A Supramodal Neural Network for Speech and Gesture Semantics: An fMRI Study

In a natural setting, speech is often accompanied by gestures. As language, speech-accompanying iconic gestures to some extent convey semantic information. However, if comprehension of the information contained in both the auditory and visual modality depends on same or different brain-networks is quite unknown. In this fMRI study, we aimed at identifying the cortical areas engaged in supramodal processing of semantic information. BOLD changes were recorded in 18 healthy right-handed male subjects watching video clips showing an actor who either performed speech (S, acoustic) or gestures (G, visual) in more (+) or less (−) meaningful varieties. In the experimental conditions familiar speech or isolated iconic gestures were presented; during the visual control condition the volunteers watched meaningless gestures (G−), while during the acoustic control condition a foreign language was presented (S−). The conjunction of the visual and acoustic semantic processing revealed activations extending from the left inferior frontal gyrus to the precentral gyrus, and included bilateral posterior temporal regions. We conclude that proclaiming this frontotemporal network the brain''s core language system is to take too narrow a view. Our results rather indicate that these regions constitute a supramodal semantic processing network.     

Beat gestures influence which speech sounds you hear

Beat gestures—spontaneously produced biphasic movements of the hand—are among the most frequently encountered co-speech gestures in human communication. They are closely temporally aligned to the prosodic characteristics of the speech signal, typically occurring on lexically stressed syllables. Despite their prevalence across speakers of the world''s languages, how beat gestures impact spoken word recognition is unclear. Can these simple ‘flicks of the hand'' influence speech perception? Across a range of experiments, we demonstrate that beat gestures influence the explicit and implicit perception of lexical stress (e.g. distinguishing OBject from obJECT), and in turn can influence what vowels listeners hear. Thus, we provide converging evidence for a manual McGurk effect: relatively simple and widely occurring hand movements influence which speech sounds we hear.     

The purity of auditory memory

Recent evidence from experiments on immediate memory indicates unambiguously that silent speech perception can produce typically 'auditory' effects while there is either active or passive mouthing of the relevant articulatory gestures. This result falsifies previous theories of auditory sensory memory (pre-categorical acoustic store) that insisted on external auditory stimulation as indispensable for access to the system. A resolution is proposed that leaves the properties of pre-categorical acoustic store much as they were assumed to be before but adds the possibility that visual information can affect the selection of auditory features in a pre-categorical stage of speech perception. In common terms, a speaker's facial gestures (or one's own) can influence auditory experience independently of determining what it was that was said. Some results in word perception that encourage this view are discussed.     

Reconstructing speech from human auditory cortex

How the human auditory system extracts perceptually relevant acoustic features of speech is unknown. To address this question, we used intracranial recordings from nonprimary auditory cortex in the human superior temporal gyrus to determine what acoustic information in speech sounds can be reconstructed from population neural activity. We found that slow and intermediate temporal fluctuations, such as those corresponding to syllable rate, were accurately reconstructed using a linear model based on the auditory spectrogram. However, reconstruction of fast temporal fluctuations, such as syllable onsets and offsets, required a nonlinear sound representation based on temporal modulation energy. Reconstruction accuracy was highest within the range of spectro-temporal fluctuations that have been found to be critical for speech intelligibility. The decoded speech representations allowed readout and identification of individual words directly from brain activity during single trial sound presentations. These findings reveal neural encoding mechanisms of speech acoustic parameters in higher order human auditory cortex.     

