"Who" Is Saying "What"? Brain-Based Decoding of Human Voice and Speech

Auditory and visual modulation of temporal lobe neurons in voice-sensitive and association cortices

Effective interactions between conspecific individuals can depend upon the receiver forming a coherent multisensory representation of communication signals, such as merging voice and face content. Neuroimaging studies have identified face- or voice-sensitive areas (Belin et al., 2000; Petkov et al., 2008; Tsao et al., 2008), some of which have been proposed as candidate regions for face and voice integration (von Kriegstein et al., 2005). However, it was unclear how multisensory influences occur at the neuronal level within voice- or face-sensitive regions, especially compared with classically defined multisensory regions in temporal association cortex (Stein and Stanford, 2008). Here, we characterize auditory (voice) and visual (face) influences on neuronal responses in a right-hemisphere voice-sensitive region in the anterior supratemporal plane (STP) of Rhesus macaques. These results were compared with those in the neighboring superior temporal sulcus (STS). Within the STP, our results show auditory sensitivity to several vocal features, which was not evident in STS units. We also newly identify a functionally distinct neuronal subpopulation in the STP that appears to carry the area's sensitivity to voice identity related features. Audiovisual interactions were prominent in both the STP and STS. However, visual influences modulated the responses of STS neurons with greater specificity and were more often associated with congruent voice-face stimulus pairings than STP neurons. Together, the results reveal the neuronal processes subserving voice-sensitive fMRI activity patterns in primates, generate hypotheses for testing in the visual modality, and clarify the position of voice-sensitive areas within the unisensory and multisensory processing hierarchies.

Characterizing the dynamics of mental representations: the temporal generalization method

Parsing a cognitive task into a sequence of operations is a central problem in cognitive neuroscience. We argue that a major advance is now possible owing to the application of pattern classifiers to time-resolved recordings of brain activity [electroencephalography (EEG), magnetoencephalography (MEG), or intracranial recordings]. By testing at which moment a specific mental content becomes decodable in brain activity, we can characterize the time course of cognitive codes. Most importantly, the manner in which the trained classifiers generalize across time, and from one experimental condition to another, sheds light on the temporal organization of information-processing stages. A repertoire of canonical dynamical patterns is observed across various experiments and brain regions. This method thus provides a novel way to understand how mental representations are manipulated and transformed.

PRoNTo: pattern recognition for neuroimaging toolbox

In the past years, mass univariate statistical analyses of neuroimaging data have been complemented by the use of multivariate pattern analyses, especially based on machine learning models. While these allow an increased sensitivity for the detection of spatially distributed effects compared to univariate techniques, they lack an established and accessible software framework. The goal of this work was to build a toolbox comprising all the necessary functionalities for multivariate analyses of neuroimaging data, based on machine learning models. The “Pattern Recognition for Neuroimaging Toolbox” (PRoNTo) is open-source, cross-platform, MATLAB-based and SPM compatible, therefore being suitable for both cognitive and clinical neuroscience research. In addition, it is designed to facilitate novel contributions from developers, aiming to improve the interaction between the neuroimaging and machine learning communities. Here, we introduce PRoNTo by presenting examples of possible research questions that can be addressed with the machine learning framework implemented in PRoNTo, and cannot be easily investigated with mass univariate statistical analysis.

Robust discrimination between EEG responses to categories of environmental sounds in early coma

Humans can recognize categories of environmental sounds, including vocalizations produced by humans and animals and the sounds of man-made objects. Most neuroimaging investigations of environmental sound discrimination have studied subjects while consciously perceiving and often explicitly recognizing the stimuli. Consequently, it remains unclear to what extent auditory object processing occurs independently of task demands and consciousness. Studies in animal models have shown that environmental sound discrimination at a neural level persists even in anesthetized preparations, whereas data from anesthetized humans has thus far provided null results. Here, we studied comatose patients as a model of environmental sound discrimination capacities during unconsciousness. We included 19 comatose patients treated with therapeutic hypothermia (TH) during the first 2 days of coma, while recording nineteen-channel electroencephalography (EEG). At the level of each individual patient, we applied a decoding algorithm to quantify the differential EEG responses to human vs. animal vocalizations as well as to sounds of living vocalizations vs. man-made objects. Discrimination between vocalization types was accurate in 11 patients and discrimination between sounds from living and man-made sources in 10 patients. At the group level, the results were significant only for the comparison between vocalization types. These results lay the groundwork for disentangling truly preferential activations in response to auditory categories, and the contribution of awareness to auditory category discrimination.

Brain-based translation: fMRI decoding of spoken words in bilinguals reveals language-independent semantic representations in anterior temporal lobe

Bilinguals derive the same semantic concepts from equivalent, but acoustically different, words in their first and second languages. The neural mechanisms underlying the representation of language-independent concepts in the brain remain unclear. Here, we measured fMRI in human bilingual listeners and reveal that response patterns to individual spoken nouns in one language (e.g., "horse" in English) accurately predict the response patterns to equivalent nouns in the other language (e.g., "paard" in Dutch). Stimuli were four monosyllabic words in both languages, all from the category of "animal" nouns. For each word, pronunciations from three different speakers were included, allowing the investigation of speaker-independent representations of individual words. We used multivariate classifiers and a searchlight method to map the informative fMRI response patterns that enable decoding spoken words within languages (within-language discrimination) and across languages (across-language generalization). Response patterns discriminative of spoken words within language were distributed in multiple cortical regions, reflecting the complexity of the neural networks recruited during speech and language processing. Response patterns discriminative of spoken words across language were limited to localized clusters in the left anterior temporal lobe, the left angular gyrus and the posterior bank of the left postcentral gyrus, the right posterior superior temporal sulcus/superior temporal gyrus, the right medial anterior temporal lobe, the right anterior insula, and bilateral occipital cortex. These results corroborate the existence of "hub" regions organizing semantic-conceptual knowledge in abstract form at the fine-grained level of within semantic category discriminations.

