Auditory induction represents one of the most striking demonstrations of perceptual synthesis in human cognitive neuroscience, showing that what we hear is not merely a passive recording of atmospheric pressure waves, but an active, predictive construction of the brain. When acoustic signals such as speech or music are momentarily occluded by louder extraneous noise, the auditory cortex dynamically reconstructs the missing fragments, creating an uninterrupted conscious experience. This illusion of continuity illustrates the sophisticated predictive coding and acoustic scene analysis mechanisms that allow organisms to navigate noisy, unpredictable acoustic environments.
Auditory Induction
1. Concise Definition
Auditory induction, often referred to interchangeably as auditory continuity or temporal induction, is a psychoacoustic phenomenon in which an interrupted sound stimulus is perceived as continuous when the intervening silent gap is filled with an appropriate masking sound of sufficient intensity. If the noise burst contains adequate spectral energy to have physically masked the missing segment, the human central nervous system synthesizes the missing auditory information, bridging the temporal discontinuity.
First documented systematically in the mid-twentieth century, the phenomenon encompasses both rudimentary acoustic stimuli, such as pure tones occluded by white noise, and highly complex semantic signals, such as conversational speech interrupted by extraneous acoustic disturbances. In cognitive psychology and linguistics, when speech forms the target stimulus, this process is widely recognized as the phonemic restoration effect. Across all domains, auditory induction highlights the brain’s commitment to perceptual constancy and ecological coherence over literal sensory fidelity.
Rather than registering physical silence or abrupt acoustic offsets, the auditory system employs top-down semantic expectations, Gestalt grouping principles, and bottom-up neural filling-in mechanisms to maintain unbroken auditory streams. As a consequence, listeners do not consciously perceive the interruption; instead, they hear a continuous primary signal running seamlessly through or beneath the interrupting noise source.
2. Etymology & Linguistic Origin
The term auditory derives from the Latin verb audire, meaning “to hear,” which progressed into Late Latin as auditorius, referring to the faculty or organs of hearing, and eventually entered Middle English via Old French in the late fourteenth century. The noun induction traces to the Latin inductio, a substantive form of the verb inducere, composed of the prefix in- (“into” or “upon”) and ducere (“to lead”). Historically employed in classical rhetoric and formal logic to describe the act of inferring general conclusions from particular instances, induction was adopted by nineteenth- and twentieth-century physics and neurophysiology to denote processes where one phenomenon induces or brings about a secondary state in an adjacent medium or physiological system.
In experimental psychoacoustics, the compound designation auditory induction was popularized in the early 1970s by auditory researcher Richard M. Warren and his contemporaries to classify the perceptual “induction” of sensory presence during physical absence. When applied to speech stimuli, Warren introduced the term phonemic restoration effect to signify the specific linguistic induction of missing phonemes within spoken words, anchoring the sensory concept securely in the intersection of linguistics and cognitive psychology.
3. Pronunciation & Grammatical Form
Pronunciation: The term is phonetically transcribed in the International Phonetic Alphabet (IPA) as /ˈɔː.dɪ.tər.i ɪnˈdʌk.ʃən/ in Received Pronunciation and /ˈɔː.dəˌtɔːr.i ɪnˈdʌk.ʃən/ in General American English.
Grammatical Form: Auditory induction functions grammatically as a compound noun phrase. The constituent word auditory serves as a non-gradable relational adjective modifying the abstract, uncountable noun induction. In specialized academic contexts, the term can be used in the plural (auditory inductions) when referring to discrete experimental trials or specific empirical manifestations of the phenomenon, though it most frequently appears as a mass or non-count noun describing the underlying cognitive-perceptual process.
Syntactically, the term frequently occurs in passive constructions depicting sensory experience (e.g., “Auditory induction was elicited when the silent interval was replaced with broad-spectrum noise”) or as an active nominal subject in cognitive processing descriptions (e.g., “Auditory induction facilitates speech comprehension in acoustically degraded environments”). Related lexical derivatives include the participial adjective induced (as in “induced continuity”) and the verb phrase to induce auditorily.
4. Detailed Conceptual Explanation
To fully grasp auditory induction, one must examine the fundamental challenge confronting the biological auditory apparatus: the acoustic waveform reaching the tympanic membrane is an undifferentiated composite of overlapping environmental sound sources. When an ambient event, such as a slamming door or a cough, momentarily eclipses a primary target sound, the physical energy of that target is either obscured or completely absent at the periphery. Under a strictly veridical bottom-up sensory model, the listener should register an abrupt termination of the signal, followed by a burst of extraneous noise, followed by an abrupt re-emergence of the signal. In everyday experience, however, human perception does not shatter into fragmented acoustic shards.
