Natural auditory behaviors invoke cognitive brain networks

Abstract

Auditory processing is traditionally framed as a hierarchical extraction of acoustic features along a canonical ascending pathway, starting in the cochlea and engaging dedicated auditory processing regions in the brainstem and midbrain on the way to the auditory cortex. Yet real-world listening scenarios, such as those found in a cocktail party (i.e., within a group of simultaneously vocalizing humans and other animals), invoke cognitive processes to facilitate the segregation of overlapping sources, suppress self-generated signals, and activate rapid sensorimotor adjustments. In this review, we argue that auditory perception, particularly in challenging listening situations, is best understood as an emergent property of the interactions between the canonical auditory pathway and distributed networks supporting attention, prediction, memory, and action selection. We propose the echolocating bat as a powerful model system with a rich behavioral repertoire and conserved mammalian neural architecture that can be leveraged to uncover general principles of natural sound processing.

Introduction

Auditory processing has traditionally been viewed as a transformation of information about incoming sound into its fundamental parameters (frequency, amplitude, and timing), which are transmitted along the ascending central pathway to form auditory percepts. However, this traditional framing of sound processing is increasingly difficult to reconcile with complex auditory behaviors in natural settings. The classic cocktail party phenomenon [1] illustrates that auditory perception cannot operate by basic acoustic feature extraction alone. In a cocktail party (or a chorus of frogs, birds, or other animals), multiple simultaneous acoustic signals produced by different speakers at different locations overlap in time, and a listener is tasked with decomposing complex acoustic stimuli into coherent information [2]. This requires computations that extend beyond the parcellation of acoustic features and must engage cognitive processes, such as attention, prediction, memory, and action selection (Figure 1A).

Figure 1.

Auditory-cognitive behaviors exemplified by the perceptual demands of the cocktail party. (A) Complex acoustic environments (e.g., a cocktail party with multiple competing sound sources (left) and echolocating bats foraging in the presence of conspecifics (right) recruit cognitive processes invoking attention, prediction, memory, and goal-directed action selection to facilitate auditory processing during natural behaviors. (B) A simplified schematic of the processing centers comprising the canonical mammalian auditory pathway (left) and the distributed processing centers (right, in orange) that integrate with the auditory pathway (in blue) to support complex auditory-cognitive behaviors. Abbreviations: AC – auditory cortex, BG – basal ganglia, Cbl – cerebellum, CN – cochlear nucleus, FAF – frontal auditory field, Hipp – hippocampus, IC – inferior colliculus, NCAT – nucleus of the central acoustic tract, NLL – lateral lemniscal nuclei, MGB – medial geniculate body, mPFC – medial prefrontal cortex, SC – superior colliculus, SOC – superior olivary complex.

Here, we posit that auditory perception in natural scenes emerges from the interaction between canonical auditory processing streams (Figure 1B, left) and cognitive networks that activate a distributed set of brain regions not traditionally considered part of the auditory system (Figure 1B, right). We highlight data from several “non-canonical” sound-processing centers, including the hippocampus, basal ganglia, cerebellum, and superior colliculus (SC), and discuss their putative roles in auditory behaviors. The importance of inputs from such non-canonical auditory processing centers is increasingly recognized [3–5], and we consider a subset of recent works that demonstrate how sound processing in behaving animals is shaped by the distributed networks subserving cognitive and motor tasks. We note, however, that this review is not intended to be exhaustive of the growing body of literature in this field.

In this review, we define auditory-cognitive behaviors as behaviors in which sound processing is dynamically modulated by cognitive variables, rather than passive stimulus processing. We feature the echolocating bat as an instructive research model to probe the neural underpinnings of auditory-cognitive behaviors, such as sonar-guided navigation and target tracking, as well as acoustic communication in noisy environments (Box 1). Research findings from echolocating bats offer a framework to compare data from more traditional animal research models, such as mice and humans, to begin differentiating species-specific observations and general mechanisms that support auditory-guided behaviors across taxonomic scales.

BOX 1: Echolocating bats as a powerful model for auditory-cognitive behaviors.