Data-driven analysis of functional brain interactions during free listening to music and speech

Natural stimulus functional magnetic resonance imaging (N-fMRI) such as fMRI acquired when participants were watching video streams or listening to audio streams has been increasingly used to investigate functional mechanisms of the human brain in recent years. One of the fundamental challenges in functional brain mapping based on N-fMRI is to model the brain's functional responses to continuous, naturalistic and dynamic natural stimuli. To address this challenge, in this paper we present a data-driven approach to exploring functional interactions in the human brain during free listening to music and speech streams. Specifically, we model the brain responses using N-fMRI by measuring the functional interactions on large-scale brain networks with intrinsically established structural correspondence, and perform music and speech classification tasks to guide the systematic identification of consistent and discriminative functional interactions when multiple subjects were listening music and speech in multiple categories. The underlying premise is that the functional interactions derived from N-fMRI data of multiple subjects should exhibit both consistency and discriminability. Our experimental results show that a variety of brain systems including attention, memory, auditory/language, emotion, and action networks are among the most relevant brain systems involved in classic music, pop music and speech differentiation. Our study provides an alternative approach to investigating the human brain's mechanism in comprehension of complex natural music and speech.

Multiclass fMRI data decoding and visualization using supervised self-organizing maps

When multivariate pattern decoding is applied to fMRI studies entailing more than two experimental conditions, a most common approach is to transform the multiclass classification problem into a series of binary problems. Furthermore, for decoding analyses, classification accuracy is often the only outcome reported although the topology of activation patterns in the high-dimensional features space may provide additional insights into underlying brain representations. Here we propose to decode and visualize voxel patterns of fMRI datasets consisting of multiple conditions with a supervised variant of self-organizing maps (SSOMs). Using simulations and real fMRI data, we evaluated the performance of our SSOM-based approach. Specifically, the analysis of simulated fMRI data with varying signal-to-noise and contrast-to-noise ratio suggested that SSOMs perform better than a k-nearest-neighbor classifier for medium and large numbers of features (i.e. 250 to 1000 or more voxels) and similar to support vector machines (SVMs) for small and medium numbers of features (i.e. 100 to 600voxels). However, for a larger number of features (>800voxels), SSOMs performed worse than SVMs. When applied to a challenging 3-class fMRI classification problem with datasets collected to examine the neural representation of three human voices at individual speaker level, the SSOM-based algorithm was able to decode speaker identity from auditory cortical activation patterns. Classification performances were similar between SSOMs and other decoding algorithms; however, the ability to visualize decoding models and underlying data topology of SSOMs promotes a more comprehensive understanding of classification outcomes. We further illustrated this visualization ability of SSOMs with a re-analysis of a dataset examining the representation of visual categories in the ventral visual cortex (Haxby et al., 2001). This analysis showed that SSOMs could retrieve and visualize topography and neighborhood relations of the brain representation of eight visual categories. We conclude that SSOMs are particularly suited for decoding datasets consisting of more than two classes and are optimally combined with approaches that reduce the number of voxels used for classification (e.g. region-of-interest or searchlight approaches).

Voice Identity Recognition: Functional Division of the Right STS and Its Behavioral Relevance

The human voice is the primary carrier of speech but also a fingerprint for person identity. Previous neuroimaging studies have revealed that speech and identity recognition is accomplished by partially different neural pathways, despite the perceptual unity of the vocal sound. Importantly, the right STS has been implicated in voice processing, with different contributions of its posterior and anterior parts. However, the time point at which vocal and speech processing diverge is currently unknown. Also, the exact role of the right STS during voice processing is so far unclear because its behavioral relevance has not yet been established. Here, we used the high temporal resolution of magnetoencephalography and a speech task control to pinpoint transient behavioral correlates: we found, at 200 msec after stimulus onset, that activity in right anterior STS predicted behavioral voice recognition performance. At the same time point, the posterior right STS showed increased activity during voice identity recognition in contrast to speech recognition whereas the left mid STS showed the reverse pattern. In contrast to the highly speech-sensitive left STS, the current results highlight the right STS as a key area for voice identity recognition and show that its anatomical-functional division emerges around 200 msec after stimulus onset. We suggest that this time point marks the speech-independent processing of vocal sounds in the posterior STS and their successful mapping to vocal identities in the anterior STS.

Sparseness of vowel category structure: Evidence from English dialect comparison

Current models of speech perception tend to emphasize either fine-grained acoustic properties or coarse-grained abstract characteristics of speech sounds. We argue for a particular kind of 'sparse' vowel representations and provide new evidence that these representations account for the successful access of the corresponding categories. In an auditory semantic priming experiment, American English listeners made lexical decisions on targets (e.g. load) preceded by semantically related primes (e.g. pack). Changes of the prime vowel that crossed a vowel-category boundary (e.g. peck) were not treated as a tolerable variation, as assessed by a lack of priming, although the phonetic categories of the two different vowels considerably overlap in American English. Compared to the outcome of the same experiment with New Zealand English listeners, where such prime variations were tolerated, our experiment supports the view that phonological representations are important in guiding the mapping process from the acoustic signal to an abstract mental representation. Our findings are discussed with regard to current models of speech perception and recent findings from brain imaging research.

Representations of Invariant Musical Categories Are Decodable by Pattern Analysis of Locally Distributed BOLD Responses in Superior Temporal and Intraparietal Sulci

In categorical perception (CP), continuous physical signals are mapped to discrete perceptual bins: mental categories not found in the physical world. CP has been demonstrated across multiple sensory modalities and, in audition, for certain over-learned speech and musical sounds. The neural basis of auditory CP, however, remains ambiguous, including its robustness in nonspeech processes and the relative roles of left/right hemispheres; primary/nonprimary cortices; and ventral/dorsal perceptual processing streams. Here, highly trained musicians listened to 2-tone musical intervals, which they perceive categorically while undergoing functional magnetic resonance imaging. Multivariate pattern analyses were performed after grouping sounds by interval quality (determined by frequency ratio between tones) or pitch height (perceived noncategorically, frequency ratios remain constant). Distributed activity patterns in spheres of voxels were used to determine sound sample identities. For intervals, significant decoding accuracy was observed in the right superior temporal and left intraparietal sulci, with smaller peaks observed homologously in contralateral hemispheres. For pitch height, no significant decoding accuracy was observed, consistent with the non-CP of this dimension. These results suggest that similar mechanisms are operative for nonspeech categories as for speech; espouse roles for 2 segregated processing streams; and support hierarchical processing models for CP.