Auditory induction resolves this fragmentation through an illusion of perceptual continuity. For induction to occur, specific psychoacoustic criteria must be satisfied. First, the interrupting noise stimulus must possess sufficient intensity and spectral distribution such that, had the target sound actually been present during the occlusion, it would have been physically masked by the noise. If the intervening sound is softer than the target tone or occupies an incompatible frequency band, the brain detects the physical discontinuity; the silent gap becomes audible, and induction fails. This phenomenon is governed by what psychoacousticians call the “potential masker rule.”
When the criteria are met, the auditory system performs a computational evaluation. The sensory evidence does not explicitly demonstrate that the target sound ceased; rather, the evidence merely indicates that a louder sound intervened that would have rendered the target inaudible anyway. Operating under Bayesian inference and predictive processing principles, the auditory cortex favors the ecologically probable interpretation: continuous physical objects rarely vanish for twenty milliseconds only to instantaneously reappear at the exact same frequency, amplitude, and phase. Consequently, neural mechanisms generate an endogenous continuation of the occluded stimulus, maintaining perceptual stream segregation.
The scope of auditory induction spans multiple layers of neural organization. At the lowest peripheral levels, cochlear mechanics and the auditory nerve faithfully register the spectral changes. However, as ascending signals traverse the cochlear nucleus, the inferior colliculus, the medial geniculate body of the thalamus, and arrive at the primary and secondary auditory cortices, neural populations begin responding to the inferred sound as though it were physically present. Neuroimaging demonstrates that primary auditory cortex neurons exhibit continuous sustained firing throughout the occlusion, effectively writing over the sensory hiatus.
5. Historical Development
The empirical foundation of auditory induction began taking shape during the golden age of psychoacoustics in the mid-twentieth century. In 1950, George A. Miller and J. C. R. Licklider published a seminal study examining the intelligibility of speech when systematically interrupted by silence or noise. They discovered that while introducing recurring silent intervals significantly degraded intelligibility, replacing those silent gaps with continuous white noise paradoxically improved listeners’ capacity to comprehend the missing linguistic message. While Miller and Licklider characterized this primarily as an intelligibility paradox, it laid the direct groundwork for understanding perceptual filling-in.
In 1970, cognitive psychologist Richard M. Warren published his landmark paper in Science detailing the phonemic restoration effect. Warren excised a forty-millisecond segment from the recording of the word legislatures, corresponding to the first /s/ sound, and replaced the missing phoneme with a recording of an extraneous cough. When presented to listeners, nineteen out of twenty reported that the word was entirely complete and intact, with the cough perceived as co-occurring in the background. Remarkably, listeners were incapable of accurately localizing when or where the cough occurred relative to the phonemes, demonstrating that conscious perception prioritized the synthesized linguistic entity over precise chronological timing.
Throughout the 1970s and 1980s, Albert S. Bregman incorporated auditory induction into his monumental framework of auditory scene analysis. Bregman and his colleagues at McGill University demonstrated that auditory induction is not an isolated linguistic quirk, but a general structural principle of auditory perception governed by acoustic Gestalt laws. Subsequent investigations by researchers such as Daniel B. Tokunaga, Christopher J. Darwin, and Yoshitaka Nakajima delineated the precise spectral, temporal, and energetic thresholds required to trigger induction across non-verbal sounds, complex harmonic tones, and frequency-modulated chirps.
With the advent of high-resolution electrophysiology and functional neuroimaging in the late 1990s and 2000s, investigators such as Mitchell Steinschneider, Christophe Micheyl, and Maria Chait traced the physiological substrates of auditory induction. Their work demonstrated that auditory continuity is directly represented in the spatiotemporal activity patterns of Heschl’s gyrus and planum temporale, cementing auditory induction as an indispensable paradigm for studying human sensory prediction.
6. Theoretical Foundations
The primary theoretical framework explaining auditory induction is Albert Bregman’s Auditory Scene Analysis (ASA). ASA posits that the auditory system faces the inverse problem: computing the nature and location of environmental events based solely on acoustic mixtures. Bregman separated this computational task into primitive grouping mechanisms and schema-based grouping mechanisms. Primitive grouping relies on innate, bottom-up heuristics, such as harmonicity, common fate (synchronous frequency and amplitude modulation), and temporal proximity. When a sound is interrupted by noise, the primitive grouping mechanism applies the principle of closure, inferring that a single ongoing acoustic stream exists behind an occluding object, just as the visual system perceives a continuous bar passing behind an opaque pillar.