The Human Brain Maintains Contradictory and Redundant Auditory Sensory Predictions

Computational and experimental research has revealed that auditory sensory predictions are derived from regularities of the current environment by using internal generative models. However, so far, what has not been addressed is how the auditory system handles situations giving rise to redundant or even contradictory predictions derived from different sources of information. To this end, we measured error signals in the event-related brain potentials (ERPs) in response to violations of auditory predictions. Sounds could be predicted on the basis of overall probability, i.e., one sound was presented frequently and another sound rarely. Furthermore, each sound was predicted by an informative visual cue. Participants’ task was to use the cue and to discriminate the two sounds as fast as possible. Violations of the probability based prediction (i.e., a rare sound) as well as violations of the visual-auditory prediction (i.e., an incongruent sound) elicited error signals in the ERPs (Mismatch Negativity [MMN] and Incongruency Response [IR]). Particular error signals were observed even in case the overall probability and the visual symbol predicted different sounds. That is, the auditory system concurrently maintains and tests contradictory predictions. Moreover, if the same sound was predicted, we observed an additive error signal (scalp potential and primary current density) equaling the sum of the specific error signals. Thus, the auditory system maintains and tolerates functionally independently represented redundant and contradictory predictions. We argue that the auditory system exploits all currently active regularities in order to optimally prepare for future events.     

The mismatch negativity as an index of the perception of speech sounds by the human brain

The present article outlines the contribution of the mismatch negativity (MMN), and its magnetic equivalent MMNm, to our understanding of the perception of speech sounds in the human brain. MMN data indicate that each sound, both speech and non-speech, develops its neural representation corresponding to the percept of this sound in the neurophysiological substrate of auditory sensory memory. The accuracy of this representation, determining the accuracy of the discrimination between different sounds, can be probed with MMN separately for any auditory feature or stimulus type such as phonemes. Furthermore, MMN data show that the perception of phonemes, and probably also of larger linguistic units (syllables and words), is based on language-specific phonetic traces developed in the posterior part of the left-hemisphere auditory cortex. These traces serve as recognition models for the corresponding speech sounds in listening to speech.     

Age Differences in the Neuroelectric Adaptation to Meaningful Sounds

Much of what we know regarding the effect of stimulus repetition on neuroelectric adaptation comes from studies using artificially produced pure tones or harmonic complex sounds. Little is known about the neural processes associated with the representation of everyday sounds and how these may be affected by aging. In this study, we used real life, meaningful sounds presented at various azimuth positions and found that auditory evoked responses peaking at about 100 and 180 ms after sound onset decreased in amplitude with stimulus repetition. This neural adaptation was greater in young than in older adults and was more pronounced when the same sound was repeated at the same location. Moreover, the P2 waves showed differential patterns of domain-specific adaptation when location and identity was repeated among young adults. Background noise decreased ERP amplitudes and modulated the magnitude of repetition effects on both the N1 and P2 amplitude, and the effects were comparable in young and older adults. These findings reveal an age-related difference in the neural processes associated with adaptation to meaningful sounds, which may relate to older adults’ difficulty in ignoring task-irrelevant stimuli.     

The Brain’s Dorsal Route for Speech Represents Word Meaning: Evidence from Gesture

The dual-route model of speech processing includes a dorsal stream that maps auditory to motor features at the sublexical level rather than at the lexico-semantic level. However, the literature on gesture is an invitation to revise this model because it suggests that the premotor cortex of the dorsal route is a major site of lexico-semantic interaction. Here we investigated lexico-semantic mapping using word-gesture pairs that were either congruent or incongruent. Using fMRI-adaptation in 28 subjects, we found that temporo-parietal and premotor activity during auditory processing of single action words was modulated by the prior audiovisual context in which the words had been repeated. The BOLD signal was suppressed following repetition of the auditory word alone, and further suppressed following repetition of the word accompanied by a congruent gesture (e.g. [“grasp” + grasping gesture]). Conversely, repetition suppression was not observed when the same action word was accompanied by an incongruent gesture (e.g. [“grasp” + sprinkle]). We propose a simple model to explain these results: auditory and visual information converge onto premotor cortex where it is represented in a comparable format to determine (in)congruence between speech and gesture. This ability of the dorsal route to detect audiovisual semantic (in)congruence suggests that its function is not restricted to the sublexical level.     