Phonetic feature encoding in human superior temporal gyrus

During speech perception, linguistic elements such as consonants and vowels are extracted from a complex acoustic speech signal. The superior temporal gyrus (STG) participates in high-order auditory processing of speech, but how it encodes phonetic information is poorly understood. We used high-density direct cortical surface recordings in humans while they listened to natural, continuous speech to reveal the STG representation of the entire English phonetic inventory. At single electrodes, we found response selectivity to distinct phonetic features. Encoding of acoustic properties was mediated by a distributed population response. Phonetic features could be directly related to tuning for spectrotemporal acoustic cues, some of which were encoded in a nonlinear fashion or by integration of multiple cues. These findings demonstrate the acoustic-phonetic representation of speech in human STG.

Advancing emotion theory with multivariate pattern classification

Characterizing how activity in the central and autonomic nervous systems corresponds to distinct emotional states is one of the central goals of affective neuroscience. Despite the ease with which individuals label their own experiences, identifying specific autonomic and neural markers of emotions remains a challenge. Here we explore how multivariate pattern classification approaches offer an advantageous framework for identifying emotion specific biomarkers and for testing predictions of theoretical models of emotion. Based on initial studies using multivariate pattern classification, we suggest that central and autonomic nervous system activity can be reliably decoded into distinct emotional states. Finally, we consider future directions in applying pattern classification to understand the nature of emotion in the nervous system.

An Investigation of Place and Voice Features Using fMRI-Adaptation

A widely accepted view of speech perception holds that in order to comprehend language, the variable acoustic signal must be parsed into a set of abstract linguistic representations. However, the neural basis of early phonological processing, including the nature of featural encoding of speech, is still poorly understood. In part, progress in this domain has been constrained by the difficulty inherent in extricating the influence of acoustic modulations from those which can be ascribed to the abstract, featural content of the stimuli. A further concern is that group averaging techniques may obscure subtle individual differences in cortical regions involved in early language processing. In this paper we present the results of an fMRI-adaptation experiment which finds evidence of areas in the superior and medial temporal lobes which respond selectively to changes in the major feature categories of voicing and place of articulation. We present both single-subject and group-averaged analyses.

The what, where and how of auditory-object perception

The fundamental perceptual unit in hearing is the 'auditory object'. Similar to visual objects, auditory objects are the computational result of the auditory system's capacity to detect, extract, segregate and group spectrotemporal regularities in the acoustic environment; the multitude of acoustic stimuli around us together form the auditory scene. However, unlike the visual scene, resolving the component objects within the auditory scene crucially depends on their temporal structure. Neural correlates of auditory objects are found throughout the auditory system. However, neural responses do not become correlated with a listener's perceptual reports until the level of the cortex. The roles of different neural structures and the contribution of different cognitive states to the perception of auditory objects are not yet fully understood.

Intact but less accessible phonetic representations in adults with dyslexia

Dyslexia is a severe and persistent reading and spelling disorder caused by impairment in the ability to manipulate speech sounds. We combined functional magnetic resonance brain imaging with multivoxel pattern analysis and functional and structural connectivity analysis in an effort to disentangle whether dyslexics' phonological deficits are caused by poor quality of the phonetic representations or by difficulties in accessing intact phonetic representations. We found that phonetic representations are hosted bilaterally in primary and secondary auditory cortices and that their neural quality (in terms of robustness and distinctness) is intact in adults with dyslexia. However, the functional and structural connectivity between the bilateral auditory cortices and the left inferior frontal gyrus (a region involved in higher-level phonological processing) is significantly hampered in dyslexics, suggesting deficient access to otherwise intact phonetic representations.

Intact but less accessible phonetic representations in adults with dyslexia

Dyslexia is a severe and persistent reading and spelling disorder caused by impairment in the ability to manipulate speech sounds. We combined functional magnetic resonance brain imaging with multivoxel pattern analysis and functional and structural connectivity analysis in an effort to disentangle whether dyslexics' phonological deficits are caused by poor quality of the phonetic representations or by difficulties in accessing intact phonetic representations. We found that phonetic representations are hosted bilaterally in primary and secondary auditory cortices and that their neural quality (in terms of robustness and distinctness) is intact in adults with dyslexia. However, the functional and structural connectivity between the bilateral auditory cortices and the left inferior frontal gyrus (a region involved in higher-level phonological processing) is significantly hampered in dyslexics, suggesting deficient access to otherwise intact phonetic representations.

Representations of modality-specific affective processing for visual and auditory stimuli derived from functional magnetic resonance imaging data

There is converging evidence that people rapidly and automatically encode affective dimensions of objects, events, and environments that they encounter in the normal course of their daily routines. An important research question is whether affective representations differ with sensory modality. This research examined the nature of the dependency of affect and sensory modality at a whole-brain level of analysis in an incidental affective processing paradigm. Participants were presented with picture and sound stimuli that differed in positive or negative valence in an event-related functional magnetic resonance imaging experiment. Global statistical tests, applied at a level of the individual, demonstrated significant sensitivity to valence within modality, but not valence across modalities. Modality-general and modality-specific valence hypotheses predict distinctly different multidimensional patterns of the stimulus conditions. Examination of lower dimensional representation of the data demonstrated separable dimensions for valence processing within each modality. These results provide support for modality-specific valence processing in an incidental affective processing paradigm at a whole-brain level of analysis. Future research should further investigate how stimulus-specific emotional decoding may be mediated by the physical properties of the stimuli.

Neural and behavioral investigations into timbre perception

Timbre is the attribute that distinguishes sounds of equal pitch, loudness and duration. It contributes to our perception and discrimination of different vowels and consonants in speech, instruments in music and environmental sounds. Here we begin by reviewing human timbre perception and the spectral and temporal acoustic features that give rise to timbre in speech, musical and environmental sounds. We also consider the perception of timbre by animals, both in the case of human vowels and non-human vocalizations. We then explore the neural representation of timbre, first within the peripheral auditory system and later at the level of the auditory cortex. We examine the neural networks that are implicated in timbre perception and the computations that may be performed in auditory cortex to enable listeners to extract information about timbre. We consider whether single neurons in auditory cortex are capable of representing spectral timbre independently of changes in other perceptual attributes and the mechanisms that may shape neural sensitivity to timbre. Finally, we conclude by outlining some of the questions that remain about the role of neural mechanisms in behavior and consider some potentially fruitful avenues for future research.

