Multiplicative Auditory Spatial Receptive Fields Created by a Hierarchy of Population Codes

A multiplicative combination of tuning to interaural time difference (ITD) and interaural level difference (ILD) contributes to the generation of spatially selective auditory neurons in the owl''s midbrain. Previous analyses of multiplicative responses in the owl have not taken into consideration the frequency-dependence of ITD and ILD cues that occur under natural listening conditions. Here, we present a model for the responses of ITD- and ILD-sensitive neurons in the barn owl''s inferior colliculus which satisfies constraints raised by experimental data on frequency convergence, multiplicative interaction of ITD and ILD, and response properties of afferent neurons. We propose that multiplication between ITD- and ILD-dependent signals occurs only within frequency channels and that frequency integration occurs using a linear-threshold mechanism. The model reproduces the experimentally observed nonlinear responses to ITD and ILD in the inferior colliculus, with greater accuracy than previous models. We show that linear-threshold frequency integration allows the system to represent multiple sound sources with natural sound localization cues, whereas multiplicative frequency integration does not. Nonlinear responses in the owl''s inferior colliculus can thus be generated using a combination of cellular and network mechanisms, showing that multiple elements of previous theories can be combined in a single system.     

Optimal Prediction of Moving Sound Source Direction in the Owl

Capturing nature’s statistical structure in behavioral responses is at the core of the ability to function adaptively in the environment. Bayesian statistical inference describes how sensory and prior information can be combined optimally to guide behavior. An outstanding open question of how neural coding supports Bayesian inference includes how sensory cues are optimally integrated over time. Here we address what neural response properties allow a neural system to perform Bayesian prediction, i.e., predicting where a source will be in the near future given sensory information and prior assumptions. The work here shows that the population vector decoder will perform Bayesian prediction when the receptive fields of the neurons encode the target dynamics with shifting receptive fields. We test the model using the system that underlies sound localization in barn owls. Neurons in the owl’s midbrain show shifting receptive fields for moving sources that are consistent with the predictions of the model. We predict that neural populations can be specialized to represent the statistics of dynamic stimuli to allow for a vector read-out of Bayes-optimal predictions.     

Maps of interaural time difference in the chicken’s brainstem nucleus laminaris

Animals, including humans, use interaural time differences (ITDs) that arise from different sound path lengths to the two ears as a cue of horizontal sound source location. The nature of the neural code for ITD is still controversial. Current models differentiate between two population codes: either a map-like rate-place code of ITD along an array of neurons, consistent with a large body of data in the barn owl, or a population rate code, consistent with data from small mammals. Recently, it was proposed that these different codes reflect optimal coding strategies that depend on head size and sound frequency. The chicken makes an excellent test case of this proposal because its physical prerequisites are similar to small mammals, yet it shares a more recent common ancestry with the owl. We show here that, like in the barn owl, the brainstem nucleus laminaris in mature chickens displayed the major features of a place code of ITD. ITD was topographically represented in the maximal responses of neurons along each isofrequency band, covering approximately the contralateral acoustic hemisphere. Furthermore, the represented ITD range appeared to change with frequency, consistent with a pressure gradient receiver mechanism in the avian middle ear. At very low frequencies, below 400 Hz, maximal neural responses were symmetrically distributed around zero ITD and it remained unclear whether there was a topographic representation. These findings do not agree with the above predictions for optimal coding and thus revive the discussion as to what determines the neural coding strategies for ITDs.     

Precognitive and cognitive elements in sound localization

Sound localization behavior is of great importance for an animal's survival. To localize a sound, animals have to detect a sound source and assign a location to it. In this review we discuss recent results on the underlying mechanisms and on modulatory influences in the barn owl, an auditory specialist with very well developed capabilities to localize sound. Information processing in the barn owl auditory pathway underlying the computations of detection and localization is well understood. This analysis of the sensory information primarily determines the following orienting behavior towards the sound source. However, orienting behavior may be modulated by cognitive (top-down) influences such as attention. We show how advanced stimulation techniques can be used to determine the importance of different cues for sound localization in quasi-realistic stimulation situations, how attentional influences can improve the response to behaviorally relevant stimuli, and how attention can modulate related neural responses. Taken together, these data indicate how sound localization might function in the usually complex natural environment.     