Coding of communication calls in the subcortical and cortical structures of the auditory system

The processing of species-specific communication signals in the auditory system represents an important aspect of animal behavior and is crucial for its social interactions, reproduction, and survival. In this article the neuronal mechanisms underlying the processing of communication signals in the higher centers of the auditory system--inferior colliculus (IC), medial geniculate body (MGB) and auditory cortex (AC)--are reviewed, with particular attention to the guinea pig. The selectivity of neuronal responses for individual calls in these auditory centers in the guinea pig is usually low--most neurons respond to calls as well as to artificial sounds; the coding of complex sounds in the central auditory nuclei is apparently based on the representation of temporal and spectral features of acoustical stimuli in neural networks. Neuronal response patterns in the IC reliably match the sound envelope for calls characterized by one or more short impulses, but do not exactly fit the envelope for long calls. Also, the main spectral peaks are represented by neuronal firing rates in the IC. In comparison to the IC, response patterns in the MGB and AC demonstrate a less precise representation of the sound envelope, especially in the case of longer calls. The spectral representation is worse in the case of low-frequency calls, but not in the case of broad-band calls. The emotional content of the call may influence neuronal responses in the auditory pathway, which can be demonstrated by stimulation with time-reversed calls or by measurements performed under different levels of anesthesia. The investigation of the principles of the neural coding of species-specific vocalizations offers some keys for understanding the neural mechanisms underlying human speech perception.     

Hearing and seeing meaning in speech and gesture: insights from brain and behaviour

As we speak, we use not only the arbitrary form–meaning mappings of the speech channel but also motivated form–meaning correspondences, i.e. iconic gestures that accompany speech (e.g. inverted V-shaped hand wiggling across gesture space to demonstrate walking). This article reviews what we know about processing of semantic information from speech and iconic gestures in spoken languages during comprehension of such composite utterances. Several studies have shown that comprehension of iconic gestures involves brain activations known to be involved in semantic processing of speech: i.e. modulation of the electrophysiological recording component N400, which is sensitive to the ease of semantic integration of a word to previous context, and recruitment of the left-lateralized frontal–posterior temporal network (left inferior frontal gyrus (IFG), medial temporal gyrus (MTG) and superior temporal gyrus/sulcus (STG/S)). Furthermore, we integrate the information coming from both channels recruiting brain areas such as left IFG, posterior superior temporal sulcus (STS)/MTG and even motor cortex. Finally, this integration is flexible: the temporal synchrony between the iconic gesture and the speech segment, as well as the perceived communicative intent of the speaker, modulate the integration process. Whether these findings are special to gestures or are shared with actions or other visual accompaniments to speech (e.g. lips) or other visual symbols such as pictures are discussed, as well as the implications for a multimodal view of language.     

No,There Is No 150 ms Lead of Visual Speech on Auditory Speech,but a Range of Audiovisual Asynchronies Varying from Small Audio Lead to Large Audio Lag

An increasing number of neuroscience papers capitalize on the assumption published in this journal that visual speech would be typically 150 ms ahead of auditory speech. It happens that the estimation of audiovisual asynchrony in the reference paper is valid only in very specific cases, for isolated consonant-vowel syllables or at the beginning of a speech utterance, in what we call “preparatory gestures”. However, when syllables are chained in sequences, as they are typically in most parts of a natural speech utterance, asynchrony should be defined in a different way. This is what we call “comodulatory gestures” providing auditory and visual events more or less in synchrony. We provide audiovisual data on sequences of plosive-vowel syllables (pa, ta, ka, ba, da, ga, ma, na) showing that audiovisual synchrony is actually rather precise, varying between 20 ms audio lead and 70 ms audio lag. We show how more complex speech material should result in a range typically varying between 40 ms audio lead and 200 ms audio lag, and we discuss how this natural coordination is reflected in the so-called temporal integration window for audiovisual speech perception. Finally we present a toy model of auditory and audiovisual predictive coding, showing that visual lead is actually not necessary for visual prediction.     