Constructing noise-invariant representations of sound in the auditory pathway

Along the auditory pathway from auditory nerve to midbrain to cortex, individual neurons adapt progressively to sound statistics, enabling the discernment of foreground sounds, such as speech, over background noise.

Identifying behaviorally relevant sounds in the presence of background noise is one of the most important and poorly understood challenges faced by the auditory system. An elegant solution to this problem would be for the auditory system to represent sounds in a noise-invariant fashion. Since a major effect of background noise is to alter the statistics of the sounds reaching the ear, noise-invariant representations could be promoted by neurons adapting to stimulus statistics. Here we investigated the extent of neuronal adaptation to the mean and contrast of auditory stimulation as one ascends the auditory pathway. We measured these forms of adaptation by presenting complex synthetic and natural sounds, recording neuronal responses in the inferior colliculus and primary fields of the auditory cortex of anaesthetized ferrets, and comparing these responses with a sophisticated model of the auditory nerve. We find that the strength of both forms of adaptation increases as one ascends the auditory pathway. To investigate whether this adaptation to stimulus statistics contributes to the construction of noise-invariant sound representations, we also presented complex, natural sounds embedded in stationary noise, and used a decoding approach to assess the noise tolerance of the neuronal population code. We find that the code for complex sounds in the periphery is affected more by the addition of noise than the cortical code. We also find that noise tolerance is correlated with adaptation to stimulus statistics, so that populations that show the strongest adaptation to stimulus statistics are also the most noise-tolerant. This suggests that the increase in adaptation to sound statistics from auditory nerve to midbrain to cortex is an important stage in the construction of noise-invariant sound representations in the higher auditory brain.

We rarely hear sounds (such as someone talking) in isolation, but rather against a background of noise. When mixtures of sounds and background noise reach the ears, peripheral auditory neurons represent the whole sound mixture. Previous evidence suggests, however, that the higher auditory brain represents just the sounds of interest, and is less affected by the presence of background noise. The neural mechanisms underlying this transformation are poorly understood. Here, we investigate these mechanisms by studying the representation of sound by populations of neurons at three stages along the auditory pathway; we simulate the auditory nerve and record from neurons in the midbrain and primary auditory cortex of anesthetized ferrets. We find that the transformation from noise-sensitive representations of sound to noise-tolerant processing takes place gradually along the pathway from auditory nerve to midbrain to cortex. Our results suggest that this results from neurons adapting to the statistics of heard sounds.

Similarity of cortical activity patterns predicts generalization behavior

Humans and animals readily generalize previously learned knowledge to new situations. Determining similarity is critical for assigning category membership to a novel stimulus. We tested the hypothesis that category membership is initially encoded by the similarity of the activity pattern evoked by a novel stimulus to the patterns from known categories. We provide behavioral and neurophysiological evidence that activity patterns in primary auditory cortex contain sufficient information to explain behavioral categorization of novel speech sounds by rats. Our results suggest that category membership might be encoded by the similarity of the activity pattern evoked by a novel speech sound to the patterns evoked by known sounds. Categorization based on featureless pattern matching may represent a general neural mechanism for ensuring accurate generalization across sensory and cognitive systems.

Auditory neuroimaging with fMRI and PET

For much of the past 30 years, investigations of auditory perception and language have been enhanced or even driven by the use of functional neuroimaging techniques that specialize in localization of central responses. Beginning with investigations using positron emission tomography (PET) and gradually shifting primarily to usage of functional magnetic resonance imaging (fMRI), auditory neuroimaging has greatly advanced our understanding of the organization and response properties of brain regions critical to the perception of and communication with the acoustic world in which we live. As the complexity of the questions being addressed has increased, the techniques, experiments and analyses applied have also become more nuanced and specialized. A brief review of the history of these investigations sets the stage for an overview and analysis of how these neuroimaging modalities are becoming ever more effective tools for understanding the auditory brain. We conclude with a brief discussion of open methodological issues as well as potential clinical applications for auditory neuroimaging. This article is part of a Special Issue entitled Human Auditory Neuroimaging.

Crossmodal integration enhances neural representation of task-relevant features in audiovisual face perception

Previous studies have shown that audiovisual integration improves identification performance and enhances neural activity in heteromodal brain areas, for example, the posterior superior temporal sulcus/middle temporal gyrus (pSTS/MTG). Furthermore, it has also been demonstrated that attention plays an important role in crossmodal integration. In this study, we considered crossmodal integration in audiovisual facial perception and explored its effect on the neural representation of features. The audiovisual stimuli in the experiment consisted of facial movie clips that could be classified into 2 gender categories (male vs. female) or 2 emotion categories (crying vs. laughing). The visual/auditory-only stimuli were created from these movie clips by removing the auditory/visual contents. The subjects needed to make a judgment about the gender/emotion category for each movie clip in the audiovisual, visual-only, or auditory-only stimulus condition as functional magnetic resonance imaging (fMRI) signals were recorded. The neural representation of the gender/emotion feature was assessed using the decoding accuracy and the brain pattern-related reproducibility indices, obtained by a multivariate pattern analysis method from the fMRI data. In comparison to the visual-only and auditory-only stimulus conditions, we found that audiovisual integration enhanced the neural representation of task-relevant features and that feature-selective attention might play a role of modulation in the audiovisual integration.

