Interacting Brain Networks
Human (non-randomized) evidence
What is it?
This hypothesis says tinnitus is not produced by one damaged spot. The sound you hear is an emergent property of several brain networks running in parallel: a minimal 'tinnitus core' that makes the sound conscious at all, plus separate networks that add its loudness, which ear it seems to come from, and the distress it causes. These networks talk to each other at shared hubs, each using its own brain rhythm. Underneath it, loss of input shifts the thalamus, cortex loop out of its normal idling rhythm, and the brain fills the missing information in from memory and prediction, which is what gets heard. Strikingly, a machine-learning classifier found the same rhythm signature across tinnitus, pain, Parkinson's and depression: the same mechanism, differing mainly in where it sits. Fear and distress are not a side effect bolted onto an otherwise purely auditory signal: connectivity from the amygdala (the brain's fear hub) to the auditory cortex is abnormally strong in tinnitus and tracks directly with how distressing it feels, while the Default Mode Network and the insula-centered Salience Network both show altered connectivity that scales with tinnitus severity, naming the specific networks behind why tinnitus so often brings anxiety and unshakeable attention along with it.
- Tinnitus coreMinimal network needed for conscious perception
- Loudness networkAdds perceived intensity
- Laterality networkWhich ear it seems to come from
- Distress networkAdds the emotional suffering
- Shared hubs synchronizeBinding the networks into one unified percept
Mechanism
Bottom-up deafferentation and/or a top-down noise-cancelling deficit alter auditory thalamocortical transmission, producing thalamocortical dysrhythmia (TCD): resting-state alpha slows to theta while surrounding gamma increases, leaving persistent theta, gamma cross-frequency coupling. Theta burst-firing raises network synchrony and recruitment, enabling the long-range synchrony that can carry focal auditory gamma activity into a consciousness-enabling ('global workspace') network. When deafferentation is narrow, the missing information is retrieved from neighbouring auditory cortex via decreased surround inhibition; when it is broad, from theta-mediated parahippocampal auditory memory. In Bayesian terms the brain is a prediction machine: tinnitus is the prediction error between predicted and sensed auditory input, with reduced sensory updating (less alpha) and the error expressed as theta, gamma and beta, gamma coupling. The unified percept then emerges from multiple parallel, partially overlapping subnetworks communicating at hubs at discrete frequencies, with a minimal 'tinnitus core' required for conscious perception, separable from the affective components. The 'tinnitus core' and its surrounding subnetworks are not purely auditory: fear- and distress-related circuitry is directly implicated. Amygdala-to-auditory-cortex connectivity is abnormally enhanced in tinnitus, and its strength correlates directly with how distressing the tinnitus feels, while hippocampus-to-auditory-cortex connectivity tracks how long the tinnitus has persisted, consistent with the amygdala driving moment-to-moment distress and the hippocampus encoding it into a lasting memory trace. A related theoretical account frames tinnitus itself as a persisting 'aversive memory' network, a failure to extinguish the brain's memory of lost sensory input, rather than a purely auditory signal. A separate limbic-gating model proposes that a subcallosal-area/nucleus-accumbens circuit normally filters phantom auditory signals out of awareness, and that chronic, bothersome tinnitus reflects a failure of this gate. Named large-scale networks are also directly implicated: enhanced connectivity within the Default Mode Network (hubs including the posterior cingulate gyrus and parahippocampus, normally active during rest and self-referential thought) and enhanced connectivity between the auditory cortex and the insula, a core hub of the Salience Network, which flags signals as worth attending to, both track tinnitus severity/handicap, with the insula proposed as the specific hub linking auditory processing, salience detection, and the default-mode network in more severe, more distressing tinnitus.
Why would tinnitus need several brain networks instead of one damaged spot?
Tinnitus is felt as one thing, but it is really a bundle of separable properties: how loud it is, which ear it seems to come from, whether it is a tone or a hiss, and how distressing it feels. Two people can have the same loudness and completely different distress. A single damaged location cannot easily account for that. The model instead proposes a minimal 'tinnitus core', the smallest set of brain areas that must fire together for you to consciously hear anything at all, with additional, partly overlapping networks layering on the loudness, the location and the emotional suffering. They exchange information at shared hub regions, each network using its own rhythm, which is how separate parts end up experienced as one unified sound.
