© 2007 Federation of European Psychophysiology Societies

T. Baldew eg: ERP Repetition

Journal ofEf fects and Mismatch
Psychophysiology Negativity
2007;
Hogrefe
Vol. Generation
& 21(3–4):204–213
Huber Publishers

ERP Repetition Effects and

Mismatch Negativity Generation
A Predictive Coding Perspective
Torsten Baldeweg
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

UCL Institute of Child Health, University College London (UCL), UK

Abstract. Neuronal adaptation is a ubiquitous property of the cortex. This review presents evidence from MMN studies that show ERP
components with similar adaptive properties. Specifically, I consider the empirical evidence from the perspective of a predictive coding
model of perceptual learning and inference. Within this framework, ERP and neuronal repetition effects (repetition suppression) are seen
as reductions in prediction error, a process that requires synaptic modifications. Repetition positivity is a human auditory ERP component,
which shows similar properties to stimulus-specific adaptation of auditory cortex neurons; a candidate mechanism for auditory trace
formation.

Keywords: mismatch negativity, auditory processing, adaptation, predictive coding, schizophrenia, repeition suppression

Background The brain is organized according to two main principles:

Functional specialization and segregation on the one hand,
Mismatch negativity is one of the most studied and best and functional integration on the other (Friston 2002,
understood human ERP components (Näätänen, 1992). Our 2003). While the former is increasingly well understood,
considerable knowledge of MMN has undoubtedly en- thanks to anatomical and neuroimaging investigations, the
riched understanding of sensory memory and promoted its mechanisms of functional integration remain a challenge
application to the study of developmental and neuropsychi- for contemporary neuroscience, requiring methods with
atric disorders. Nevertheless, there are still critical issues to resolution in both time and space. Given the considerable
be resolved if MMN research is to deepen our understand- architectural complexity of the human brain, such an en-
ing of the biological mechanisms involved. This is particu- deavor appears to be possible only when informed by com-
larly true for research into schizophrenia, a disorder that, putational principles of neural mass action and functional
so far, has escaped a thorough description of the underlying interaction between multiple cortical sites.
pathophysiology. MMN deficits in patients with schizo- Here I will review ERP evidence in the light of a theory
phrenia have been replicated many times since the original of perceptual learning (inference) proposed by Friston
observations by Shelley et al. (1991), making MMN a ro- (2003), which attempts to tackle the problem of functional
bust biological marker of this complex mental disorder (re- integration from a computational perspective, while being
viewed in Umbricht & Krljes, 2005). constrained by realistic neuroanatomical principles (Felle-
man & Van Essen, 1991). The main attraction of this theory
In this brief review I argue that computational models of
for the present review is that it makes specific predictions
perceptual learning can provide a framework to refine neu-
for the behavior of ERP components elicited in oddball
rophysiological investigations of MMN generation, to fos-
tasks such as those commonly used to elicit MMN (Friston,
ter understanding of its pharmacological modulation, and
2005a).
ultimately to help identify cortical mechanisms at fault in
schizophrenia and other disorders. Here, I will focus on
ERP repetition effects to standard tones that precede the
elicitation of mismatch potentials by deviants. Our motiva-
tion for studying such effects, using MMN, was to illumi-
A Model of Perceptual Learning and
nate the process of stimulus encoding into sensory memory Predictions for MMN Generation
thought to be impaired in schizophrenia as evidenced by
behavioral data (Strous, Cowan, Ritter, & Javitt, 1995). In- The reader is referred to Friston (2003, 2005a) for a com-
terest in stimulus repetition effects has also arisen from the prehensive neurobiological and mathematical treatment of
quest for neural correlates of priming (Henson, 2003; Grill- the model (see also Rao & Ballard, 1999). Here I review
Spector, Henson, & Martin, 2005). some of its basic tenets as they appear relevant to MMN

Journal of Psychophysiology 2007; Vol. 21(3–4):204–213 Hogrefe & Huber Publishers