Echolocating bats rely heavily on hearing to guide a wide range of natural behaviors, including prey capture and obstacle avoidance. In addition to passive listening (e.g., to communication calls or to prey-generated sounds), bats also engage in active sonar sampling to represent objects in their environment. Crucially, they actively control the timing, duration, intensity, and frequency content of their sonar emissions while tracking dynamic acoustic feedback from returning echoes. Therefore, their movements directly influence the sensory information available for processing [6]. In turn, the bat’s perception of the acoustic scene guides the production of subsequent vocalizations, forming a closed-loop feedback system between action and perception. Beyond control of the spectro-temporal features of their echolocation calls, bats actively shift the directional aim of their sonar signals and coordinate head and ear movements, yielding an explicit metric of overt spatial attention to objects in their surroundings [6]. Since echo reception and call production are temporally discrete and directly measurable events, the echolocating bat presents an exceptionally powerful model system to probe interactions between sensing and action, and to understand how cognitive variables shape auditory behavior in natural environments.

Auditory Attention

The sensory challenges highlighted by the cocktail party may be distilled into the key task of selective attention. When multiple auditory inputs converge on a receiver, including those from sources that may be closer and louder than the sound source of interest, it is essential that the auditory system perceptually segregate and reorganize complex multi-source sounds into auditory ‘streams’ (sensu Bregman [7]). This is generally seen as an enhancement of neural responses elicited by the attended sound and the relative suppression of responses to unattended sounds, mediated by both bottom-up sensory gating and top-down control of neural responses [8]. An echolocating bat may encounter multiple sound sources, its own echo returns, as well as the calls and echoes produced by other bats, which it can parse through active control over the directional aim, duration, and spectral content of sonar calls [6] and pinna movements [9]. The bat’s active adjustments in sonar call features not only minimize acoustic interference from neighboring bats but also reveal its attention to objects in its surroundings.

Auditory selective attention engages cortical networks that govern attentional control of other sensory modalities [10,11]. For example, the anterior cingulate cortex (ACC) contributes to goal-directed attentional processes that are often explored in the context of visual- or motor-based decision-making tasks [12,13]. Notably, the ACC also demonstrates functional connectivity to auditory regions, and its activity is selectively enhanced by conspecific vocalizations in humans and non-human primates [14,15], during auditory attentional tasks [16,17] and under challenging listening conditions [18–20]. Activity in the rodent homolog of the primate ACC, the medial prefrontal cortex (mPFC), is similarly enhanced during challenging auditory tasks [21] and by unexpected auditory stimuli [22]. Strong reciprocal connections between the mPFC and auditory cortex (AC) and between the mPFC and inferior colliculus (IC) of echolocating bats [23,24] may support similar context-dependent attentional shifts during auditory-cognitive tasks.

Electrophysiological recordings from a region of the prefrontal cortex termed the frontal auditory field (FAF) in bats reveal strong selectivity for species-specific call types [25–27]. Not only does the FAF have reciprocal connections with the AC, but it is also part of a specialized, non-lemniscal pathway (‘the central acoustic tract’) linking brainstem processing centers directly to the SC and thalamus, bypassing the canonical route though the IC [28]. This noncanonical pathway is hypothesized to contribute to fast transmission of auditory information to premotor and motor control centers, enabling synchronization of the vocal-motor behaviors required to rapidly emit and receive acoustic information. Such a direct pathway can facilitate the coordination of pinna movements with the expected direction and timing of returning echoes and, in species that emit constant frequency sonar signals, may support rapid modulation of call frequency to stabilize the spectral content of returning echoes (Doppler shift compensation) [29]. Although the FAF in bats is proximal to the medial prefrontal region of other mammals [23], and may similarly contribute to contextual modulation of auditory attention, we note that the homology of this structure with the mPFC or ACC has not yet been confirmed. Nevertheless, we emphasize that the reciprocal circuits connecting auditory centers to attention networks, in some cases bypassing canonical auditory nuclei, enable highly dynamic and adaptable responses under challenging listening conditions. The inherent utility of these auditory attention networks is exemplified by the echolocating bat that must integrate sound production and reception behaviors with rapidly changing acoustic information as it navigates through complex soundscapes comprised of target echoes and environmental noise.