Adaptation of the communicative brain to post-lingual deafness. Evidence from functional imaging

Not having access to one sense profoundly modifies our interactions with the environment, in turn producing changes in brain organization. Deafness and its rehabilitation by cochlear implantation offer a unique model of brain adaptation during sensory deprivation and recovery. Functional imaging allows the study of brain plasticity as a function of the times of deafness and implantation. Even long after the end of the sensitive period for auditory brain physiological maturation, some plasticity may be observed. In this way the mature brain that becomes deaf after language acquisition can adapt to its modified sensory inputs. Oral communication difficulties induced by post-lingual deafness shape cortical reorganization of brain networks already specialized for processing oral language. Left hemisphere language specialization tends to be more preserved than functions of the right hemisphere. We hypothesize that the right hemisphere offers cognitive resources re-purposed to palliate difficulties in left hemisphere speech processing due to sensory and auditory memory degradation. If cochlear implantation is considered, this reorganization during deafness may influence speech understanding outcomes positively or negatively. Understanding brain plasticity during post-lingual deafness should thus inform the development of cognitive rehabilitation, which promotes positive reorganization of the brain networks that process oral language before surgery. This article is part of a Special Issue entitled Human Auditory Neuroimaging.

Representation of frequency-modulated sounds in the human brain

Frequency-modulation is a ubiquitous sound feature present in communicative sounds of various animal species and humans. Functional imaging of the human auditory system has seen remarkable advances in the last two decades and studies pertaining to frequency-modulation have centered around two major questions: a) are there dedicated feature-detectors encoding frequency-modulation in the brain and b) is there concurrent representation with amplitude-modulation, another temporal sound feature? In this review, we first describe how these two questions are motivated by psychophysical studies and neurophysiology in animal models. We then review how human non-invasive neuroimaging studies have furthered our understanding of the representation of frequency-modulated sounds in the brain. Finally, we conclude with some suggestions on how human neuroimaging could be used in future studies to address currently still open questions on this fundamental sound feature. This article is part of a Special Issue entitled Human Auditory Neuroimaging.

Decoding speech perception by native and non-native speakers using single-trial electrophysiological data

Brain-computer interfaces (BCIs) are systems that use real-time analysis of neuroimaging data to determine the mental state of their user for purposes such as providing neurofeedback. Here, we investigate the feasibility of a BCI based on speech perception. Multivariate pattern classification methods were applied to single-trial EEG data collected during speech perception by native and non-native speakers. Two principal questions were asked: 1) Can differences in the perceived categories of pairs of phonemes be decoded at the single-trial level? 2) Can these same categorical differences be decoded across participants, within or between native-language groups? Results indicated that classification performance progressively increased with respect to the categorical status (within, boundary or across) of the stimulus contrast, and was also influenced by the native language of individual participants. Classifier performance showed strong relationships with traditional event-related potential measures and behavioral responses. The results of the cross-participant analysis indicated an overall increase in average classifier performance when trained on data from all participants (native and non-native). A second cross-participant classifier trained only on data from native speakers led to an overall improvement in performance for native speakers, but a reduction in performance for non-native speakers. We also found that the native language of a given participant could be decoded on the basis of EEG data with accuracy above 80%. These results indicate that electrophysiological responses underlying speech perception can be decoded at the single-trial level, and that decoding performance systematically reflects graded changes in the responses related to the phonological status of the stimuli. This approach could be used in extensions of the BCI paradigm to support perceptual learning during second language acquisition.

Cognitive-motor brain-machine interfaces

Brain-machine interfaces (BMIs) open new horizons for the treatment of paralyzed persons, giving hope for the artificial restoration of lost physiological functions. Whereas BMI development has mainly focused on motor rehabilitation, recent studies have suggested that higher cognitive functions can also be deciphered from brain activity, bypassing low level planning and execution functions, and replacing them by computer-controlled effectors. This review describes the new generation of cognitive-motor BMIs, focusing on three BMI types: By outlining recent progress in developing these BMI types, we aim to provide a unified view of contemporary research towards the replacement of behavioral outputs of cognitive processes by direct interaction with the brain.

Decoding the neural mechanisms of human tool use

Sophisticated tool use is a defining characteristic of the primate species but how is it supported by the brain, particularly the human brain? Here we show, using functional MRI and pattern classification methods, that tool use is subserved by multiple distributed action-centred neural representations that are both shared with and distinct from those of the hand. In areas of frontoparietal cortex we found a common representation for planned hand- and tool-related actions. In contrast, in parietal and occipitotemporal regions implicated in hand actions and body perception we found that coding remained selectively linked to upcoming actions of the hand whereas in parietal and occipitotemporal regions implicated in tool-related processing the coding remained selectively linked to upcoming actions of the tool. The highly specialized and hierarchical nature of this coding suggests that hand- and tool-related actions are represented separately at earlier levels of sensorimotor processing before becoming integrated in frontoparietal cortex. DOI:http://dx.doi.org/10.7554/eLife.00425.001.

Norm-based coding of voice identity in human auditory cortex

Listeners exploit small interindividual variations around a generic acoustical structure to discriminate and identify individuals from their voice—a key requirement for social interactions. The human brain contains temporal voice areas (TVA) [1] involved in an acoustic-based representation of voice identity [2, 3, 4, 5, 6], but the underlying coding mechanisms remain unknown. Indirect evidence suggests that identity representation in these areas could rely on a norm-based coding mechanism [4, 7, 8, 9, 10, 11]. Here, we show by using fMRI that voice identity is coded in the TVA as a function of acoustical distance to two internal voice prototypes (one male, one female)—approximated here by averaging a large number of same-gender voices by using morphing [12]. Voices more distant from their prototype are perceived as more distinctive and elicit greater neuronal activity in voice-sensitive cortex than closer voices—a phenomenon not merely explained by neuronal adaptation [13, 14]. Moreover, explicit manipulations of distance-to-mean by morphing voices toward (or away from) their prototype elicit reduced (or enhanced) neuronal activity. These results indicate that voice-sensitive cortex integrates relevant acoustical features into a complex representation referenced to idealized male and female voice prototypes. More generally, they shed light on remarkable similarities in cerebral representations of facial and vocal identity.