DOI 10.1027/0269-8803.21.34.204
 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation 205

generation. In brief, the principal function of perception is ta, and (d) invasive ERP recordings in human and animal
to determine the most likely causes of sensory data. To that studies (for further discussion see Deouell, this issue).
end, hierarchically organized sensory systems employ pre- Näätänen and Michie (1979) were first to suggest that
dictive coding, where each level of the hierarchy receives MMN is composed of two subcomponents: a sensory-spe-
bottom-up input from the level below and top-down predic- cific component generated in auditory cortices and a sepa-
tions from the level above. If sensory input deviates from rate frontal component. It has been noted that the frontal
the current prediction an error signal is generated, which (scalp negative) and temporal (scalp positive) MMN com-
propagates up the processing hierarchy to adjust the predic- ponents are differentially dependent on stimulus character-
tion at higher levels. This prediction is passed down the istics such as the side of the stimulated ear (Paavilainen,
hierarchy to explain the error away in a top-down fashion. Alho, Reinikainen, Sams, & Näätänen, 1991). These two
In other words, higher sensory levels provide a prediction components also show distinct developmental changes
signal to lower levels about the expected sensory input, sup- (Gomot, Giard, Roux, Barthelemy, & Bruneau, 2000) and
pressing prediction error in those lower levels. A deviation are differentially affected in patients with schizophrenia,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

from the predicted input elicits an error (i.e., mismatch) with predominant impairment of the frontal MMN, while
This document is copyrighted by the American Psychological Association or one of its allied publishers.

signal, which initiates recurrent, self-organizing exchange the temporal component is largely intact (Sato et al., 2003;
of signals in the hierarchy until the error is explained away. Todd, Michie, & Jablensky, 2003; Baldeweg, Klugman,
This corresponds to perceptual inference. Perceptual learn- Gruzelier, & Hirsch, 2002, 2004). These two components
ing corresponds to an optimization of inference that is me- can also be distinguished by current source density distri-
diated by changes in synaptic connections: Prediction error bution (Giard, Perrin, Pernier, & Bouchet, 1990), peak la-
is reduced with stimulus repetition by adjusting connection tency (Rinne, Alho, Ilmoniemi, Virtanen, & Näätänen,
strengths, through synaptic plasticity, in the connections 2000), and in response to attentional competition (Shalgi
among levels. These changes enable backward connections & Deouell, 2007). The label of “frontal” MMN does not
to suppress prediction error in lower levels more efficiently. necessarily imply a frontal cortex origin; it merely refers to
The neuronal correlate of the ensuing reduction in predic- its scalp distribution, which could also arise from genera-
tion error is repetition suppression, i.e., reduction of neu- tors along the lateral (Scherg, Vajsar, & Picton, 1989) and
ronal firing that encodes prediction error. Critically, predic- rostral portions of the superior temporal gyrus (STG), areas
tions are based on nonlinear generative models, which map which contains high-order auditory cortices (Kass & Hack-
nicely onto the nonlinear or modulatory nature of backward ett, 1998). Invasive MMN recordings also point to multiple
projections, i.e., NMDA glutamate receptor dependent pro- generator sites, including auditory cortex, association cor-
jections. tex of the STG (Kropotov et al., 1995, Halgren et al., 1995;
This model makes several predictions for the MMN, Rosburg et al., 2005), as well as lateral (Baudena, Halgren,
which I examine in more detail below (for further details Heit, & Clarke, 1995; Liasis, Towell, Alho, & Boyd, 2001)
see also Friston, 2005a; Stephan, Baldeweg, & Friston, and medial frontal cortex (Rosburg et al., 2005). While the
2006). The model postulates: (1) the existence of several latter locations require confirmation, the existence of STG
hierarchical mismatch generators; (2) a dependence of generator sites outside A1 appear to be a more robust find-
MMN on the probabilistic structure of the stimulus se- ing across animal (Csepe, Karmos, & Molnar, 1987) and
quence; (3) the existence of ERP repetition effects to stan- human studies. Converging evidence for multiple genera-
dards; and (4) a role for short-term synaptic plasticity, ex- tors comes also from neuroimaging studies (e.g., Mueller,
pressed via NMDA receptors and its modulation by acetyl- Jueptner, Jentzen, & Muller, 2002; Opitz, Rinne, Mecklin-
choline (Ach). I now review the evidence for and against ger, von Cramon, & Schroger, 2002; Doeller et al., 2003).
each of those predictions in the context of the existing It is less certain, at present, if there are functional inter-
MMN literature (mainly for Predictions 1 and 2) as well as actions between these multiple generators, and if so, what
studies that address Predictions 3 and 4. their nature is. Notwithstanding the possibility that they
could be activated in parallel by common input, their func-
tional characteristics, such as dependence on physical stim-
ulus characteristics and time course (Näätänen, 1992), so
far support the hypothesis that the frontal activity is initi-
Multiple MMN Generators ated by auditory cortex change-detection mechanisms
(Näätänen & Michie, 1979; Rinne et al., 2000). In contra-
A central assumption of the model is that perceptual learn- distinction to this bottom-up attentional function is a pro-
ing is implemented via changes in connection strength be- posal based on fMRI studies that the fronto-temporal net-
tween hierarchical sensory processing stages. This implies work activated during frequency deviance detection is in-
that prediction error (i.e., mismatch) signals can be elicited volved in top-down contrast enhancement (Opitz et al.,
at each stage. The evidence for multiple mismatch genera- 2002; Doeller et al., 2003). However, the best evidence for
tors is so far based on (a) the topographical distribution of such top-down modulation of auditory change-detection
MMN components, (b) the dependence of MMN compo- comes from studies showing MMN attenuation in patients
nents on experimental factors, (c) recent neuroimaging da- with lesions of the dorsolateral prefrontal cortex (Alho,

