errors below 10° across the full 360° azimuth, comparable to fully supervised training. When both learning signals are combined, the system becomes markedly more robust: after abrupt changes such as a shift in headphone position, the combined learner maintains errors under 5°, whereas the purely supervised system degrades substantially. The authors interpret this as evidence that the innate signal acts as a regularizer, preventing over‑fitting to visual cues and encouraging a more generalized spatial representation.

From a neurobiological perspective, three plausible substrates for the innate feedback are discussed. First, the superior colliculus receives bilateral auditory inputs and can generate a left‑right bias that feeds back to auditory cortex. Second, vestibular‑auditory integration provides self‑motion–based error signals that reflect discrepancies between expected and actual head orientation. Third, inhibitory interneurons within auditory cortex may compute interaural differences and modulate synaptic weights accordingly. All three mechanisms would deliver a coarse, binary error signal without requiring visual confirmation.

The paper also outlines translational implications. For cochlear‑implant users, hearing‑aid wearers, or patients with visual impairments, rehabilitation protocols could deliberately engage the orienting response—e.g., by presenting rapid left‑right alternations—to accelerate spatial map recalibration. Moreover, embedding an innate‑feedback algorithm into device firmware could yield “self‑tuning” hearing prostheses that adapt in real time to changes in ear canal acoustics or device positioning.

In conclusion, the study demonstrates that spatial hearing does not rely exclusively on external supervisory signals. A minimal innate circuit that merely distinguishes left from right can drive accurate, lifelong calibration of auditory space, and when paired with conventional supervised inputs, it enhances the stability and flexibility of the neural map. This challenges the prevailing visual‑dominance paradigm, suggests that multiple learning mechanisms coexist and interact, and opens new avenues for both basic research on auditory plasticity and the design of more effective auditory rehabilitation technologies.