Responses to simulated echoes by neurons in the barn owl's auditory space map

The natural acoustical environment contains many reflective surfaces that give rise to echoes, complicating the task of sound localization and identification. The barn owl (Tyto alba), as a nocturnal predator, relies heavily on its auditory system for tracking and capturing prey in this highly echoic environment. The external nucleus of the owl's inferior colliculus (ICx) contains a retina-like map of space composed of space-specific auditory neurons that have spatially limited receptive fields. We recorded extracellularly from individual space-specific neurons in an attempt to understand the pattern of activity across the ICx in response to a brief direct sound and a simulated echo. Space-specific neurons responded strongly to the direct sound, but their response to a simulated echo was suppressed, typically, if the echo arrived within 5 ms or less of the direct sound. Thus we expect there to be little or no representation within the ICx of echoes arriving within such short delays.Behavioral tests using the owl's natural tendency to turn their head toward a sound source suggested that owls, like their space-specific neurons, similarly localize only the first of two brief sounds. Naive, untrained owls were presented with a pair of sounds in rapid succession from two horizontally-separated speakers. With interstimulus delays of less than 10 ms, the owl consistently turned its head toward the leading speaker. Longer delays elicited head turns to either speaker with approximately equal frequency and in some cases to both speakers sequentially.Abbreviations IC inferior colliculus - ICx external nucleus of the inferior colliculus - ITD interaural time difference - ISI interstimulus interval - LS left speaker - RS right speaker - CS centering speaker - RF receptive field     

A Dominance Hierarchy of Auditory Spatial Cues in Barn Owls

Background

Barn owls integrate spatial information across frequency channels to localize sounds in space.

Methodology/Principal Findings

We presented barn owls with synchronous sounds that contained different bands of frequencies (3–5 kHz and 7–9 kHz) from different locations in space. When the owls were confronted with the conflicting localization cues from two synchronous sounds of equal level, their orienting responses were dominated by one of the sounds: they oriented toward the location of the low frequency sound when the sources were separated in azimuth; in contrast, they oriented toward the location of the high frequency sound when the sources were separated in elevation. We identified neural correlates of this behavioral effect in the optic tectum (OT, superior colliculus in mammals), which contains a map of auditory space and is involved in generating orienting movements to sounds. We found that low frequency cues dominate the representation of sound azimuth in the OT space map, whereas high frequency cues dominate the representation of sound elevation.

Conclusions/Significance

We argue that the dominance hierarchy of localization cues reflects several factors: 1) the relative amplitude of the sound providing the cue, 2) the resolution with which the auditory system measures the value of a cue, and 3) the spatial ambiguity in interpreting the cue. These same factors may contribute to the relative weighting of sound localization cues in other species, including humans.     

Sound-localization experiments with barn owls in virtual space: influence of broadband interaural level difference on head-turning behavior

Interaural level differences play an important role for elevational sound localization in barn owls. The changes of this cue with sound location are complex and frequency dependent. We exploited the opportunities offered by the virtual space technique to investigate the behavioral relevance of the overall interaural level difference by fixing this parameter in virtual stimuli to a constant value or introducing additional broadband level differences to normal virtual stimuli. Frequency-specific monaural cues in the stimuli were not manipulated. We observed an influence of the broadband interaural level differences on elevational, but not on azimuthal sound localization. Since results obtained with our manipulations explained only part of the variance in elevational turning angle, we conclude that frequency-specific cues are also important. The behavioral consequences of changes of the overall interaural level difference in a virtual sound depended on the combined interaural time difference contained in the stimulus, indicating an indirect influence of temporal cues on elevational sound localization as well. Thus, elevational sound localization is influenced by a combination of many spatial cues including frequency-dependent and temporal features.     

Mechanisms of experience-dependent plasticity in the auditory localization pathway of the barn owl

Sound localization is a computational process that requires the central nervous system to measure various auditory cues and then associate particular cue values with appropriate locations in space. Behavioral experiments show that barn owls learn to associate values of cues with locations in space based on experience. The capacity for experience-driven changes in sound localization behavior is particularly great during a sensitive period that lasts until the approach of adulthood. Neurophysiological techniques have been used to determine underlying sites of plasticity in the auditory space-processing pathway. The external nucleus of the inferior colliculus (ICX), where a map of auditory space is synthesized, is a major site of plasticity. Experience during the sensitive period can cause large-scale, adaptive changes in the tuning of ICX neurons for sound localization cues. Large-scale physiological changes are accompanied by anatomical remodeling of afferent axons to the ICX. Changes in the tuning of ICX neurons for cue values involve two stages: (1) the instructed acquisition of neuronal responses to novel cue values and (2) the elimination of responses to inappropriate cue values. Newly acquired neuronal responses depend differentially on NMDA receptor currents for their expression. A model is presented that can account for this adaptive plasticity in terms of plausible cellular mechanisms. Accepted: 17 April 1999     