Hogrefe & Huber Publishers Journal of Psychophysiology 2007; Vol. 21(3–4):204–213

 206 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation

Woods, Algazi, Knight, & Näätänen, 1994; Alain, Woods, 2000; Haenschel, Vernon, Dwivedi, Gruzelier, & Balde-
& Knight, 1998). weg, 2005).
The predictive coding model is compatible with both Another view on the differential adaptivity of the frontal
bottom-up change-detection and top-down contrast en- and temporal MMN components suggests the possibility
hancement: The change-detection corresponds to the ex- that the frontal MMN generator encodes the probability
pression of greater prediction error relative to standard structure of the stimulus environment (i.e., generates top-
stimuli, whose prediction error is suppressed more effi- down predictions), while the temporal component repre-
ciently. This difference in suppression is mediated by more sents an error signal generated in a lower sensory level
efficient top-down predictions. Whether the repetition-de- (Baldeweg et al., 2004; Haenschel et al., 2005). Clearly
pendent suppression of prediction error depends on plastic- more evidence is needed to determine functional roles for
ity in extrinsic cortico-cortical or just intrinsic connections MMN components and such progress will depend on the
remains an open question; it will require animal models of ability to accurately separate and localize MMN subcom-
MMN and methods designed to measure changes in con- ponents (Rinne et al., 2000; Sato et al., 2000; Shalgi &
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

nectivity in vivo, such as dynamic causal modeling (DCM; Deouell, 2007).

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Friston, Harrison, & Penny, 2003; David, Harrison, & Fris-

ton, 2005), to address this question.