A second foundational pillar is the predictive coding framework developed within contemporary computational cognitive neuroscience. Under this paradigm, the brain is conceptualized as a hierarchical Bayesian inference engine. Higher-order cortical structures constantly generate top-down predictions regarding sensory inputs based on prior likelihood and learned environmental statistics. These predictive templates descend through the auditory hierarchy to meet incoming sensory data. When noise interrupts a predictable signal, the bottom-up sensory prediction error is suppressed because the high-amplitude noise provides ambiguous information regarding whether the target has stopped. In the absence of a contradictory prediction error, the top-down prior dominates conscious perception, resulting in the subjective experience of uninterrupted sound.
In the linguistic sphere, the theoretical architecture relies upon models of lexical access, such as the TRACE model of speech perception developed by James McClelland and Jeffrey Elman, and the Cohort model formulated by William Marslen-Wilson. These models demonstrate that speech recognition operates via bidirectional activation. When acoustic-phonetic cues are obscured, surrounding phonemic context, lexical constraints, and syntactic expectancies activate candidates within the mental lexicon. This top-down activation feeds back down to pre-lexical processing levels, directly generating the perceptual restoration of missing phonetic features.
7. Key Components, Types & Dimensions
Auditory induction manifests across a spectrum of acoustic and cognitive contexts, characterized by distinct dimensions, operational types, and structural components:
- Temporal Continuity Induction (Tonal Induction): The baseline psychoacoustic manifestation wherein an elementary target sound, such as a sinusoidal pure tone or narrow-band noise, is cyclically alternated with an interrupting broadband noise burst. If the noise burst satisfies energetic masking thresholds, the pure tone is heard as a steady, uninterrupted pitch.
- Phonemic Restoration (Linguistic Induction): The high-order cognitive manifestation wherein phonemes, syllables, or morphemes excised from natural running speech are perceptually reconstituted when replaced with non-speech maskers (e.g., coughs, buzzes, or pink noise), drawing upon syntactic, semantic, and pragmatic context.
- Harmonic and Spectral Induction: The perceptual synthesis of missing spectral components within complex acoustic waveforms. This includes the “missing fundamental” phenomenon, where the perceived pitch of a complex tone corresponds to a fundamental frequency that has been physically removed, reconstructed via the presence of integer harmonic overtones.
- Melodic and Musical Induction: The continuity of melodic lines, rhythmic pulses, and musical timbres across interruptions. Musicians and experienced listeners seamlessly bridge musical gaps when occluded by percussive transients or audience noise, heavily mediated by tonal syntax and musical schemas.
- Acoustic Masker Compatibility: A governing dimension dictated by the physical attributes of the interrupting stimulus. Effective induction requires that the masker possess sufficient amplitude (energetic masking) and overlap the critical auditory filter band of the occluded target (spectral overlap).
- Temporal Duration Limits: The biological bounds of the continuity illusion. While induction functions robustly across brief intervals (typically 20 to 200 milliseconds), it progressively collapses as the occluding interval exceeds 300 to 500 milliseconds, at which point the auditory system flags the temporal hiatus as genuine physical cessation.
8. Examples & Illustrative Cases
A classic experimental demonstration of auditory induction involves a pure 1,000 Hz tone that alternates every 100 milliseconds with a 100-millisecond silent gap. Under these conditions, the listener hears an unmistakable, staccato series of isolated beeps separated by silence. If the silent intervals are subsequently replaced with bursts of broad-spectrum white noise played at an amplitude ten decibels higher than the tone, the perceptual experience shifts instantly. The intermittent beeping vanishes entirely; in its place, the subject hears an unbroken, steady 1,000 Hz tone sounding continuously in the background, while the bursts of noise appear to pulsate in the foreground.
In an illustrative linguistic case, consider the classical sentence designed by Richard Warren: “The state governors met with their respective legi*latures convening in the capital.” Here, the asterisk represents a complete acoustic excision of the /s/ phoneme, replaced by a loud burst of static or a digitized cough. Naive listeners invariably report hearing the word legislatures in its entirety. When questioned, subjects frequently hypothesize that the speaker coughed in the room while speaking, unable to discern that the speech stream was physically mutilated.