Auditory Prediction

A fundamental challenge in complex auditory tasks lies in analyzing the auditory scene in the presence of self-generated vocalizations and other competing sound sources. Predictive processing of acoustic signals allows animals to build coherent percepts by distinguishing expected sounds (e.g., sounds that are self-generated or that are otherwise anticipated within the context of known/learned patterns) from unexpected sounds. Separating received sounds into self-generated versus external sources relies on efference copy, in which incoming sounds are compared with an internal representation of the motor command driving sound generation. This mechanism modulates the processing of self-produced speech in human listeners [30]. Similarly, peripheral and central mechanisms attenuate responses to self-produced sonar signals in bats [31–33]. In both situations, prediction serves as a foundational operation to represent the auditory scene.

Canonical models of auditory processing have long emphasized the suppression of self-generated sounds within the auditory pathway itself. During speech production in humans, responses in the AC to one’s own vocalizations are suppressed [34]. This mechanism is widely conserved across taxa, with similar vocal-induced suppression observed in the AC of non-human primates and mice, as well as the AC-analog of songbirds [35–37]. Notably, a small population of neurons in the IC of echolocating bats exhibit suppressed responses during self-produced vocalizations [32], suggesting that predictive processing modulates auditory representations to maintain perceptual stability. While these responses are typically measured in the AC, their precise timing and rapid error correction raise a fundamental question: what brain networks give rise to the predictive models that enable such suppression to be computed?

Predictive computations are hallmark features of the cerebellum, a region outside of the canonical auditory processing pathway. Audio-motor integration in the cerebellum enables predictive filtering of self-generated sounds in addition to rapid correction of vocalization errors. A growing body of evidence suggests that the cerebellum plays a direct causal role in shaping auditory processing by generating predictive models of the sensory consequences of actions [38–40]. In humans, perturbation of cerebellar function impairs auditory feedback during speech production through reduced attenuation of self-generated sounds [38]. Further, the cerebellar system mediates temporal prediction even during passive listening tasks [41], indicating that cerebellar timing mechanisms operate broadly across auditory contexts. Recent work in echolocating bats support this view, showing that cerebellar activity predicts the class of vocalization (echolocation/communication) that the bat is about to produce [42]. Together, these findings suggest that cerebellar predictive mechanisms form an integral component of distributed circuits supporting auditory perception.

Prediction in auditory processing is not implemented by the cerebellum alone. Predictive mechanisms also include basal ganglia circuits, which generate expectations of when auditory events will occur and shape action selection based on learned contingencies. Disruptions to the basal ganglia reduce sensitivity to temporal regularities in auditory sequences [43] and affect the ability to generate temporal predictions in both simple and complex auditory sequences [44]. In parallel, neuromodulatory systems further influence processing by regulating the gain of auditory representations and determine how strongly prediction errors are influenced by behavioral state [45]. These computations are particularly important to foraging in echolocating bats, where auditory representations would be dominated by self-vocalizations and that of nearby conspecifics without predictive filtering.

Predictive mechanisms in bats are tailored to the behavioral function of the vocalizations they emit. They produce two distinct classes of vocalizations, communication and echolocation, which serve distinct behavioral goals. Recent work demonstrates that information flow is predominantly top-down from frontal to auditory regions during the production of communication calls, while a reversal to bottom-up flow occurs during echolocation to prioritize the processing of self-directed acoustic feedback [46]. Crucially, the class of vocalizations that the bat is about to produce can be predicted from frontal and auditory cortical activity before emission, demonstrating that the auditory processing network dynamically reconfigures by shifting between top-down and bottom-up information flow based on vocal output [47]. Collectively, these findings highlight predictive categorization as a fundamental operation distributed across the auditory hierarchy for processing externally- and self-generated sounds.