ERP Correlates of Repetition

Suppression
Dependence on the Probabilistic
Within the predictive coding framework, higher sensory
Structure of the Stimulus Sequence levels provide a prediction signal to lower levels about the
expected percept, suppressing prediction error in lower lev-
The oddball paradigm used to elicit MMN is a convenient els, whether stimuli are attended or not. Such repetition
model of perceptual (i.e., stimulus-to-stimulus) learning, suppression (RS) of neuronal activity is well described in
where temporal information (of past stimulus) is used as a monkey inferior temporal cortex in response to repetitions
contextual (top-down) cue that must be weighted against of attended visual stimuli (Desimone, 1996); however, no
the sensory (bottom-up) inputs. The model, therefore, pre- such correlates are known for human auditory ERP. Hence,
dicts that the probabilistic properties of the stimulus se- we set out to test if systematic changes in ERP could be
quence (global probability of events, local sequence ef- demonstrated with increasing the number of standard stim-
fects) are encoded in the form of prior probability estimates ulus repetitions (Haenschel et al., 2005). We expected that
(priors). Indeed, the dependence of MMN on deviant prob- these effects would occur rapidly; hence we used a “roving”
ability is a well-established experimental fact (Näätänen, standard stimulation (Cowan, Winkler, Teder, & Näätänen,
Sams, Jarvilehto, & Soininen, 1983; Sams, Alho, & Nää- 1993) where the frequency of the standard tones changed
tänen, 1983; Imada, Hari, Loveless, McEvoy, & Sams, between stimulus trains. This allowed us to observe repeti-
1993), independent of interdeviant interval effects (Javitt, tion-related effects within each stimulus train, which is not
Grochowski, Shelley, & Ritter, 1998). Furthermore, neuro- possible with a single constant standard.
nal activity in the auditory cortex of the cat also shows the We observed a slow positive wave in the standard ERP
characteristic differential response to probabilistic (odd- from 50–250 ms poststimulus, which increased with stand-
ball) stimuli (Pincze, Lakatos, Rajkai, Ulbert, & Karmos, ard stimulus repetition. This repetition positivity (RP) was
2002; Ulanovsky, Las, & Nelken, 2003; Ulanovsky, Las, recorded from frontocentral electrodes when participants
Farkas, & Nelken, 2004), as identified in human MMN re- were passively listening to (Figure 1) or actively discrimi-
cordings. nating changes in tone-frequency (Haenschel et al., 2005).
According to Friston (2003) cortical hierarchies con- These ERP effects are similar to adaptation effects in
struct their own priors based on Bayesian statistical princi- neuronal spike firing observed in response to sound repe-
ples (empirical Bayes). At each level of the hierarchy the tition in primary auditory cortex (A1) neurons, called stim-
most likely perceptual representations are based on an op- ulus-specific adaptation (SSA): Both occur without overt
timum balance between the priors from the level above and attention to sounds, are stimulus-specific, and develop rap-
the sensory evidence from the level below. The empirical idly (Ulanovsky et al., 2003, 2004; Nelken & Ulanovsky,
nature of these priors can be seen easily by noting that sen- this issue). Indeed, the onset of RP in the latency range of
sory evidence at one level is also the prior for the level P50 strongly implicates the A1 in its generation, based on
below. This notion is compatible with the observation that a latency comparison with intracranial generators of human
MMN subcomponents are differently sensitive to deviant AEP (Liegeois-Chauvel, Musolino, & Chauvel, 1991).
probability. It is mainly the frontal MMN component that Similar repetition effects in auditory mid-latency response
is probability dependent while the temporal component is peaks (P30, P50, N70, and P90) were reported by Dyson,
much less so or not at all (Sams et al., 1983; Sato et al., Alain, & He (2005). The extended latency range of RP and

Journal of Psychophysiology 2007; Vol. 21(3–4):204–213 Hogrefe & Huber Publishers

 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation 207
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 1. Repetition effects on standard ERP and the repetition positivity (RP). (A) The standard ERP at Fz after 2, 6, and
36 repetitions, recorded in a roving standard experiment of Haenschel et al. (2005) during passive listening. The inset
indicates the time window for early ERP components (P30 and P50) shown in (C). (B) RP is the difference wave between
the standard ERPs after 36 and 2 (black line) and 36 and 6 (dashed line) repetitions, respectively. (C) The enlarged early
time window in (A) showing the standard ERP at electrode Cz, here referenced to average mastoids to improve detection
of early components (Dyson et al., 2005). (D) Mean P50 amplitude (error bars represent SE) to standards (as shown in
[C]) and deviants (not shown) for different numbers of standard repetitions, revealing a stimulus by repetition interaction
effect (p < .001); (modified with permission after Haenschel et al., 2005; data for the active condition were omitted).

