Contribution of the efferent system to auditory perception

Abstract

This third lecture focused on the mechanisms by which a listener's perception of a sound is modulated by feedback from the efferent auditory system on the ascending auditory pathways, once the sensory information (sound, visual, tactile, or other) has been processed. This modulation may take place consciously during attentive listening, but it is also largely unconscious, acting as a high-level filtering that the auditory system exercises continuously, and which depends crucially on multisensory integration, some aspects of which we saw in the previous lecture. The question of where and how this integration takes place has led researchers to question the impact of the disparity and reliability of the parameters that characterize the different sensory signals. How are disparate sound and visual parameters taken into account and processed by the brain to form a distinct audio-visual object ? This question naturally led us to consider the role of statistical inference in multisensory integration. The first part of the lecture was therefore devoted to introducing this notion through its most natural model, Bayesian inference.

Behavioral studies indicate that, when subjects are faced with a perceptual task involving several sensory signals of different nature, origin and reliability (as in the coktail party experience), they arbitrate between two processing strategies : 1) segregation of the distinct signals, which are processed independently ; 2) fusion of signals from the same source, which then form multimodal representations or percepts. Computationally, these trade-offs can be modeled by a type of Bayesian inference known as causal inference . In these models, brain representations of the various sensory signals are combined with plausibility constraints to enable a posteriori evaluation of the most probable value of the parameters characterizing these signals (source positions, intensities, frequencies, etc.). This value can then be taken into account when performing a task such as causal inference (to decide whether to process the signals by segregation or fusion). The strength of the Bayesian approach lies in the fact that it enables these probabilities to be continuously updated as a function of the flow of sensory signals with which the subject is confronted.

Following this introduction, we discussed a recent study by Yinan Cao and colleagues [10] which reveals the existence of causal inference mechanisms in audio-visual multisensory integration, while characterizing for the first time the locus and temporal course of this integration. To this end, the authors used magnetoencephalography to probe the brain activity of human subjects with high temporal resolution, while they performed an audiovisual perceptual categorization task (they were asked to characterize the frequency of a train of sound pulses perceived in the presence of a similar train of light flashes). Through an elegant series of experiments and advanced statistical analyses, they were able to demonstrate a gradual, sequential progression in the functional sites and temporality of activation of the different modes of integration at work in their experiment. Their results reveal that processing corresponding to the segregated model (strategy 1) is activated first in the primary auditory and visual cortexes ; it is rapidly followed by the fusion model (strategy 2) whose activation takes place mainly in the associative parietal areas of the auditory cortex ; then comes the causal inference model (in which a Bayesian trade-off is made between the segregated model and the fusion model), which is activated in regions of the frontal cortex. We are thus in the presence of a temporal and spatial hierarchy in multisensory processing. A privileged role for the frontal lobe in causal inference, a role hinted at by previous anatomical and behavioral studies, is, moreover, revealed for the first time in a very direct way.

The second half of the lecture was devoted to the auditory efferent system proper, i.e. all the descending pathways of the auditory system (still referred to as " corticofugal pathways "), present in mammals and also in certain invertebrates such as mosquitoes and spiders. We began with some general anatomical and neurophysiological information, focusing first on the role of the olivocochlear system, first described by Grant Rasmussen in the 1940s and 1960s, and still the most studied, having long been considered synonymous with the efferent auditory system. It should be remembered that this system is made up of two sub-systems, known as the " olivocochlear medial and lateral ", which are made up of very different neurons that project onto the cochlea in different ways : the medial bundle projects mainly and directly onto the outer hair cells, while the lateral bundle projects mainly onto the afferent neurons of the inner hair cells. It seems well established that the role of the olivocochlear efferent system is to modulate cochlear activity. This modulation is far from understood in detail, but it includes control of the gain of cochlear amplification produced by the outer hair cells, as well as modulation of the response of the primary auditory neurons that transmit auditory information encoded by the inner hair cells. Similar effects have so far not been clearly established in humans, where direct measurement or manipulation of olivocochlear system activity is not possible. The physiological role of the efferent system in humans (and even in animals) therefore remains controversial. However, it is thought to play an important role in speech perception in the presence of noise, by reducing the masking effect of noise through control of cochlear gain by the medial olivocochlear bundle. The role of lateral olivocochlear efferences is even less clear ; these may be involved in protecting the auditory system from loud sounds, as well as in maturing the electrical response of inner hair cells during development.

Nevertheless, the efferent system is not reduced to the olivocochlear system (which itself receives information from efferent relays located at higher levels), but comprises a set of feedback loops running from the auditory cortex to the cochlea. Recent attention has focused on the role of the corticofugal projections of the auditory efferent system, in particular those projecting to the inferior colliculus. It's not the least paradoxical that this role has been uncovered in recent studies, while that of the olivocochlear system, which has been studied for much longer, is still largely unknown. This is probably because optogenetic tools have so far been neglected in the study of the olivocochlear system, whereas they have been used systematically over the last five years in studies of the upper stages of the auditory efferent system.

To illustrate this point, we ended the lecture by briefly discussing a study by Li Zhang's group at the University of California [11], focusing on the role of corticofugal projections on the inferior colliculus in innate sound-induced escape behavior in mice. By finely deciphering, using optogenetic tools, not only the activation of corticofugal efferent pathways but also the effect of this activation in the escape reflex, the authors were able to highlight a circuit that directly conditions this reflex. This circuit involves ascending pathways from the central inferior colliculus to the auditory cortex (via the median geniculate body of the thalamus), and within the descending corticofugal pathways, pyramidal neurons from the auditory cortex projecting to the inferior colliculus cortex. As it exits the inferior colliculus cortex, this loop controls the response of the dorsolateral part of the periaqueductal gray matter, whose involvement in other non-innate (Pavlovian-type) defense reflexes has been demonstrated in previous studies.

Thus, a morphofunctional deciphering of the auditory cortex's efferent projections to both the thalamus and inferior colliculus is gradually taking shape.