Contextual semantic modulation provides another striking experimental demonstration, as shown by Arthur Wingfield and colleagues. Consider the following carrier sentences where the final word contains an occluded initial consonant cluster (*ill):
- Sentence A: “There was time to spare, so the driver checked the orange *ill.” (The listener perceptually restores the word pill).
- Sentence B: “There was time to spare, so the painter climbed the high *ill.” (The listener perceptually restores the word hill).
- Sentence C: “There was time to spare, so the banker examined the dollar *ill.” (The listener perceptually restores the word bill).
In each scenario, the physical acoustic input at the obscured locus is entirely identical—a burst of identical noise. Yet, the listener’s subjective perceptual awareness constructs three completely distinct phonetic segments, guided entirely by semantic constraints embedded earlier in the sentence frame.
9. Measurement & Assessment
Assessing auditory induction requires psychoacoustic paradigms that quantify both subjective awareness and objective neurophysiological signatures. In classic psychophysical measurement, researchers employ the method of constant stimuli or adaptive staircase procedures to determine the “continuity threshold.” Listeners are presented with alternating target-noise sequences and must perform a two-alternative forced-choice (2AFC) task, deciding whether the target sound is “continuous” or “interrupted.” By systematically modulating the signal-to-noise ratio (SNR) and gap durations, experimenters plot psychometric functions tracking the transition from perceived discontinuity to robust induction.
To bypass subjective response biases, psychoacousticians utilize objective performance measures, such as gap detection thresholds. If auditory induction is functionally genuine, an occluded gap should be rendered inaccessible to the listener. If a miniature target-state change (such as an imperceptible phase shift or frequency modulation) is embedded directly within the noise-filled interval, the occurrence of auditory induction can paradoxically impair the listener’s ability to detect this localized modification, providing objective proof of sensory neural completion.
In clinical and cognitive neuroscience laboratories, electroencephalography (EEG) and magnetoencephalography (MEG) track the physiological correlates of auditory induction. The Auditory Evoked Potential (AEP) components, particularly the P1-N1-P2 complex, provide quantitative metrics of sensory encoding. When an interruption is perceived as continuous, the abrupt cortical offset- and onset-responses (N1-P2 complexes) that typically mark the boundaries of a silent gap are substantially attenuated or completely absent, mirroring the waveform observed during an uninterrupted physical stimulus. Furthermore, the Mismatch Negativity (MMN)—a pre-attentive electrophysiological index of acoustic deviation—is triggered when a continuous sound unexpectedly drops out beneath noise, demonstrating that the primary auditory cortex actively expects signal preservation.
10. Applications & Practical Significance
Auditory induction has direct practical ramifications across clinical audiology, hearing aid development, telecommunications, and speech recognition technology. In telecommunications and voice-over-IP (VoIP) protocols, packet loss creates dropouts that compromise communication clarity. Engineers design packet loss concealment (PLC) algorithms that mimic the biological mechanisms of auditory induction. By inserting synthetic comfort noise or mathematically extrapolating spectral parameters across dropped frames, digital signal processors induce the human listener’s auditory system to fill in the missing speech information, vastly improving perceived quality without expanding bandwidth requirements.
In audiological medicine and hearing assistive technology, auditory induction explains why individuals with mild-to-moderate sensorineural hearing loss experience catastrophic communication breakdowns in noisy social settings, such as restaurants. Sensorineural damage degrades spectral resolution and cochlear tuning curves. Because auditory induction depends on precise frequency-to-space mapping to determine whether an occluding noise meets the potential masker rule, damaged peripheral filters impair temporal continuity processing. Consequently, modern hearing aids incorporate advanced directional microphones and dynamic noise reduction algorithms specifically tailored to preserve the acoustic contrasts required for the central auditory system to execute natural phonemic restoration.
In everyday human environments, auditory induction represents a critical evolutionary adaptation. In the natural world, auditory lines of sight are routinely interrupted by wind, rustling foliage, animal vocalizations, and weather phenomena. Auditory induction endows organisms with perceptual resilience, allowing predators and prey alike to maintain uninterrupted acoustic surveillance on salient environmental targets without continuous, pristine sensory input.
11. Research & Empirical Evidence
Decades of empirical studies have rigorously mapped the behavioral and neurological boundaries of auditory induction. A landmark study by Riecke, van Atteveldt, Formisano, and colleagues used functional Magnetic Resonance Imaging (fMRI) combined with sophisticated behavioral tracking to examine cortical responses during auditory continuity. Their neuroimaging data revealed that when listeners reported the illusion of a continuous tone through noise, activity within the primary auditory cortex (A1) and adjacent superior temporal gyrus mirrored the sustained activation observed during physically continuous tones, validating that auditory induction is not a mere post-perceptual cognitive judgment, but an authentic sensory restoration.