•

Identity coding in temporal voice area in terms of acoustical distance to prototypes

•

Description of the “voice space” in terms of simple acoustical measures

•

Male and female prototypes are ideally smooth versions of the population means

•

Comparable coding mechanism for identity across sensory modalities

Sex matters: Neural correlates of voice gender perception

The basis for different neural activations in response to male and female voices as well as the question, whether men and women perceive male and female voices differently, has not been thoroughly investigated. Therefore, the aim of the present study was to examine the behavioral and neural correlates of gender-related voice perception in healthy male and female volunteers. fMRI data were collected while 39 participants (19 female) were asked to indicate the gender of 240 voice stimuli. These stimuli included recordings of 3-syllable nouns as well as the same recordings pitch-shifted in 2, 4 and 6 semitone steps in the direction of the other gender. Data analysis revealed a) equal voice discrimination sensitivity in men and women but better performance in the categorization of opposite-sex stimuli at least in men, b) increased responses to increasing gender ambiguity in the mid cingulate cortex and bilateral inferior frontal gyri, and c) stronger activation in a fronto-temporal neural network in response to voices of the opposite sex. Our results indicate a gender specific processing for male and female voices on a behavioral and neuronal level. We suggest that our results reflect higher sensitivity probably due to the evolutionary relevance of voice perception in mate selection.

Activity patterns in the category-selective occipitotemporal cortex predict upcoming motor actions

Converging lines of evidence point to the occipitotemporal cortex (OTC) as a critical structure in visual perception. For instance, human functional magnetic resonance imaging (fMRI) has revealed a modular organisation of object-selective, face-selective, body-selective and scene-selective visual areas in the OTC, and disruptions to the processing within these regions, either in neuropsychological patients or through transcranial magnetic stimulation, can produce category-specific deficits in visual recognition. Here we show, using fMRI and pattern classification methods, that the activity in the OTC also represents how stimuli will be interacted with by the body--a level of processing more traditionally associated with the preparatory activity in sensorimotor circuits of the brain. Combining functional mapping of different OTC areas with a real object-directed delayed movement task, we found that the pre-movement spatial activity patterns across the OTC could be used to predict both the action of an upcoming hand movement (grasping vs. reaching) and the effector (left hand vs. right hand) to be used. Interestingly, we were able to extract this wide range of predictive movement information even though nearly all OTC areas showed either baseline-level or below baseline-level activity prior to action onset. Our characterisation of different OTC areas according to the features of upcoming movements that they could predict also revealed a general gradient of effector-to-action-dependent movement representations along the posterior-anterior OTC axis. These findings suggest that the ventral visual pathway, which is well known to be involved in object recognition and perceptual processing, plays a larger than previously expected role in preparing object-directed hand actions.

Where one hand meets the other: limb-specific and action-dependent movement plans decoded from preparatory signals in single human frontoparietal brain areas

Planning object-directed hand actions requires successful integration of the movement goal with the acting limb. Exactly where and how this sensorimotor integration occurs in the brain has been studied extensively with neurophysiological recordings in nonhuman primates, yet to date, because of limitations of non-invasive methodologies, the ability to examine the same types of planning-related signals in humans has been challenging. Here we show, using a multivoxel pattern analysis of functional MRI (fMRI) data, that the preparatory activity patterns in several frontoparietal brain regions can be used to predict both the limb used and hand action performed in an upcoming movement. Participants performed an event-related delayed movement task whereby they planned and executed grasp or reach actions with either their left or right hand toward a single target object. We found that, although the majority of frontoparietal areas represented hand actions (grasping vs reaching) for the contralateral limb, several areas additionally coded hand actions for the ipsilateral limb. Notable among these were subregions within the posterior parietal cortex (PPC), dorsal premotor cortex (PMd), ventral premotor cortex, dorsolateral prefrontal cortex, presupplementary motor area, and motor cortex, a region more traditionally implicated in contralateral movement generation. Additional analyses suggest that hand actions are represented independently of the intended limb in PPC and PMd. In addition to providing a unique mapping of limb-specific and action-dependent intention-related signals across the human cortical motor system, these findings uncover a much stronger representation of the ipsilateral limb than expected from previous fMRI findings.

Consciousness and the invention of Morel

A scientific study of consciousness should take into consideration both objective and subjective measures of conscious experiences. To this date, very few studies have tried to integrate third-person data, or data about the neurophysiological correlates of conscious states, with first-person data, or data about subjective experience. Inspired by Morel's invention (Casares, 1940), a literary machine capable of reproducing sensory-dependent external reality, this article suggests that combination of virtual reality techniques and brain reading technologies, that is, decoding of conscious states by brain activity alone, can offer this integration. It is also proposed that the multimodal, simulating, and integrative capacities of the dreaming brain render it an “endogenous” Morel's machine, which can potentially be used in studying consciousness, but not always in a reliable way. Both the literary machine and dreaming could contribute to a better understanding of conscious states.

The free-energy self: a predictive coding account of self-recognition

Recognising and representing one's self as distinct from others is a fundamental component of self-awareness. However, current theories of self-recognition are not embedded within global theories of cortical function and therefore fail to provide a compelling explanation of how the self is processed. We present a theoretical account of the neural and computational basis of self-recognition that is embedded within the free-energy account of cortical function. In this account one's body is processed in a Bayesian manner as the most likely to be "me". Such probabilistic representation arises through the integration of information from hierarchically organised unimodal systems in higher-level multimodal areas. This information takes the form of bottom-up "surprise" signals from unimodal sensory systems that are explained away by top-down processes that minimise the level of surprise across the brain. We present evidence that this theoretical perspective may account for the findings of psychological and neuroimaging investigations into self-recognition and particularly evidence that representations of the self are malleable, rather than fixed as previous accounts of self-recognition might suggest.

Dynamic and static facial expressions decoded from motion-sensitive areas in the macaque monkey

Humans adeptly use visual motion to recognize socially relevant facial information. The macaque provides a model visual system for studying neural coding of expression movements, as its superior temporal sulcus (STS) possesses brain areas selective for faces and areas sensitive to visual motion. We used functional magnetic resonance imaging and facial stimuli to localize motion-sensitive areas [motion in faces (Mf) areas], which responded more to dynamic faces compared with static faces, and face-selective areas, which responded selectively to faces compared with objects and places. Using multivariate analysis, we found that information about both dynamic and static facial expressions could be robustly decoded from Mf areas. By contrast, face-selective areas exhibited relatively less facial expression information. Classifiers trained with expressions from one motion type (dynamic or static) showed poor generalization to the other motion type, suggesting that Mf areas employ separate and nonconfusable neural codes for dynamic and static presentations of the same expressions. We also show that some of the motion sensitivity elicited by facial stimuli was not specific to faces but could also be elicited by moving dots, particularly in fundus of the superior temporal and middle superior temporal polysensory/lower superior temporal areas, confirming their already well established low-level motion sensitivity. A different pattern was found in anterior STS, which responded more to dynamic than to static faces but was not sensitive to dot motion. Overall, we show that emotional expressions are mostly represented outside of face-selective cortex, in areas sensitive to motion. These regions may play a fundamental role in enhancing recognition of facial expression despite the complex stimulus changes associated with motion.