its modulation by attention suggests that RP may have mul- sentation of the standard stimulus (Näätänen, 1992). We
tiple generators outside A1 – in analogy to MMN – but predicted that if new afferent input contributes to MMN in
these remain to be determined. the P50 latency window, then a stimulus (i.e., deviance)
Furthermore, because we changed the number of standards effect should be visible irrespective of the number of repe-
before the deviant in each stimulus train (i.e., the first of the titions (i.e., a difference in P50 between frequency deviant
new frequency standards), the ensuing MMN showed the ex- and standards). However, no such effect was found; for nei-
pected memory trace effect, i.e., the MMN difference wave ther passive nor active task conditions. In contrast, the stim-
increased with the number of preceding standards. However, ulus (deviance) effect in the P50 window was significantly
this effect was entirely because of the increase in RP with dependent on the number of standard repetitions (Figure
repetition, while the deviant negativity did not change. Again 1D): P50 to deviants was larger (more positive) after less
there is agreement with neuronal responses in cat A1: Rapid or equal than 6 standards (i.e., showing a repetition effect)
and stimulus-specific adaptation occurred only to repeated but smaller after 36 (i.e., showing a mismatch effect). This
standards, but not to deviants (Ulanovsky et al., 2003). This suggests that the early detection of a deviant stimulus in A1
is entirely consistent with perceptual learning of the standard depends on a comparison with a memory trace of the stand-
as the main explanation for differences between responses to ard, rather than new afferent input to A1.
the standard and oddball stimuli. These data also imply that increasing trace strength (i.e.,
The P50 repetition effect also allowed us to examine the number of repetitions) leads to more rapid mismatch detec-
controversial issue of the contribution of new (fresh) affer- tion starting from around 30–50 ms onward, visible also in
ent input to MMN at the level of primary auditory cortex the shortening of the onset latency of mismatch difference
(A1) (see Näätänen, Jacobsen, & Winkler, 2005 for further waves (data not shown in Figure 1), an effect accelerated
discussion). At issue is the question if MMN is a differential by overt attention to the sounds (Haenschel et al., 2005).
response in the auditory cortex between habituated (by re- This early latency effect is also visible in other roving
peated standards) and nonhabituated (responding to devi- standard studies (Baldeweg et al., 2004; Baldeweg, Wong,
ants) N1 neurons (adaptation hypothesis proposed by Jääs- & Stephan, 2006) as well as during the early (perceptual
keläinen et al., 2004) compared to the alternative notion of learning) phase of constant-standard experiments (Balde-
MMN being elicited by neurons coding a memory repre- weg, Williams, & Gruzelier, 1999). This effect could be

Hogrefe & Huber Publishers Journal of Psychophysiology 2007; Vol. 21(3–4):204–213

 208 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 2. Standard ERP, repetition positivity (RP), and MMN to persistent memory traces of an 800 Hz standard tone
compared to other roving frequency standards. Results of a pilot experiment (Baldeweg, unpublished observation) in 9
healthy volunteers (4 female, age range 24–35 years) using a roving standard protocol as in Baldeweg et al. (2004) using
either n = 2, 6, or 18 standard repetitions in each stimulus train. The 800 Hz tone was presented 14 times more often than
any of the other 12 roving frequencies (ranging from 750–1200 Hz) and a minimum of three other stimulus trains intervened
before its repetition. (A) mean standard ERP to n = 6 and 18 repetitions; (B) mean RP obtained by subtracting standard
ERP to 2 repetitions from those to 18 and 6, respectively (according to Figure 1). Note the enhanced RP to 800 Hz tone
compared to other roving standards. (C) mean MMN difference waves to 50 ms duration deviants immediately following
the last standard (combined after 6 and 18 repetitions, all 25 ms duration). Note that the MMN following 800 Hz standard
is enhanced and shows shorter onset latency compared when following other roving standards.

caused by two separate processes: either an acceleration of amid a sequence of 12 roving frequencies. Because at least
neural computation in the same auditory area or the back- three other roving frequency trains intervened before a rep-
propagation of auditory memory traces from higher to low- etition of the 800 Hz tone, progressively more persistent
er sensory levels with increasing trace strength resulting in traces of this sound must have survived over the duration
earlier mismatch detection at lower processing levels. Pre- of the intervening trains. Further study is needed to identify
liminary findings from a combined ERP/fMRI study are the RP correlates of very short-term adaptation (over sec-
indeed supportive of the latter hypothesis (Blankenburg et onds and without intervening stimuli) from those of more
al., in preparation). long-term trace formation (i.e., a preattentive form of fa-
It is important to stress that the reported repetition effects miliarity, lasting over tens of seconds to minutes and across
in human ERP have so far only been observed for tone fre- intervening items). Again, there are parallels with neuronal
quency, while in vivo recordings in A1 also showed SSA to adaptation (SSA) effects in A1 lasting over multiple, nested
sound intensity (Ulanovsky et al., 2003) and interaural time-scales of similar duration (Ulanovsky et al., 2004).
phase disparity (Malone, Scott, & Semple, 2002). Clearly,
further human studies are needed to test if similar repetition
effects can be observed to other stimulus features.
Another interesting question concerns the RP correlate Neuromodulatory Effects on RP and
of long-term auditory memory traces, which can be probed
passively using MMN as demonstrated by Näätänen, MMN Generation
Schroeger, Karakas, Tervaniemi, and Paavilainen (1993).
Indeed, preliminary evidence (Figure 2) suggests that more The predictive coding model postulates that learning ef-
persistent traces result in larger RP and enhanced MMN fects (suppression of prediction error) must involve modu-
with shorter onset latency. Such traces were elicited in an latory backward connections between hierarchical sensory
experiment where one frequency standard (800 Hz) was levels. This implicates NMDA-type glutamate receptor and
presented 14 times more often than any of the other tones Ach receptor-dependent plasticity. In contrast, the propaga-

