The mismatch negativity, abbreviated as MMN, is an event-related potential component derived from electroencephalogram (EEG) recordings. Unlike many evoked potentials that require active attention, the MMN emerges automatically when an infrequent “deviant” sound violates a sequence of repeated “standard” sounds. This automatic quality makes it particularly valuable as a clinical and research tool, because it removes the confounding variable of whether a patient is cooperating or maintaining focus during testing.
What Is the Mismatch Negativity?
Mismatch negativity (MMN) is an event-related potential component recorded via electroencephalogram (EEG) that reflects the brain’s automatic, pre-attentive detection of auditory changes. This automatic response is typically evoked by presenting a sequence of identical standard tones, followed by an infrequent 'deviant' tone that differs in frequency, duration, or other acoustic features.
The brain detects this deviation, generating a negative-going voltage deflection that peaks approximately 100 to 200 milliseconds after the change. Recorded most prominently over frontal and central scalp electrodes, this characteristic waveform serves as the primary indicator of the mismatch negativity.
Two properties distinguish the MMN from other brain responses:
First, it is automatic. A person can be reading a book, watching a silent film, or actively ignoring the sounds entirely, and the MMN will still appear.
Second, it is pre-attentive. The brain performs the comparison between the deviant sound and the memory trace of the standard sounds before attention ever engages.
These characteristics make the MMN an index of the brain's fundamental capacity to detect change, independent of higher cognitive processes that might be impaired or variable across individuals.
The magnetic counterpart of the MMN, recorded with magnetoencephalography and abbreviated MMNm, represents the same underlying auditory change-detection process. Because MEG offers different spatial sensitivity than EEG, combining both recording modalities help researchers separate contributions from different brain regions. The MMNm localizes to auditory cortex and surrounding temporal lobe structures, confirming that the response originates in sensory processing areas rather than higher associative regions.
How the Brain Generates the Mismatch Negativity
MMN generation is driven by complex, dynamic neuronal mechanisms in the auditory cortex, which researchers are actively exploring through sophisticated neurocomputational models. While there is no single cellular-level consensus on exactly what occurs when a deviant sound triggers this response, two major hypotheses have shaped decades of scientific inquiry: the adaptation hypothesis and the model adjustment hypothesis.
The adaptation hypothesis suggests that neurons in the auditory cortex respond to repeated standard sounds by gradually reducing their firing rates—a process known as stimulus-specific adaptation. When a rare deviant sound is introduced, it activates a separate, fresh population of non-adapted neurons. Under this view, the MMN is essentially a differential response between adapted and non-adapted neural populations.
Conversely, the model adjustment hypothesis proposes an active cognitive mechanism rather than passive neural fatigue. It posits that the brain continuously constructs a predictive internal model of its auditory environment. When standard sounds repeat, they reinforce this expectation; when a deviant sound occurs, it violates the expectation and forces the brain to update its internal representation. The MMN waveform, therefore, reflects this active model-updating process.
The Predictive Coding Framework and Spiking-Neuron Models
To reconcile these competing accounts, researchers increasingly look to predictive coding as a unifying framework. Predictive coding views the brain as an active prediction engine that anticipates sensory inputs to process information more efficiently.
The auditory cortex learns statistical regularities from preceding sounds to generate predictions. When incoming stimuli match these expectations, processing proceeds smoothly; when a sound deviates, a prediction error propagates through the cortical hierarchy. As a result, the MMN is the electrophysiological signature of a prediction error signal.
A detailed, biologically plausible spiking-neuron model of the auditory cortex supports this framework. This model structures excitatory and inhibitory neurons into a layered cortical architecture containing distinct functional processing units:
Input units that receive raw sensory data.
Prediction units that represent learned environmental expectations.
Prediction-error units that signal mismatches when inputs and predictions conflict.
By incorporating a synaptic learning rule dependent on NMDA receptor transmission, this network dynamically adjusts its predictions based on the transition statistics of recent sounds.
This computational model successfully accounts for several key, empirically observed MMN characteristics:
Frequency-dependent scaling: The MMN amplitude increases proportionally with how much the deviant sound differs from the standard tone.