Auditory Memory

Auditory processing in natural environments is strongly shaped by prior experience. For example, humans can separate strangers’ voices using acoustic and spatial cues, but recognition improves when a voice is familiar [48]. Similarly, during prey interception, echolocating bats interpret the arriving stream of echoes by leveraging prior experience with prey motion trajectories and the spatial layout of their environment [49,50]. In both scenarios, memory actively influences auditory-cognitive processes.

Classical models of the medial temporal lobe, specifically the hippocampus, have long been implicated in episodic and spatial memory [51]. However, emerging evidence supports a broader, domain-general function of medial temporal lobe circuits, namely the integration of sensory input with learned structure [52,53]. In humans, neuroimaging studies have shown that the hippocampus exhibits sustained activation and functional connectivity with the AC during auditory working memory maintenance [54,55]. Furthermore, recent findings show bidirectional interactions between the hippocampus and AC, including hippocampus-to-AC information flow [56], and increased cortico-hippocampal coupling during auditory working memory maintenance [57]. Additionally, hippocampal networks participate in learning and recognition of structured auditory sequences [58]. Collectively, these findings implicate hippocampal networks in conferring contextual and relational information that shapes auditory-cognitive processing.

Beyond working memory, the hippocampus is implicated in associative novelty detection [59]. Neuroimaging and electrophysiological data show that the hippocampus responds strongly to unexpected auditory events that violate learned temporal or sequential regularities in humans [60] and animals [61]. Critically, causal evidence from animal studies demonstrate that hippocampal circuits are necessary for auditory tasks that require bridging temporal intervals, but not tasks that involve immediate stimulus-response associations [62]. Together, these findings indicate that hippocampal mismatch signals reflect violations of internal predictive models, rather than simple detection of acoustic change. More generally, these results imply that the memory circuits become integral during auditory processing of sound sequences with natural temporal structure, such as human speech and bat sonar tracking signals.

These principles naturally extend to echolocating bats, where auditory processing is intrinsically linked to memory-based behaviors. During prey interception, echolocating bats must rely on working memory to track auditory targets across successive call-echo sequences in the presence of background clutter and obstacles. This challenge is further exacerbated during group foraging, when faint echoes can be masked by intense conspecific calls. Memory circuits may alleviate this challenge by linking incoming echo information to stored representations of object identity and environmental layout. Recent work demonstrates that the bat hippocampus encodes the spatial positions of moving auditory objects during active sonar tracking, and these representations emerge only when bats actively probe targets with echolocation calls, not during passive listening [63,64]. These findings align with the previously described role of memory circuits in maintaining perceptual continuity of sound over time, rather than purely encoding auditory features [65]. The bat AC is embedded within large-scale networks that include memory-related regions, positioning auditory processing to interact dynamically with mnemonic systems [24]. These findings compliment previous work emphasizing the integration of sensory and mnemonic processing in active sensing systems [66] and highlight echolocating bats as a powerful model to study this interaction.

While we have focused primarily on the hippocampal formation, memory-related influences on auditory processing are distributed across prefrontal, parahippocampal, striatal, and basal forebrain circuits [54]. Collectively, this large-scale network encodes contextual knowledge and learned associations that bias how auditory scenes are constructed and represented. It is essential to note that memory systems do not operate in isolation but interact closely with predictive and attention networks. It is thus imperative to consider auditory processing as a distributed circuit modulated by multiple cognitive processes and brain regions, rather than confining it to the canonical network.

Auditory action selection

Navigation has been described as the output of sensory processing of environmental features, which are transformed into motor commands that guide movement through space. However, more recent perspectives suggest that navigation invokes a bidirectional inference loop, tightly coupled with perception, prediction, and action [67]. In this view, active sampling of sensory information is shaped by internal models of space and expected sensory outcomes. This framing becomes particularly relevant in auditory-cognitive behaviors, where spatial information must be inferred over time from dynamic, and sometimes ambiguous acoustic signals.