Journal of Psychophysiology 2007; Vol. 21(3–4):204–213 Hogrefe & Huber Publishers

 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation 209

tion of prediction error per se, via forward (bottom-up) con- different MMN generators are under way using the DCM
nections is mediated mainly via fast AMPA-type glutamate method (Stephan et al., 2006, Garrido, Kilner, Kiebel, Ste-
receptor-dependent processes (see Friston, 2005a, for fur- phan, & Friston, 2007).
ther details). In the model Ach acts via modulation of
NMDA-type glutamate receptors (Gu, 2000; Stephan,
Baldeweg, & Friston, 2006). We, therefore, predicted that
neuromodulation using antagonists and agonists of the Discussion
NMDA and Ach receptors would influence the RP, inas-
much as it reflects the process of repetition suppression
Many of the points raised and interim conclusions drawn in
(and learning), while the negativity in response to the devi-
this review are still speculative. However, the prospect of ad-
ants (i.e., prediction error and inference) should be much
vancing our knowledge about the principles of functional in-
less affected by such drugs.
tegration in hierarchical sensory systems using MMN will
The findings of two separate placebo-controlled studies
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

undoubtedly stimulate further research. For example, model

in healthy volunteers confirm this hypothesis. The first
This document is copyrighted by the American Psychological Association or one of its allied publishers.

generalization across other sensory modalities can be tested

study tested the influence of acute nicotine administration
using MMN to visual and somatosensory stimuli. While the
on RP and MMN using the roving-standard experiment
predictions made at the outset have been supported by avail-
(Baldeweg et al., 2006). Twenty healthy adult volunteers
able empirical evidence, further experimental tests are needed
were assigned randomly to receive either a nicotine gum or
in both human and animal models of MMN. It is possible that
placebo after a baseline ERP recording. Nicotine adminis-
the more conventional, purely feedforward models, based on
tration augmented MMN amplitude in the treatment group
Hubel- and Wiesel-type cell properties of increasing recep-
compared to the baseline recording, while no MMN change
tive field complexity (Koch & Poggio, 1999), can account for
was found in the placebo group. The drug effect was the
the observed MMN effects. For example, feedback connec-
result of a selective enhancement of the frontal RP while
tions may only serve to modulate response magnitude in low-
the negativity to deviants remained unaffected. Further-
er sensory levels, i.e., mediating selective attention (Koch &
more, under nicotine stimulation this repetition positivity
Poggio, 1999; Felleman & Van Essen, 1991). However, spe-
showed a more marked increase with stimulus repetition
cific proposals would need to be developed, which can be
compared to baseline and placebo.
tested experimentally, as suggested here for MMN. Before
A second study tested the effect of the anesthetic keta-
concluding I will briefly review other relevant theoretical
mine (i.v. infusion, placebo: saline) in 18 healthy volunteers
considerations regarding MMN generation and repetition
in a double-blind, cross-over design on RP and MMN using
suppression (RS).
the same stimulation protocol (Baldeweg, Moelle, Merle,
& Born, in preparation). Ketamine, like phencyclidine, is a
noncompetitive antagonist of the NMDA receptor, which
opens in response to binding of the neurotransmitter gluta- Theoretical Models of MMN Generation and
mate. Again, drug effects were found exclusively in ERPs RS
to standards: The RP was reduced in the time window of
50–200 ms (p < .001) while the deviant negativity was un- The notion of predictive coding is conceptually similar to the
affected. In contrast to the nicotine study, the ketamine ef- model adjustment hypothesis proposed by Winkler, Karmos,
fect on RP was more pronounced after few repetitions (less and Näätänen (1996) and that of primitive sensory intelli-
or equal to 6) than after many (n = 36), suggesting that these gence (Näätänen, Tervaniemi, Sussman, Paavilainen, &
two neuromodulators influence short-term auditory plastic- Winkler, 2001) where MMN reflects the deviation from a
ity with somewhat different time scales. preattentive extrapolation of invariant trends in the sound
Independent evidence for a NMDA receptor role in stream and the corresponding adjustments to this model (see
MMN generation comes from studies in monkeys (Javitt, also Winkler, 2007). A critical difference, however, is that
Steinschneider, Schroeder, & Arezzo, 1996) and humans Friston (2003, 2005a) makes specific proposals of how such
(Umbricht et al., 2000), however, no specific test of RP a predictive sensory system is implemented in the brain.
effects was possible in those studies (see also reviews by There have been few attempts at modeling neural repe-
Kahkonen, 2006; Umbricht & Krljes, 2005; Leung, Croft, tition effects using artificial neural systems (see Henson,
Baldeweg, & Nathan, 2007). We have argued elsewhere 2003). Sohal & Hasselmo (2000) describe a realistic model
(Stephan et al., 2006) for the utility of MMN in investigat- of inferior temporal (IT) neurons that show short-term RS
ing the neuromodulatory effects on synaptic plasticity dur- (recency effects over few seconds) as well as long lasting
ing human perceptual learning. familiarity effects across intervening items. Here, recency
In summary, the evidence for neuromodulatory effects effects are caused by (nonsynaptic) neuronal adaptation and
on RP and MMN available so far suggest that RP is partic- familiarity effects by synaptic modifications in feedfor-
ularly sensitive to such effects, in agreement with the pre- ward synapses and local recurrent connections, modulated
dictions outlined above. More direct tests of neuromodula- by cholinergic input from the basal forebrain. A different
tory effects on the cortico-cortical connectivity between theoretical account attributes a role for cholinergic systems