Sensitivity to local patterns: The model reproduces responses to unexpected repetitions in alternating sequences (e.g., registering a prediction error when an alternating ABABAB sequence suddenly changes to ABABAA), reflecting a reliance on local transition probabilities rather than global statistical rules.
Omission responses: The model generates a prediction error even when an expected sound is completely omitted.
Pharmacological sensitivity: The model is highly sensitive to NMDA receptor antagonists, aligning with clinical studies showing that blocking these receptors reduces or eliminates MMN amplitude.
Moreover, a complementary magnetoencephalography (MEG) experiment validated these model claims, showing that MMN responses stem from active cortical predictions rather than simple synaptic habituation, providing strong empirical support for the predictive coding account.
Generative Models and Bayesian Inference in Auditory Processing
Another prominent neurocomputational model approaches the auditory cortex through the lens of Bayesian inference. This neuroscientific framework models the brain as continually updating a probabilistic generative representation of the sensory world using generalized Bayesian filtering.
In this system, the MMN represents the collective electrical activity generated by prediction-error neurons as they perform inference on the hidden states of hierarchical dynamics. This Bayesian approach produces highly realistic MMN waveforms that mirror empirical data, accurately capturing how deviant probability and magnitude shape the response:
As a deviant sound becomes more probable, the expectation violation decreases, reducing MMN amplitude.
As the deviant becomes less acoustically distinct from standards, the latency of the MMN response increases while its amplitude falls.
Collectively, both predictive coding and Bayesian models converge on the same fundamental truth: the MMN is the electrical signature of a brain actively forecasting its auditory environment and signaling when sensory reality conflicts with internal expectations.
Hypothesis | Mechanism | Core Idea |
|---|---|---|
Adaptation | Neurons fatigue from repeated sounds | Differential response between adapted populations |
Model adjustment | Brain actively updates internal model | MMN reflects expectation violation |
Predictive coding | Brain predicts upcoming sensory input | MMN signals prediction error |
MMN as a Biomarker in Schizophrenia
Since the early 1990s, studies have documented reduced MMN amplitude in patients with schizophrenia compared to healthy individuals. This attenuation appears to be robust and systematic, appearing across laboratories, recording protocols, and patient populations.
A conventional auditory oddball paradigm reveals significant deficits in both duration-deviant and frequency-deviant MMN in schizophrenia patients. This impairment is so reliable that MMN is widely proposed as a clinical biomarker—an objective, physiological measure capable of indexing disease progression and testing therapeutic efficacy.
Different acoustic deviants track distinct dimensions of the pathology:
Frequency-Deviant MMN: The amplitude reduction worsens as the illness progresses, closely correlating with illness duration and cognitive and functional decline. This makes frequency-deviant MMN a valuable marker of disease progression.
Duration-Deviant MMN: Deficits in this area follow a different trajectory and are more strongly linked to genetic vulnerability, serving as a trait marker that remains relatively stable over time.
Additionally, the distinct neural generators contributing to MMN are differentially affected in schizophrenia:
Temporal Lobe Generators: Impairments here are linked to deficits in basic auditory perception and discrimination, which can contribute to positive symptoms like auditory hallucinations.
Frontal Lobe Generators: Reduced activity in the frontal cortex is tied to impaired attention-switching, correlating with negative symptoms such as social withdrawal and flat affect.
Neuronal Adaptation vs. Deviance Detection in Schizophrenia
A critical question is whether the MMN deficit in schizophrenia stems from a failure of sensory adaptation or a failure of active deviance detection. In a standard oddball paradigm, these two processes are confounded because the brain's response to the deviant tone depends on both how it adapts to the repeating standard and how it detects the novel change.
To isolate these mechanisms, Koshiyama et al. utilized a many-standards control paradigm. In this design, multiple tones of varying frequencies are presented with equal probability so that no single tone repeats enough to trigger adaptation. The deviant tone is then compared to this varied background, isolating deviance detection from adaptation effects.
When patients with schizophrenia were evaluated using both paradigms, the results revealed a selective impairment in deviance detection. While their neural adaptation to repeated tones remained entirely intact, their ability to detect acoustic changes and generate a prediction error in response to an altered environment was significantly compromised.