Spatial inference of complex auditory scenes requires coordinated orienting behaviors, predictive filtering, and sensorimotor coupling. For example, head turns, eye movements, and postural adjustments dynamically improve the segregation and tracking of behaviorally-relevant sound stimuli in humans [68]. Similarly, bats coordinate changes in sonar call production with goal-directed head and ear movements to enhance target localization [9]. Critically, the acoustic benefits of these behaviors are amplified during goal-directed actions, in which movements are actively driven by sensory expectations [69,70]. Collectively, these results underscore spatial hearing as an active process in which self-motion modifies perceptual processing.

At the circuit level, motion-related modulation of auditory processing is supported by interactions between sensory, motor, and spatial networks. A central hub is the SC, an evolutionarily-conserved midbrain structure that integrates visual, auditory, and somatosensory information to guide species-specific orienting movements [71]. Comparative work across species has demonstrated that this region exhibits robust auditory responses [72]. The SC contains neurons selective for sound location and temporal features that contribute to a representation of egocentric auditory space [73,74]. Auditory responses in the SC are not purely stimulus driven but are shaped by behavioral context and internal goals. Projections from the AC to the SC provide top-down modulation of neural response selectivity, thereby shaping representations of salient auditory stimuli [75].

Through its extensive connectivity with cortical, thalamic, and brainstem motor circuits [76], the SC is anatomically and functionally situated to direct anticipatory orienting and sensory sampling. Indeed, neurons in the SC encode planned movements before movement onset [77] and optogenetic activation of the SC produces directionally-biased head and eye movements [78]. These (anticipatory) movements, in turn, modulate auditory processing. For example, auditory cortical activity is modulated by head movements, locomotion, and motor commands [79]. These movement signals are interpreted within feed-forward model frameworks, where predicted consequences of actions are used to update internal representations [80]. These findings underscore the putative role of attention, prediction and memory in shaping auditory perception and action selection in humans and other animals. The SC, along with a broader network encompassing hippocampal and parahippocampal circuits, parietal regions, and motor systems, collectively coordinates anticipatory body movements, which in turn modulate sensory processing. Such anticipatory control can facilitate sound processing in natural environments, where stimulus input may be noisy or intermittent.

Conclusions

In this review, we have highlighted evidence that structures external to the canonical auditory pathway, such as the hippocampus, cerebellum, basal ganglia, and SC, play a critical role in shaping auditory processing during natural behavior. Compared to the visual system, the classical auditory pathway contains many more central processing stations dedicated to accurate representation of stimuli, providing numerous opportunities for afferent and efferent projections to integrate with cognitive networks to refine auditory perception with respect to behavioral state. We argue that auditory perception is best understood as an emergent property of the dynamic interaction between auditory processing centers and the distributed networks that support attention, prediction, memory, and action selection.

Several open questions remain. First, how are predictive models implemented and coordinated across brain regions? While the cerebellum, hippocampus, superior colliculus and frontal circuits each contribute to auditory perception, it remains unclear how signals from these brain structures are integrated with the canonical auditory processing pathway across different timescales. Second, how do neuromodulatory systems gate the relative contribution of attention, prediction, and memory in different behavioral contexts? Third, how does the auditory system dynamically mediate the rapid transitions between top-down and bottom-up processing as animals switch between echolocation and communication calls during navigation in natural environments? Finally, how are egocentric (body-based) sensory representations transformed into allocentric or relational representations within auditory-cognitive circuits to guide navigation?

Addressing these questions will require linking circuit-level mechanisms to species-specific behaviors under natural conditions. Combining large-scale neural recordings across multiple regions, causal manipulations, and quantitative behavioral readouts will be essential to unravel the contributions of attention, prediction, memory and action selection to auditory processing mechanisms and perception. The echolocating bat, with its active sensing system and conserved mammalian neural architecture, offers a powerful research model to uncover general principles of auditory-cognitive processes. Their ability to rapidly switch between specialized modes for echolocation and social communication further provides a robust window into how neural circuits are dynamically reconfigured across behavioral states.

Highlights.

• Brain networks outside the canonical auditory pathway shape sound perception

• Attention, prediction, memory and action networks modulate auditory processing

• Echolocating bats provide a powerful model to study auditory-cognitive processes