Hogrefe & Huber Publishers Journal of Psychophysiology 2007; Vol. 21(3–4):204–213

 210 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation

in balancing bottom-up sensory evidence and top-down pri- form with coarse stimulus specificity, requiring continuous
ors by encoding their relative uncertainty (Friston, 2005b; stimulation or only short gaps, and cortical SSA, which de-
Yu & Dayan, 2002; see Baldeweg et al., 2006 for further velops rapidly, has a long onset latency, long memory (more
discussion regarding MMN). than 1.7 s), and high stimulus specificity. They argue that
More recently Grill-Spector et al. (2005) reviewed evi- MMN generation in A1 could be supported by the cortical
dence for three qualitative neural models of RS in the visual form of SSA. SSA is also likely to exist in other auditory
system: fatigue, sharpening and facilitation. They point out cortical fields, and could even be larger in A2 than in A1
that different RS models could operate at different time scales (Pincze, Lakatos, Rajkai, Ulbert, & Karmos, 2001). The
and brain systems, perhaps supported by a variety of neuronal study of functional interactions between generators of MMN
mechanisms. Some implications for audition have been dis- and SSA in different cortical fields will be critically informa-
cussed in Baldeweg (2006) and further parallels to auditory tive for theoretical models. The current model would predict
repetition effects are the high stimulus specificity and rapid that A2 exerts top-down influence on SSA in A1, in analogy
onset, the asymptotic increase of RS with number of stimulus to the corticofugal effects of A1 on neuronal activity in the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

repetitions, and the persistence of RS despite intervening auditory thalamus (Villa et al., 1991; Villa, Tetko, Dutoit, de
This document is copyrighted by the American Psychological Association or one of its allied publishers.