What is thalamocortical dysrhythmia, in plain terms?
The thalamus is the relay that feeds sensory information to the cortex, and the loop between them normally idles at a rhythm called alpha. When the ear stops delivering its usual input, that idling slows down into a slower theta rhythm, and around it, fast gamma activity increases. The slow and fast rhythms then lock together (cross-frequency coupling), a pattern that should not persist. Those slow theta bursts recruit ever-wider stretches of brain, and that recruitment is thought to be how a small patch of abnormal auditory activity gets promoted into something you consciously hear rather than something the brain quietly ignores.
What does 'the brain is a prediction machine' mean here?
Under the Bayesian view, your brain is constantly predicting what it is about to hear and correcting those predictions with what the ears actually deliver. When hearing is damaged, the correction signal weakens, so the brain leans on its predictions and fills the gap from neighbouring frequencies, or, if the gap is wide, from stored auditory memory. What you perceive is the mismatch: the difference between what the brain expected and what it actually received. That is why the model treats tinnitus as a prediction error rather than as a sound being generated from scratch.
How do we know this rhythm pattern is real and not just theory?
The pattern was tested rather than assumed. Researchers fed resting-state EEG recordings from patients with four different conditions, Parkinson's disease, neuropathic pain, tinnitus and depression, into a machine-learning classifier, without telling it what to look for. It found the same *spectral* dysrhythmia signature in all four, but located in *different* brain areas depending on the disorder, plus a set of regions common to all of them. That is meaningful because a purely theoretical construct would not be expected to fall out of the data this cleanly, and it explains why one oscillatory mechanism can underlie several very different phantom conditions.
Why does tinnitus so often come with anxiety and distress, not just sound?
Because tinnitus is so tightly linked to emotional distress, researchers tested whether the amygdala, the brain's central fear-processing hub, is abnormally wired into the auditory system. Using resting-state fMRI, they found abnormal directional connectivity FROM the amygdala TO the auditory cortex in tinnitus patients, and the strength of that connection tracked almost exactly with how distressing people rated their tinnitus. A related connection from the hippocampus (the brain's memory hub) to auditory cortex tracked how long the tinnitus had lasted, consistent with the amygdala driving in-the-moment distress and the hippocampus locking that distress into a lasting memory.
What is the Default Mode Network and the Salience Network, and why do they matter for tinnitus?
The Default Mode Network is a set of brain regions (including the posterior cingulate gyrus and parahippocampus) that are normally active when the mind is at rest or turned inward, not focused on the outside world. The Salience Network, centered on a region called the insula, is what flags a signal, internal or external, as important enough to demand attention. An EEG study found that both networks show abnormal connectivity in people with more severe, more handicapping tinnitus, with the insula acting as a hub connecting the auditory cortex to both networks. In plain terms: the more the brain's 'this matters, pay attention' network and its 'turn inward' network get pulled into processing the phantom sound, the worse tinnitus tends to feel.
Could tinnitus be a memory that never fades, rather than a sound the brain keeps generating?
One theoretical account reframes tinnitus (and chronic pain) as a kind of aversive memory that fails to extinguish. Normally, the brain learns to stop reacting to a stimulus that carries no new information, but if a piece of sensory input is suddenly missing (as after hearing damage) rather than merely repeated, the brain may instead keep 're-presenting' it through memory-related circuitry, the way an unresolved memory keeps intruding. A separate model proposes a specific limbic gate, a circuit including the subcallosal area and nucleus accumbens, both involved in emotional valuation, that normally filters phantom signals out before they reach awareness, and that chronic tinnitus reflects this gate failing to close.
Evidence (9)
An integrative model of auditory phantom perception: tinnitus as a unified percept of interacting separable subnetworks.
Neuroscience and biobehavioral reviews · 2014
The flagship model: tinnitus is not one damaged locus but an emergent percept of multiple parallel brain subnetworks, communicating at shared hubs at distinct oscillatory frequencies.
Thalamocortical Dysrhythmia: A Theoretical Update in Tinnitus.
Frontiers in neurology · 2015
Theoretical update on thalamocortical dysrhythmia: after input loss the thalamus, cortex loop slows from alpha to theta with surrounding gamma, and that theta, gamma coupling carries the phantom sound into consciousness.
Thalamocortical dysrhythmia detected by machine learning.