Diverse processing underlying frequency integration in midbrain neurons of barn owls

Emergent response properties of sensory neurons depend on circuit connectivity and somatodendritic processing. Neurons of the barn owl’s external nucleus of the inferior colliculus (ICx) display emergence of spatial selectivity. These neurons use interaural time difference (ITD) as a cue for the horizontal direction of sound sources. ITD is detected by upstream brainstem neurons with narrow frequency tuning, resulting in spatially ambiguous responses. This spatial ambiguity is resolved by ICx neurons integrating inputs over frequency, a relevant processing in sound localization across species. Previous models have predicted that ICx neurons function as point neurons that linearly integrate inputs across frequency. However, the complex dendritic trees and spines of ICx neurons raises the question of whether this prediction is accurate. Data from in vivo intracellular recordings of ICx neurons were used to address this question. Results revealed diverse frequency integration properties, where some ICx neurons showed responses consistent with the point neuron hypothesis and others with nonlinear dendritic integration. Modeling showed that varied connectivity patterns and forms of dendritic processing may underlie observed ICx neurons’ frequency integration processing. These results corroborate the ability of neurons with complex dendritic trees to implement diverse linear and nonlinear integration of synaptic inputs, of relevance for adaptive coding and learning, and supporting a fundamental mechanism in sound localization.     

Sensitivity to spectral interaural intensity difference cues in space-specific neurons of the barn owl

Barn owls use interaural intensity differences to localize sounds in the vertical plane. At a given elevation the magnitude of the interaural intensity difference cue varies with frequency, creating an interaural intensity difference spectrum of cues which is characteristic of that direction. To test whether space-specific cells are sensitive to spectral interaural intensity difference cues, pure-tone interaural intensity difference tuning curves were taken at multiple different frequencies for single neurons in the external nucleus of the inferior colliculus. For a given neuron, the interaural intensity differences eliciting the maximum response (the best interaural intensity differences) changed with the frequency of the stimulus by an average maximal difference of 9.4±6.2 dB. The resulting spectral patterns of these neurally preferred interaural intensity differences exhibited a high degree of similarity to the acoustic interaural intensity difference spectra characteristic of restricted regions in space. Compared to stimuli whose interaural intensity difference spectra matched the preferred spectra, stimuli with inverted spectra elicited a smaller response, showing that space-specific neurons are sensitive to the shape of the spectrum. The underlying mechanism is an inhibition for frequency-specific interaural intensity differences which differ from the preferred spectral pattern. Collectively, these data show that space-specific neurons are sensitive to spectral interaural intensity difference cues and support the idea that behaving barn owls use such cues to precisely localize sounds.Abbreviations ABI average binaural intensity - HRTF head-related transfer function - ICx external nucleus of the inferior colliculus - IID interaural intensity difference - ITD interaural time difference - OT optic tectum - RMS root mean square - VLVp nucleus ventralis lemnisci laterale, pars posterior     

The implications of categorical and category-free mixed selectivity on representational geometries

The firing rates of individual neurons displaying mixed selectivity are modulated by multiple task variables. When mixed selectivity is nonlinear, it confers an advantage by generating a high-dimensional neural representation that can be flexibly decoded by linear classifiers. Although the advantages of this coding scheme are well accepted, the means of designing an experiment and analyzing the data to test for and characterize mixed selectivity remain unclear. With the growing number of large datasets collected during complex tasks, the mixed selectivity is increasingly observed and is challenging to interpret correctly. We review recent approaches for analyzing and interpreting neural datasets and clarify the theoretical implications of mixed selectivity in the variety of forms that have been reported in the literature. We also aim to provide a practical guide for determining whether a neural population has linear or nonlinear mixed selectivity and whether this mixing leads to a categorical or category-free representation.     