In cellular neurophysiology, invasive multi-unit recordings in animal models (such as non-human primates and cats) conducted by researchers including Mitchell Steinschneider and colleagues have traced how single neurons in the auditory cortex behave during occluded intervals. Neurons tuned to the specific frequency of the uninterrupted baseline tone sustain their characteristic firing rates straight through the masking noise burst, provided the noise contains energy within the neuron’s receptive field. When the identical noise burst is presented alone without the flanking tone, the firing pattern drops significantly, revealing that top-down predictive persistence actively biases peripheral inputs in real time.
Research conducted by Matthew Leonard and colleagues at the University of California, San Francisco, utilized direct high-density electrocorticography (ECoG) in human surgical patients during phonemic restoration tasks. Their findings demonstrated that neural populations in the superior temporal gyrus (STG) reconstruct missing speech sounds within hundreds of milliseconds following noise onset. Intriguingly, the neural representation of the restored phoneme aligned with the neural signature of the physical phoneme, with top-down activity originating in frontal linguistic areas directing the acoustic-phonetic reconstruction occurring within auditory receptive fields.
12. Cultural & Cross-Cultural Considerations
While the basic psychoacoustic architecture of auditory induction—such as the continuity of pure tones through white noise—is a human universal rooted in the biophysical evolution of the vertebrate auditory pathway, linguistic auditory induction (phonemic restoration) exhibits intriguing cross-cultural and cross-linguistic variations. These variations are driven by the disparate phonological structures, tonal systems, and morphological rules characterizing the world’s languages.
In tonal languages such as Mandarin Chinese, Vietnamese, or Yoruba, lexical meaning is governed not merely by segmental phonemes (consonants and vowels) but by pitch contours (lexical tones). Research evaluating phonemic restoration in native Mandarin speakers indicates that the auditory system must restore both segmental phonetic information and dynamic fundamental frequency (F0) trajectories simultaneously. When noise interrupts a Mandarin syllable, the linguistic restoration mechanism relies heavily on the overarching sentential pitch contour to infer the missing tone, demonstrating that top-down semantic constraints must coordinate tightly with musical-pitch processing networks in tonal language speakers.
Cross-linguistic studies also reveal distinctions linked to linguistic redundancy and phonotactic constraints. Languages with high degrees of inflectional morphology and strict phonotactic rules, such as Russian or German, constrain possible lexical completions more aggressively than languages with simpler syllabic structures, such as Japanese. Consequently, the speed and behavioral efficacy of phonemic restoration vary in accordance with how narrow the statistical probability matrix is for a missing phoneme given its surrounding phonetic neighborhood. Despite these structural differences, the fundamental propensity of the human brain to reconstruct missing communicative signals across acoustic gaps remains a universal cognitive constant.
13. Criticisms, Debates & Limitations
Despite its broad acceptance, auditory induction has generated notable debates within psychoacoustics and the philosophy of perception. One prominent theoretical controversy centers on the “perceptual vs. post-perceptual” debate. Early skeptics argued that phonemic restoration and auditory continuity do not represent true sensory illusions occurring within early auditory processing; instead, they proposed that the phenomenon is an artifact of retrospective memory distortion or post-perceptual decision-making. Under this view, listeners do not genuinely *hear* the missing sound; rather, they rapidly infer what the sound must have been and mistakenly report having heard it due to response biases.
Extensive neurophysiological and electrophysiological evidence has largely refuted the extreme post-perceptual position by demonstrating early cortical activation in sensory areas within 100 milliseconds of interruption. Nonetheless, the precise point of division between bottom-up sensory filling-in and top-down cognitive judgment remains contentious. Researchers continue to debate whether auditory induction in complex sentences is driven strictly pre-attentively or whether it requires conscious, top-down attention and working memory reserves.
Another significant limitation of auditory induction research lies in ecological validity. Most historical investigations have relied on highly artificial stimuli: static pure tones occluded by rectangular bursts of Gaussian noise presented through closed-back headphones in sound-isolated acoustic chambers. In natural environments, target sounds and interrupting maskers are rarely static or neatly separated; environmental acoustics involve reverberation, spatial movement, head-related transfer functions, and dynamic amplitude fluctuations. Critics argue that laboratory paradigms may overestimate the robustness of auditory induction in everyday life, where real-world maskers frequently fail to fulfill the strict psychoacoustic criteria required for seamless continuity.