Temporal predictive codes for spoken words in auditory cortex

Humans can recognize spoken words with unmatched speed and accuracy. Hearing the initial portion of a word such as "formu…" is sufficient for the brain to identify "formula" from the thousands of other words that partially match. Two alternative computational accounts propose that partially matching words (1) inhibit each other until a single word is selected ("formula" inhibits "formal" by lexical competition) or (2) are used to predict upcoming speech sounds more accurately (segment prediction error is minimal after sequences like "formu…"). To distinguish these theories we taught participants novel words (e.g., "formubo") that sound like existing words ("formula") on two successive days. Computational simulations show that knowing "formubo" increases lexical competition when hearing "formu…", but reduces segment prediction error. Conversely, when the sounds in "formula" and "formubo" diverge, the reverse is observed. The time course of magnetoencephalographic brain responses in the superior temporal gyrus (STG) is uniquely consistent with a segment prediction account. We propose a predictive coding model of spoken word recognition in which STG neurons represent the difference between predicted and heard speech sounds. This prediction error signal explains the efficiency of human word recognition and simulates neural responses in auditory regions.     

Basic auditory processes involved in the analysis of speech sounds

This paper reviews the basic aspects of auditory processing that play a role in the perception of speech. The frequency selectivity of the auditory system, as measured using masking experiments, is described and used to derive the internal representation of the spectrum (the excitation pattern) of speech sounds. The perception of timbre and distinctions in quality between vowels are related to both static and dynamic aspects of the spectra of sounds. The perception of pitch and its role in speech perception are described. Measures of the temporal resolution of the auditory system are described and a model of temporal resolution based on a sliding temporal integrator is outlined. The combined effects of frequency and temporal resolution can be modelled by calculation of the spectro-temporal excitation pattern, which gives good insight into the internal representation of speech sounds. For speech presented in quiet, the resolution of the auditory system in frequency and time usually markedly exceeds the resolution necessary for the identification or discrimination of speech sounds, which partly accounts for the robust nature of speech perception. However, for people with impaired hearing, speech perception is often much less robust.     

Auditory Cortex Processes Variation in Our Own Speech

As we talk, we unconsciously adjust our speech to ensure it sounds the way we intend it to sound. However, because speech production involves complex motor planning and execution, no two utterances of the same sound will be exactly the same. Here, we show that auditory cortex is sensitive to natural variations in self-produced speech from utterance to utterance. We recorded event-related potentials (ERPs) from ninety-nine subjects while they uttered “ah” and while they listened to those speech sounds played back. Subjects'' utterances were sorted based on their formant deviations from the previous utterance. Typically, the N1 ERP component is suppressed during talking compared to listening. By comparing ERPs to the least and most variable utterances, we found that N1 was less suppressed to utterances that differed greatly from their preceding neighbors. In contrast, an utterance''s difference from the median formant values did not affect N1. Trial-to-trial pitch (f0) deviation and pitch difference from the median similarly did not affect N1. We discuss mechanisms that may underlie the change in N1 suppression resulting from trial-to-trial formant change. Deviant utterances require additional auditory cortical processing, suggesting that speaking-induced suppression mechanisms are optimally tuned for a specific production.     

Dynamics of Vocalization-Induced Modulation of Auditory Cortical Activity at Mid-utterance

Background

Recent research has addressed the suppression of cortical sensory responses to altered auditory feedback that occurs at utterance onset regarding speech. However, there is reason to assume that the mechanisms underlying sensorimotor processing at mid-utterance are different than those involved in sensorimotor control at utterance onset. The present study attempted to examine the dynamics of event-related potentials (ERPs) to different acoustic versions of auditory feedback at mid-utterance.

Methodology/Principal findings

Subjects produced a vowel sound while hearing their pitch-shifted voice (100 cents), a sum of their vocalization and pure tones, or a sum of their vocalization and white noise at mid-utterance via headphones. Subjects also passively listened to playback of what they heard during active vocalization. Cortical ERPs were recorded in response to different acoustic versions of feedback changes during both active vocalization and passive listening. The results showed that, relative to passive listening, active vocalization yielded enhanced P2 responses to the 100 cents pitch shifts, whereas suppression effects of P2 responses were observed when voice auditory feedback was distorted by pure tones or white noise.