Auditory modulation of visual stimulus encoding in human retinotopic cortex

Sounds can modulate visual perception as well as neural activity in retinotopic cortex. Most studies in this context investigated how sounds change neural amplitude and oscillatory phase reset in visual cortex. However, recent studies in macaque monkeys show that congruence of audio-visual stimuli also modulates the amount of stimulus information carried by spiking activity of primary auditory and visual neurons. Here, we used naturalistic video stimuli and recorded the spatial patterns of functional MRI signals in human retinotopic cortex to test whether the discriminability of such patterns varied with the presence and congruence of co-occurring sounds. We found that incongruent sounds significantly impaired stimulus decoding from area V2 and there was a similar trend for V3. This effect was associated with reduced inter-trial reliability of patterns (i.e. higher levels of noise), but was not accompanied by any detectable modulation of overall signal amplitude. We conclude that sounds modulate naturalistic stimulus encoding in early human retinotopic cortex without affecting overall signal amplitude. Subthreshold modulation, oscillatory phase reset and dynamic attentional modulation are candidate neural and cognitive mechanisms mediating these effects.

Processing of natural sounds in human auditory cortex: tonotopy, spectral tuning, and relation to voice sensitivity

Auditory cortical processing of complex meaningful sounds entails the transformation of sensory (tonotopic) representations of incoming acoustic waveforms into higher-level sound representations (e.g., their category). However, the precise neural mechanisms enabling such transformations remain largely unknown. In the present study, we use functional magnetic resonance imaging (fMRI) and natural sounds stimulation to examine these two levels of sound representation (and their relation) in the human auditory cortex. In a first experiment, we derive cortical maps of frequency preference (tonotopy) and selectivity (tuning width) by mathematical modeling of fMRI responses to natural sounds. The tuning width maps highlight a region of narrow tuning that follows the main axis of Heschl's gyrus and is flanked by regions of broader tuning. The narrowly tuned portion on Heschl's gyrus contains two mirror-symmetric frequency gradients, presumably defining two distinct primary auditory areas. In addition, our analysis indicates that spectral preference and selectivity (and their topographical organization) extend well beyond the primary regions and also cover higher-order and category-selective auditory regions. In particular, regions with preferential responses to human voice and speech occupy the low-frequency portions of the tonotopic map. We confirm this observation in a second experiment, where we find that speech/voice selective regions exhibit a response bias toward the low frequencies characteristic of human voice and speech, even when responding to simple tones. We propose that this frequency bias reflects the selective amplification of relevant and category-characteristic spectral bands, a useful processing step for transforming a sensory (tonotopic) sound image into higher level neural representations.

Decoding speech for understanding and treating aphasia

Aphasia is an acquired language disorder with a diverse set of symptoms that can affect virtually any linguistic modality across both the comprehension and production of spoken language. Partial recovery of language function after injury is common but typically incomplete. Rehabilitation strategies focus on behavioral training to induce plasticity in underlying neural circuits to maximize linguistic recovery. Understanding the different neural circuits underlying diverse language functions is a key to developing more effective treatment strategies. This chapter discusses a systems identification analytic approach to the study of linguistic neural representation. The focus of this framework is a quantitative, model-based characterization of speech and language neural representations that can be used to decode, or predict, speech representations from measured brain activity. Recent results of this approach are discussed in the context of applications to understanding the neural basis of aphasia symptoms and the potential to optimize plasticity during the rehabilitation process.

A sparse representation-based algorithm for pattern localization in brain imaging data analysis

Considering the two-class classification problem in brain imaging data analysis, we propose a sparse representation-based multi-variate pattern analysis (MVPA) algorithm to localize brain activation patterns corresponding to different stimulus classes/brain states respectively. Feature selection can be modeled as a sparse representation (or sparse regression) problem. Such technique has been successfully applied to voxel selection in fMRI data analysis. However, single selection based on sparse representation or other methods is prone to obtain a subset of the most informative features rather than all. Herein, our proposed algorithm recursively eliminates informative features selected by a sparse regression method until the decoding accuracy based on the remaining features drops to a threshold close to chance level. In this way, the resultant feature set including all the identified features is expected to involve all the informative features for discrimination. According to the signs of the sparse regression weights, these selected features are separated into two sets corresponding to two stimulus classes/brain states. Next, in order to remove irrelevant/noisy features in the two selected feature sets, we perform a nonparametric permutation test at the individual subject level or the group level. In data analysis, we verified our algorithm with a toy data set and an intrinsic signal optical imaging data set. The results show that our algorithm has accurately localized two class-related patterns. As an application example, we used our algorithm on a functional magnetic resonance imaging (fMRI) data set. Two sets of informative voxels, corresponding to two semantic categories (i.e., “old people” and “young people”), respectively, are obtained in the human brain.

Learning of new sound categories shapes neural response patterns in human auditory cortex

The formation of new sound categories is fundamental to everyday goal-directed behavior. Categorization requires the abstraction of discrete classes from continuous physical features as required by context and task. Electrophysiology in animals has shown that learning to categorize novel sounds alters their spatiotemporal neural representation at the level of early auditory cortex. However, functional magnetic resonance imaging (fMRI) studies so far did not yield insight into the effects of category learning on sound representations in human auditory cortex. This may be due to the use of overlearned speech-like categories and fMRI subtraction paradigms, leading to insufficient sensitivity to distinguish the responses to learning-induced, novel sound categories. Here, we used fMRI pattern analysis to investigate changes in human auditory cortical response patterns induced by category learning. We created complex novel sound categories and analyzed distributed activation patterns during passive listening to a sound continuum before and after category learning. We show that only after training, sound categories could be successfully decoded from early auditory areas and that learning-induced pattern changes were specific to the category-distinctive sound feature (i.e., pitch). Notably, the similarity between fMRI response patterns for the sound continuum mirrored the sigmoid shape of the behavioral category identification function. Our results indicate that perceptual representations of novel sound categories emerge from neural changes at early levels of the human auditory processing hierarchy.