Nature communications · 2018
Human evidence: machine learning on resting EEG detected a thalamocortical dysrhythmia signature that is spectrally equivalent but spatially distinct across tinnitus, pain, Parkinson's and depression.
Predictive coding and stochastic resonance as fundamental principles of auditory phantom perception.
Brain : a journal of neurology · 2023
Review: tinnitus and other auditory phantom percepts may reflect the brain's own predictive-coding machinery amplifying random neural noise (stochastic resonance) into a false percept.
Phantom percepts: tinnitus and pain as persisting aversive memory networks.
Proceedings of the National Academy of Sciences of the United States of America · 2011
Theoretical model proposing tinnitus and chronic pain are persisting 'aversive memory' networks, engaging parahippocampal/memory circuitry alongside the sensory pathway.
Tuning out the noise: limbic-auditory interactions in tinnitus.
Neuron · 2010
Proposes a limbic (subcallosal area / nucleus accumbens) gate that normally filters out the tinnitus signal before it reaches consciousness, and that chronic tinnitus reflects a failure of this gate.
Tinnitus distress is linked to enhanced resting-state functional connectivity from the limbic system to the auditory cortex.
Human brain mapping · 2017
Human fMRI: connectivity FROM the amygdala TO auditory cortex is abnormally enhanced in tinnitus and correlates directly with distress and duration, a measurable fear/emotion-to-hearing pathway.
Abnormal Functional Connectivity Within Default Mode Network and Salience Network Related to Tinnitus Severity.
Journal of the Association for Research in Otolaryngology : JARO · 2023
Human EEG study naming the exact networks: connectivity within the Default Mode Network and Salience Network (via the insula) tracks tinnitus severity/handicap.
EEG microstates and dynamic functional connectivity reveal stage-specific brain networks in subjective tinnitus.
iScience · 2026
Small human EEG study finds distinct brain-network signatures for acute vs. chronic tinnitus, an early salience/executive imbalance replaced by a more stable, low-frequency compensatory pattern as tinnitus becomes chronic.
Timeline
- 2010
Proposes a limbic (subcallosal area / nucleus accumbens) gate that normally filters out the tinnitus signal before it reaches consciousness, and that chronic tinnitus reflects a failure of this gate.
- 2011
Theoretical model proposing tinnitus and chronic pain are persisting 'aversive memory' networks, engaging parahippocampal/memory circuitry alongside the sensory pathway.
- 2014
The flagship model: tinnitus is not one damaged locus but an emergent percept of multiple parallel brain subnetworks, communicating at shared hubs at distinct oscillatory frequencies.
- 2015
Theoretical update on thalamocortical dysrhythmia: after input loss the thalamus, cortex loop slows from alpha to theta with surrounding gamma, and that theta, gamma coupling carries the phantom sound into consciousness.
- 2017
Human fMRI: connectivity FROM the amygdala TO auditory cortex is abnormally enhanced in tinnitus and correlates directly with distress and duration, a measurable fear/emotion-to-hearing pathway.
- 2018
Human evidence: machine learning on resting EEG detected a thalamocortical dysrhythmia signature that is spectrally equivalent but spatially distinct across tinnitus, pain, Parkinson's and depression.
- 2023
Review: tinnitus and other auditory phantom percepts may reflect the brain's own predictive-coding machinery amplifying random neural noise (stochastic resonance) into a false percept.
- 2023
Human EEG study naming the exact networks: connectivity within the Default Mode Network and Salience Network (via the insula) tracks tinnitus severity/handicap.
- 2026
Small human EEG study finds distinct brain-network signatures for acute vs. chronic tinnitus, an early salience/executive imbalance replaced by a more stable, low-frequency compensatory pattern as tinnitus becomes chronic.
- Is the 'tinnitus core' subnetwork the same across patients, or does each person have a different core?
- Thalamocortical dysrhythmia also appears in Parkinson's, pain and depression, what makes the tinnitus form specifically auditory?
- Can normalising the thalamocortical rhythm abolish the percept itself, rather than only the distress that accompanies it?
- If amygdala-to-auditory-cortex connectivity tracks distress, could directly targeting that connection (rather than the auditory pathway) relieve tinnitus-related suffering without changing the perceived loudness?
- Are the Default Mode Network and Salience Network changes a cause of tinnitus distress, or a consequence of living with a persistent phantom sound?