Behavioral guides for sensory neurophysiology

The study of natural behavior is important for understanding the coding schemes of sensory systems. The jamming avoidance response of the weakly electric fish Eigenmannia is an excellent example of a bottom–up approach, in which behavioral analyses guided neurophysiological studies. These studies started from the electroreceptive sense organs to the motor output consisting of pacemaker neurons. Going in the opposite direction, from the central nervous system to lower centers, is the characteristic of the top–down approach. Although this approach is perhaps more difficult than the bottom–up approach, it was successfully employed in the neuroethological analysis of sound localization in the barn owl. In the latter studies, high-order neurons selective for complex natural stimuli led to the discovery of neural pathways and networks responsible for the genesis of the stimulus selectivity. Comparison of Eigenmannia and barn owls, and their neural systems, has revealed similarities in network designs, such as parallel pathways and their convergence to produce stimulus selectivity necessary for detection of natural stimuli.     

Sensory cues for sound localization in the cricket<Emphasis Type="Italic"> Teleogryllus oceanicus</Emphasis>: interaural difference in response strength versus interaural latency difference

Two potential sensory cues for sound location are interaural difference in response strength (firing rate and/or spike count) and in response latency of auditory receptor neurons. Previous experiments showed that these two cues are affected differently by intense prior stimulation; the difference in response strength declines and may even reverse in sign, but the difference in latency is unaffected. Here, I use an intense, constant tone to disrupt localization cues generated by a subsequent train of sound pulses. Recordings from the auditory nerve confirm that tone stimulation reduces, and sometimes reverses, the interaural difference in response strength to subsequent sound pulses, but that it enhances the interaural latency difference. If sound location is determined mainly from latency comparison, then behavioral responses to a pulse train following tone stimulation should be normal, but if the main cue for sound location is interaural difference in response strength, then post-tone behavioral responses should sometimes be misdirected. Initial phonotactic responses to the post-tone pulse train were frequently directed away from, rather than towards, the sound source, indicating that the dominant sensory cue for sound location is interaural difference in response strength.     

Multisensory integration in macaque visual cortex depends on cue reliability

Responses of multisensory neurons to combinations of sensory cues are generally enhanced or depressed relative to single cues presented alone, but the rules that govern these interactions have remained unclear. We examined integration of visual and vestibular self-motion cues in macaque area MSTd in response to unimodal as well as congruent and conflicting bimodal stimuli in order to evaluate hypothetical combination rules employed by multisensory neurons. Bimodal responses were well fit by weighted linear sums of unimodal responses, with weights typically less than one (subadditive). Surprisingly, our results indicate that weights change with the relative reliabilities of the two cues: visual weights decrease and vestibular weights increase when visual stimuli are degraded. Moreover, both modulation depth and neuronal discrimination thresholds improve for matched bimodal compared to unimodal stimuli, which might allow for increased neural sensitivity during multisensory stimulation. These findings establish important new constraints for neural models of cue integration.     

Causal inference in multisensory perception

Perceptual events derive their significance to an animal from their meaning about the world, that is from the information they carry about their causes. The brain should thus be able to efficiently infer the causes underlying our sensory events. Here we use multisensory cue combination to study causal inference in perception. We formulate an ideal-observer model that infers whether two sensory cues originate from the same location and that also estimates their location(s). This model accurately predicts the nonlinear integration of cues by human subjects in two auditory-visual localization tasks. The results show that indeed humans can efficiently infer the causal structure as well as the location of causes. By combining insights from the study of causal inference with the ideal-observer approach to sensory cue combination, we show that the capacity to infer causal structure is not limited to conscious, high-level cognition; it is also performed continually and effortlessly in perception.     

An ideal-observer model of human sound localization

In recent years, a great deal of research within the field of sound localization has been aimed at finding the acoustic cues that human listeners use to localize sounds and understanding the mechanisms by which they process these cues. In this paper, we propose a complementary approach by constructing an ideal-observer model, by which we mean a model that performs optimal information processing within a Bayesian context. The model considers all available spatial information contained within the acoustic signals encoded by each ear. Parameters for the optimal Bayesian model are determined based on psychoacoustic discrimination experiments on interaural time difference and sound intensity. Without regard as to how the human auditory system actually processes information, we examine the best possible localization performance that could be achieved based only on analysis of the input information, given the constraints of the normal auditory system. We show that the model performance is generally in good agreement with the actual human localization performance, as assessed in a meta-analysis of many localization experiments (Best et al. in Principles and applications of spatial hearing, pp 14–23. World Scientific Publishing, Singapore, 2011). We believe this approach can shed new light on the optimality (or otherwise) of human sound localization, especially with regard to the level of uncertainty in the input information. Moreover, the proposed model allows one to study the relative importance of various (combinations of) acoustic cues for spatial localization and enables a prediction of which cues are most informative and therefore likely to be used by humans in various circumstances.     