Distinguishing grammatical constructions with fMRI pattern analysis

All linguistic and psycholinguistic theories aim to provide psychologically valid analyses of particular grammatical patterns and the relationships that hold among them. Until recently, no tools were available to distinguish neural correlates of particular grammatical constructions that shared the same content words, propositional meaning, and degree of surface complexity, such as the dative (e.g., Sally gave the book to Joe) and the ditransitive (e.g., Sally gave Joe a book). We report the first fMRI data that distinguish such closely related, abstract grammatical patterns. Multi-voxel pattern analysis (MVPA) proved capable of discriminating at above-chance levels between activity patterns arising during reading of dative and ditransitive sentences. Region-of-interest analyses reveal that the union of certain language-relevant areas, anterior and posterior BA22, BA44/45 and BA47, yield classification accuracy above chance and above that of control conditions in the left hemisphere but not in the right. Looking more closely at the LH ROIs, we find that the combination of areas aBA22 and BA47 is sufficient to distinguish the two constructions better than the controls and better than chance. The fact that both of these areas-particularly BA47-have been implicated in semantics, lends support to claims that the two constructions are distinguishable semantically. More generally, the ability to distinguish closely related grammatical constructions using MVPA offers the promise of addressing traditional theoretical questions on a neuroscientifically grounded basis.

Using phase to recognize English phonemes and their distinctive features in the brain

The neural mechanisms used by the human brain to identify phonemes remain unclear. We recorded the EEG signals evoked by repeated presentation of 12 American English phonemes. A support vector machine model correctly recognized a high percentage of the EEG brain wave recordings represented by their phases, which were expressed in discrete Fourier transform coefficients. We show that phases of the oscillations restricted to the frequency range of 2-9 Hz can be used to successfully recognize brain processing of these phonemes. The recognition rates can be further improved using the scalp tangential electric field and the surface Laplacian around the auditory cortical area, which were derived from the original potential signal. The best rate for the eight initial consonants was 66.7%. Moreover, we found a distinctive phase pattern in the brain for each of these consonants. We then used these phase patterns to recognize the consonants, with a correct rate of 48.7%. In addition, in the analysis of the confusion matrices, we found significant similarity-differences were invariant between brain and perceptual representations of phonemes. These latter results supported the importance of phonological distinctive features in the neural representation of phonemes.

What makes a pattern? Matching decoding methods to data in multivariate pattern analysis

Research in neuroscience faces the challenge of integrating information across different spatial scales of brain function. A promising technique for harnessing information at a range of spatial scales is multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) data. While the prevalence of MVPA has increased dramatically in recent years, its typical implementations for classification of mental states utilize only a subset of the information encoded in local fMRI signals. We review published studies employing multivariate pattern classification since the technique’s introduction, which reveal an extensive focus on the improved detection power that linear classifiers provide over traditional analysis techniques. We demonstrate using simulations and a searchlight approach, however, that non-linear classifiers are capable of extracting distinct information about interactions within a local region. We conclude that for spatially localized analyses, such as searchlight and region of interest, multiple classification approaches should be compared in order to match fMRI analyses to the properties of local circuits.

Music in our ears: the biological bases of musical timbre perception

Timbre is the attribute of sound that allows humans and other animals to distinguish among different sound sources. Studies based on psychophysical judgments of musical timbre, ecological analyses of sound's physical characteristics as well as machine learning approaches have all suggested that timbre is a multifaceted attribute that invokes both spectral and temporal sound features. Here, we explored the neural underpinnings of musical timbre. We used a neuro-computational framework based on spectro-temporal receptive fields, recorded from over a thousand neurons in the mammalian primary auditory cortex as well as from simulated cortical neurons, augmented with a nonlinear classifier. The model was able to perform robust instrument classification irrespective of pitch and playing style, with an accuracy of 98.7%. Using the same front end, the model was also able to reproduce perceptual distance judgments between timbres as perceived by human listeners. The study demonstrates that joint spectro-temporal features, such as those observed in the mammalian primary auditory cortex, are critical to provide the rich-enough representation necessary to account for perceptual judgments of timbre by human listeners, as well as recognition of musical instruments.

Processing of speech and non-speech sounds in the supratemporal plane: auditory input preference does not predict sensitivity to statistical structure

The supratemporal plane contains several functionally heterogeneous subregions that respond strongly to speech. Much of the prior work on the issue of speech processing in the supratemporal plane has focused on neural responses to single speech vs. non-speech sounds rather than focusing on higher-level computations that are required to process more complex auditory sequences. Here we examined how information is integrated over time for speech and non-speech sounds by quantifying the BOLD fMRI response to stochastic (non-deterministic) sequences of speech and non-speech naturalistic sounds that varied in their statistical structure (from random to highly structured sequences) during passive listening. Behaviorally, the participants were accurate in segmenting speech and non-speech sequences, though they were more accurate for speech. Several supratemporal regions showed increased activation magnitude for speech sequences (preference), but, importantly, this did not predict sensitivity to statistical structure: (i) several areas showing a speech preference were sensitive to statistical structure in both speech and non-speech sequences, and (ii) several regions that responded to both speech and non-speech sounds showed distinct responses to statistical structure in speech and non-speech sequences. While the behavioral findings highlight the tight relation between statistical structure and segmentation processes, the neuroimaging results suggest that the supratemporal plane mediates complex statistical processing for both speech and non-speech sequences and emphasize the importance of studying the neurocomputations associated with auditory sequence processing. These findings identify new partitions of functionally distinct areas in the supratemporal plane that cannot be evoked by single stimuli. The findings demonstrate the importance of going beyond input preference to examine the neural computations implemented in the superior temporal plane.