Conclusion/Significance

The present findings, for the first time, demonstrate a dynamic modulation of cortical activity as a function of the quality of acoustic feedback at mid-utterance, suggesting that auditory cortical responses can be enhanced or suppressed to distinguish self-produced speech from externally-produced sounds.     

Speech recognition by an artificial neural network using findings on the afferent auditory system

An artificial neural network which uses anatomical and physiological findings on the afferent pathway from the ear to the cortex is presented and the roles of the constituent functions in recognition of continuous speech are examined. The network deals with successive spectra of speech sounds by a cascade of several neural layers: lateral excitation layer (LEL), lateral inhibition layer (LIL), and a pile of feature detection layers (FDL's). These layers are shown to be effective for recognizing spoken words. Namely, first, LEL reduces the distortion of sound spectrum caused by the pitch of speech sounds. Next, LIL emphasizes the major energy peaks of sound spectrum, the formants. Last, FDL's detect syllables and words in successive formants, where two functions, time-delay and strong adaptation, play important roles: time-delay makes it possible to retain the pattern of formant changes for a period to detect spoken words successively; strong adaptation contributes to removing the time-warp of formant changes. Digital computer simulations show that the network detect isolated syllables, isolated words, and connected words in continuous speech, while reproducing the fundamental responses found in the auditory system such as ON, OFF, ON-OFF, and SUSTAINED patterns.     

From Birdsong to Human Speech Recognition: Bayesian Inference on a Hierarchy of Nonlinear Dynamical Systems

Our knowledge about the computational mechanisms underlying human learning and recognition of sound sequences, especially speech, is still very limited. One difficulty in deciphering the exact means by which humans recognize speech is that there are scarce experimental findings at a neuronal, microscopic level. Here, we show that our neuronal-computational understanding of speech learning and recognition may be vastly improved by looking at an animal model, i.e., the songbird, which faces the same challenge as humans: to learn and decode complex auditory input, in an online fashion. Motivated by striking similarities between the human and songbird neural recognition systems at the macroscopic level, we assumed that the human brain uses the same computational principles at a microscopic level and translated a birdsong model into a novel human sound learning and recognition model with an emphasis on speech. We show that the resulting Bayesian model with a hierarchy of nonlinear dynamical systems can learn speech samples such as words rapidly and recognize them robustly, even in adverse conditions. In addition, we show that recognition can be performed even when words are spoken by different speakers and with different accents—an everyday situation in which current state-of-the-art speech recognition models often fail. The model can also be used to qualitatively explain behavioral data on human speech learning and derive predictions for future experiments.     

Hemispheric asymmetries in speech perception: sense, nonsense and modulations

Background

The well-established left hemisphere specialisation for language processing has long been claimed to be based on a low-level auditory specialization for specific acoustic features in speech, particularly regarding ‘rapid temporal processing’.

Methodology

A novel analysis/synthesis technique was used to construct a variety of sounds based on simple sentences which could be manipulated in spectro-temporal complexity, and whether they were intelligible or not. All sounds consisted of two noise-excited spectral prominences (based on the lower two formants in the original speech) which could be static or varying in frequency and/or amplitude independently. Dynamically varying both acoustic features based on the same sentence led to intelligible speech but when either or both acoustic features were static, the stimuli were not intelligible. Using the frequency dynamics from one sentence with the amplitude dynamics of another led to unintelligible sounds of comparable spectro-temporal complexity to the intelligible ones. Positron emission tomography (PET) was used to compare which brain regions were active when participants listened to the different sounds.

Conclusions

Neural activity to spectral and amplitude modulations sufficient to support speech intelligibility (without actually being intelligible) was seen bilaterally, with a right temporal lobe dominance. A left dominant response was seen only to intelligible sounds. It thus appears that the left hemisphere specialisation for speech is based on the linguistic properties of utterances, not on particular acoustic features.