Mutation in the Kv3.3 Voltage-Gated Potassium Channel Causing Spinocerebellar Ataxia 13 Disrupts Sound-Localization Mechanisms

Normal sound localization requires precise comparisons of sound timing and pressure levels between the two ears. The primary localization cues are interaural time differences, ITD, and interaural level differences, ILD. Voltage-gated potassium channels, including Kv3.3, are highly expressed in the auditory brainstem and are thought to underlie the exquisite temporal precision and rapid spike rates that characterize brainstem binaural pathways. An autosomal dominant mutation in the gene encoding Kv3.3 has been demonstrated in a large Filipino kindred manifesting as spinocerebellar ataxia type 13 (SCA13). This kindred provides a rare opportunity to test in vivo the importance of a specific channel subunit for human hearing. Here, we demonstrate psychophysically that individuals with the mutant allele exhibit profound deficits in both ITD and ILD sensitivity, despite showing no obvious impairment in pure-tone sensitivity with either ear. Surprisingly, several individuals exhibited the auditory deficits even though they were pre-symptomatic for SCA13. We would expect that impairments of binaural processing as great as those observed in this family would result in prominent deficits in localization of sound sources and in loss of the "spatial release from masking" that aids in understanding speech in the presence of competing sounds.     

A comparison of neural computations underlying stereo vision and sound localization.

Stereo vision and sound localization are behaviors that help organisms to orient in space. Stereo vision provides information about the distance of an object. In sound localization mainly directional signals are analyzed. Two specific cues, binocular disparities underlying stereo vision, and the interaural time difference, suited to represent azimuth in sound localization, have many computational problems in common, although they are generated in two different modalities. The extraction of both cues requires a comparison of signals that arise from two independent sensors, called two-sensor comparison here. This two-sensor comparison is achieved by algorithms similar to summation, half-wave rectification and multiplication. Since the underlying neurons are band-pass filters, the two-sensor comparison results in ambiguities. These are removed in a hierarchical way in several computational steps, involving squaring, inhibition and across-frequency integration.     

Probabilistic inference in general graphical models through sampling in stochastic networks of spiking neurons

An important open problem of computational neuroscience is the generic organization of computations in networks of neurons in the brain. We show here through rigorous theoretical analysis that inherent stochastic features of spiking neurons, in combination with simple nonlinear computational operations in specific network motifs and dendritic arbors, enable networks of spiking neurons to carry out probabilistic inference through sampling in general graphical models. In particular, it enables them to carry out probabilistic inference in Bayesian networks with converging arrows ("explaining away") and with undirected loops, that occur in many real-world tasks. Ubiquitous stochastic features of networks of spiking neurons, such as trial-to-trial variability and spontaneous activity, are necessary ingredients of the underlying computational organization. We demonstrate through computer simulations that this approach can be scaled up to neural emulations of probabilistic inference in fairly large graphical models, yielding some of the most complex computations that have been carried out so far in networks of spiking neurons.     

The synthesis and use of the owl’s auditory space map

The barn owl (Tyto alba) is capable of capturing prey by passive hearing alone, guided by a topographic map of auditory space in the external nucleus of its inferior colliculus. The neurons of this auditory space map have discrete spatial receptive fields that result from the computation of interaural differences in the level (ILD) and time-of-arrival (ITD) of sounds. Below we review the synthesis of the spatial receptive fields from the frequency-specific ITDs and ILDs to which the neurons are tuned, concentrating on recent studies exploiting virtual auditory space techniques to analyze the contribution of ILD. We then compared the owls spatial discrimination, assessed behaviorally, with that of its space map neurons. Spatial discrimination was assessed using a novel paradigm involving the pupillary dilation response (PDR), and neuronal acuity was assessed by measuring the changes in firing rate resulting from changes in source location, scaled to the variance. This signal-detection-based approach revealed that the change in the position of the neural image on this map best explains the spatial discrimination measured using the PDR. We compare this result to recent studies in mammalian systems.