14. Related Terms & Distinctions
To avoid conceptual ambiguity, auditory induction must be distinguished from several related auditory and perceptual phenomena:
- Auditory Induction vs. Auditory Scene Analysis (ASA): Auditory scene analysis is the broad conceptual framework describing how the auditory system segments composite acoustic environments into discrete sound objects. Auditory induction is an operational mechanism *within* ASA that handles temporal occlusion using Gestalt closure.
- Auditory Induction vs. Auditory Masking: Auditory masking occurs when the perception of one sound is attenuated or completely obliterated by the presence of another sound. While auditory induction requires the *potential* for masking to function, induction represents the perceptual *preservation* of the occluded sound, whereas masking denotes its perceptual elimination.
- Auditory Induction vs. Phonemic Restoration: Phonemic restoration is a specialized, linguistic subcategory of auditory induction. While auditory induction describes the general psychoacoustic continuity of any sound (including tones, noise bands, and environmental noises), phonemic restoration specifically denotes the reconstruction of missing speech phonemes mediated by lexical and semantic knowledge.
- Auditory Induction vs. The McGurk Effect: The McGurk effect is a cross-modal illusion where conflicting visual speech information (lip movements) alters the auditory perception of a spoken syllable. In contrast, auditory induction does not require multisensory visual input; it is fundamentally an intra-auditory or auditory-cognitive constructive process.
- Auditory Induction vs. Auditory Preattentive Mismatch: Mismatch responses (such as the MMN) index the brain’s detection of unexpected physical transitions. Auditory induction prevents or dampens this mismatch response by generating an endogenous continuation template that maintains stream integrity.
15. Summary & Key Takeaways
Auditory induction stands as a foundational concept in sensory psychology, psychoacoustics, and cognitive neuroscience. It demonstrates that auditory perception is an active, predictive inference rather than a passive, literal recording of mechanical sound waves. By leveraging Gestalt grouping heuristics, hierarchical Bayesian predictive coding, and semantic context, the brain ensures continuity in a noisy world.
The key takeaways regarding auditory induction include:
- The Illusion of Continuity: When an interrupted sound is occluded by an acoustic stimulus with sufficient spectral energy and amplitude (the potential masker rule), listeners experience the interrupted sound as unbroken and continuous.
- Linguistic Relevance: In speech perception, this mechanism manifests as the phonemic restoration effect, where listeners subconsciously fill in missing phonemes using contextual, lexical, and syntactic cues.
- Cortical Reality: Functional neuroimaging and intracranial recordings demonstrate that auditory induction is accompanied by authentic, sustained neural firing within primary and association auditory cortices, proving it is a genuine sensory synthesis rather than a mere retrospective cognitive guess.
- Clinical and Technological Utility: Understanding the rules governing auditory induction informs packet loss concealment in digital communications, noise reduction algorithms in hearing prosthetics, and rehabilitative approaches for patients with central auditory processing disorders.
Ultimately, auditory induction illustrates the remarkable adaptive capacity of human perception. Faced with an imperfect, noisy sensory environment, our brains draw upon statistical regularities, prior knowledge, and acoustic cues to transform fragmented acoustic streams into a coherent, continuous conscious reality.
References
- Bregman, A. S. (1990). Auditory scene analysis: The perceptual organization of sound. MIT Press. https://mitpress.mit.edu/9780262521956/auditory-scene-analysis/
- Leonard, M. K., Baud, M. O., Sjerps, M. J., & Chang, E. F. (2016). Perceptual restoration of masked speech in human superior temporal gyrus. Nature Communications, 7, Article 13619. https://www.nature.com/articles/ncomms13619
- Miller, G. A., & Licklider, J. C. R. (1950). The intelligibility of interrupted speech. The Journal of the Acoustical Society of America, 22(2), 167–173. https://doi.org/10.1121/1.1906584
- Riecke, L., van Atteveldt, N., & Formisano, E. (2007). Listening to sound continuity: How the brain generates illusory auditory perception. Journal of Neuroscience, 27(46), 12684–12689. https://doi.org/10.1523/JNEUROSCI.2940-07.2007
- Warren, R. M. (1970). Perceptual restoration of missing speech sounds. Science, 167(3917), 392–393. https://doi.org/10.1126/science.167.3917.392