Methodological Best Practices for Eliciting Mismatch Negativity
The core principle of recording a clean, interpretable MMN is to present an auditory regularity (a repeating standard sound) and occasionally violate that regularity with an infrequent deviant sound. How this violation is constructed dictates what aspect of neural processing can be evaluated.
1. Core Principles: Standard Oddball Paradigm
The standard oddball paradigm relies on three key stimulus parameters that systematically shape the resulting MMN response:
Deviant Type: Deviants can differ from standards across various acoustic dimensions, such as frequency or duration. Including both frequency and duration deviants is often needed in clinical research because they reflect different dimensions of neural pathology and disease progression.
Deviant Probability: Lower-probability deviants generate larger MMN amplitudes because they represent a more unexpected break in environmental regularity.
Deviant Magnitude: Larger acoustic differences between standard and deviant sounds elicit greater MMN amplitudes and shorter response latencies.
2. Distinguishing Processes: Oddball vs. Many-Standards Paradigm
A single-standard oddball paradigm inherently confounds two separate neural processes: passive adaptation to the repeating standard and active detection of the deviant.
To isolate deviance detection, researchers use the many-standards paradigm. By presenting a wide variety of tone frequencies at equal probabilities, adaptation is distributed evenly across the stimulus set.
Comparing oddball and many-standards paradigms within the same experiment help researchers to determine whether an MMN abnormality stems from impaired adaptation, impaired deviance detection, or a combination of both.
3. Testing Active Prediction: Beyond the Standard Oddball
Evaluating predictive coding accounts requires designs that go beyond simple physical feature changes:
Pattern Violations: Presenting repeating patterns (e.g., ABABAB) establishes a prediction of alternation. An unexpected repetition (e.g., ABABAA) triggers an MMN, demonstrating detection of complex structural rules.
Omission Paradigms: When a predictable sound in a rhythmic sequence is omitted entirely, the brain still generates a prediction error signal, proving that MMN reflects active expectation rather than mere sensory habituation.
4. Source Localization: EEG vs. MEG
While standard EEG provides robust surface recordings, pinpointing the specific neural generators of the MMN often requires MEG. MEG records the magnetic counterpart (MMNm) and offers superior spatial resolution to disentangle contributions from temporal lobe generators (sensory discrimination) and frontal lobe generators (attention switching).
Conclusion
The mismatch negativity is an automatic, pre-attentive brain response to auditory change. It emerges when a sound violates a regularity established by preceding stimuli, indexing the brain's capacity for sensory memory and perceptual accuracy. Decades of research have established it as a reliable electrophysiological phenomenon with well-characterized properties.
Two major hypotheses have competed to explain the neural mechanisms underlying the MMN. The adaptation hypothesis attributes it to differential neural fatigue, while the model adjustment hypothesis posits active updating of internal representations. Predictive coding has emerged as a unifying framework that accounts for a broad range of empirical features. Detailed computational models built on predictive coding principles reproduce the MMN's sensitivity to deviant probability, deviant magnitude, unexpected repetitions, sound omission, and NMDA receptor antagonists. MEG experiments also support the core claim that the MMN reflects active cortical prediction rather than passive habituation.
In schizophrenia, reduced MMN amplitude is a robust and systematic finding. Frequency-deviant MMN tracks illness progression and cognitive decline, while duration-deviant MMN may index genetic vulnerability. The deficit appears to be specific to deviance detection, with neuronal adaptation remaining intact. This specificity aims to identify a clear target for treatment development, and the cross-species translatability of the MMN aims to make it a practical biomarker for preclinical drug testing.
Recording meaningful MMN data requires careful methodology. Deviant type, probability, and magnitude must be selected with deliberate reference to the computational principles that govern the response. The many-standards paradigm should accompany the conventional oddball when separating adaptation from deviance detection is necessary. Unconventional designs using unexpected repeats or sound omissions can test specific predictions of the predictive coding framework. When source localization matters, MEG recording of the MMNm complements EEG. These methodological considerations ensure that the MMN continues to serve as a precise tool for both basic neuroscience and clinical translation.
References
Garrido, M. I., Kilner, J. M., Stephan, K. E., & Friston, K. J. (2009). The mismatch negativity: a review of underlying mechanisms. Clinical neurophysiology, 120(3), 453-463. https://doi.org/10.1016/j.clinph.2008.11.029
Wacongne, C., Changeux, J. P., & Dehaene, S. (2012). A neuronal model of predictive coding accounting for the mismatch negativity. The Journal of neuroscience, 32(11), 3665-3678. https://doi.org/10.1523/JNEUROSCI.5003-11.2012
Lieder, F., Stephan, K. E., Daunizeau, J., Garrido, M. I., & Friston, K. J. (2013). A neurocomputational model of the mismatch negativity. PLoS computational biology, 9(11), e1003288. https://doi.org/10.1371/journal.pcbi.1003288
Näätänen, R., & Kähkönen, S. (2009). Central auditory dysfunction in schizophrenia as revealed by the mismatch negativity (MMN) and its magnetic equivalent MMNm: a review. International Journal of Neuropsychopharmacology, 12(1), 125-135. https://doi.org/10.1017/S1461145708009322
Koshiyama, D., Kirihara, K., Tada, M., Nagai, T., Fujioka, M., Usui, K., ... & Kasai, K. (2020). Reduced auditory mismatch negativity reflects impaired deviance detection in schizophrenia. Schizophrenia bulletin, 46(4), 937-946. https://doi.org/10.1093/schbul/sbaa006
Frequently Asked Questions
What is the mismatch negativity (MMN)?
The mismatch negativity, or MMN, is an electrical brain signal recorded with EEG that occurs when an infrequent “deviant” sound breaks a pattern of repeated “standard” sounds. It reflects the brain's automatic ability to detect change and is generated even when a person is not paying attention to the sounds.
Why is the MMN considered automatic and pre-attentive?
The MMN is automatic because it appears even when a person is reading, watching a film, or actively ignoring the sounds. It is pre-attentive because the brain compares the deviant sound against the memory trace of standard sounds before conscious attention ever engages.
What are the two main hypotheses for how the brain generates the MMN?
The adaptation hypothesis proposes that repeated standard sounds cause neurons to reduce their firing rates, so a rare deviant sound activates “fresh” neurons that produce a larger response. The model adjustment hypothesis proposes that the brain actively builds an internal model of the auditory environment, and the MMN reflects the process of updating that model when a deviant violates expectations.
How does predictive coding explain the MMN?
Predictive coding is a unifying framework that views the brain as actively predicting upcoming sounds based on learned regularities. When actual input deviates from the prediction, a prediction error signal is generated, and the MMN is understood as the electrophysiological manifestation of that prediction error.
What is the many-standards paradigm and why is it used?
The many-standards paradigm presents many different tones at equal probability so that no single tone repeats enough to cause neural adaptation. This allows researchers to isolate deviance detection from adaptation effects, which are confounded in a conventional oddball paradigm.
What is the MMNm and how does it differ from the MMN?
The MMNm is the magnetic counterpart of the MMN, recorded with magnetoencephalography (MEG) instead of EEG. It offers better spatial resolution for separating contributions from different brain regions, such as temporal lobe and frontal lobe generators.
What factors should be considered when designing an MMN experiment?
Researchers must deliberately select deviant type, probability, and magnitude because these parameters systematically shape the MMN response. Lower probability and larger magnitude deviants produce larger amplitudes, and using both frequency and duration deviants is essential for clinical studies. The many-standards paradigm should accompany the conventional oddball when separating adaptation from deviance detection is necessary.
Emotiv is a neurotechnology leader helping advance neuroscience research through accessible EEG and brain data tools.
Medical Disclaimer: The information provided on this website is for educational and informational purposes only and is not intended as medical or health advice. This content may contain errors and should not be relied upon to make life-altering health, medical, or lifestyle choices. Always seek the advice of your physician or other qualified health provider with any questions you may have regarding a medical condition or treatment.
Christian Burgos