items. In addition, repetition effects were also observed in Ribaupierre, & de Ribaupierre, 1999).
stimulus-evoked and induced oscillatory EEG activity at Furthermore, Ulanovsky et al. (2004) reviewed two classes
gamma and beta frequencies (Haenschel, Baldeweg, Croft, of neuronal adaptation mechanisms: (1) activity-dependent
Whittington, & Gruzelier, 2000), however the EEG and ERP mechanisms operating at the output of neurons and (2) stim-
(or local field potential) correlates of SSA remain to be estab- ulus-specific adaptation of neuronal inputs. Their findings
lished (Nelken & Ulanovsky, 2007). In addition, the P50 rep- argue against mechanism (1) and suggest that input mecha-
etition effects of Haenschel et al. (2005) show some similarity nism such as synaptic depression and facilitation play a major
to fMRI adaptation effects in vision (Box 1 in Grill-Spector role in SSA, mediated via intracortical processing rather than
et al., 2005). Hence, a combination of electromagnetic and depression of afferent thalamocortical synapses. This conclu-
hemodynamic imaging is most likely to show which auditory sion is in agreement with the findings of neuromodulatory
regions show RS and if, indeed, they show hierarchical rela- effects on RP reviewed here, if we assume that RP is indeed
tionships, as predicted from the predominant rostral-to-cau- an ERP correlate of SSA.
dal direction of descending (feedback) connections in STG Unravelling the network mechanism responsible for SSA,
(Galaburda & Pandya, 1983; Felleman & Van Essen, 1991). RP, and MMN are likely to be of considerable theoretical
The hypothesis that neuronal adaptation might account for interest (see Grill-Spector et al., 2005) and clinical relevance.
aspects of MMN generation was first suggested by Näätänen First, identification of the mechanisms mediating the cortical
(1984) and later revised by May et al. (1999). Näätänen as- form of auditory SSA could lead to a reconciliation of the
sumed that repetition of standards leads to frequency specific debate on the existence of separate MMN “memory” neurons
inhibition of the tonotopic representation of the standard and (see also Näätänen et al., 2005). While the high frequency
in parallel to release from inhibition of all other (nonadapted) resolution of MMN could be acquired by neurons contribut-
frequency representations. This model assumed the existence ing to N1 generation through stimulus repetition, as shown
of separate “memory” neurons with sufficiently sharp tuning for SSA in A1 neurons (Ulanovsky et al., 2003), the latter
to explain the frequency resolution of MMN. studies also concluded that SSA is an intracortical phenome-
May and colleagues explored a neural network model of non mediated by specific synaptic mechanisms rather than
tonotopically organized auditory cortex, which included two neuronal fatigue. The predictive coding model would suggest
separate inhibitory components: (1) local inhibition via neu- that “memory” is a property of the (predominantly top-down)
ronal adaptation (nonsynaptic spike frequency adaptation) connections between higher and lower sensory levels.
and (2) nonlocal (i.e., synaptic) inhibition mediated via lateral Secondly, there is also a clinical perspective. MMN and RP
inhibition. Experimental evidence on MMN latency and am- are severely diminished in patients with schizophrenia and the
plitude differences to varying degrees of frequency separation degree of impairment in such basic auditory plasticity corre-
between standard and deviant supported the model’s predic- lates with markers of disease severity, such as cognitive and
tions that MMN generation is the result of a combination of functional impairment (Baldeweg et al., 2004; Light & Braff,
both mechanisms. In contrast to Näätänen’s original model, 2005). Given the etiological heterogeneity of schizophrenia it
May et al. suggest that MMN is generated by neural responses would be advantageous to identify neurophysiological corre-
(N1) to the deviant in a delayed and attenuated form, propor- lates of distinct cellular and molecular aberrations (Stephan
tional to the strength of lateral inhibition caused by the stand- et al., 2006). For example, there are two established neuro-
ard. As discussed above, direct evidence for neuronal adapta- pathological features of schizophrenia that have direct impli-
tion underlying MMN generation is now available from single cations for MMN generation and predictive coding models:
neuron studies in A1. Indeed, this important property of cor- (1) the loss of dendritic spines in supragranular layer (II/III)
tical neurons has long been suspected to be a dominant mech- pyramidal neurons (Garey et al., 1998; Glantz & Lewis, 2000)
anism of cortical memory functions (Creutzfeld, 1977). and (2) functional impairment of cortical inhibitory interneu-
Specifically, Ulanovsky et al. (2003) discussed the exis- rons (Lewis, Hashimoto, & Volk, 2005). Mechanism (1) im-
tence of two separate forms of adaptation (SSA): a subcortical pacts directly on the intracortical generation of the deviant

Journal of Psychophysiology 2007; Vol. 21(3–4):204–213 Hogrefe & Huber Publishers

 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation 211

negativity (MMN) in supragranular layers of auditory cortex predictive coding in the auditory system. Trends Cognitive Sci-
(Javitt et al., 1996) and, according to predictive coding mod- ences, 10, 93–94.
els, serves as an error detector projecting forward to the next Baldeweg, T., Klugman, A., Gruzelier, J.H., & Hirsch, S.R.
cortical level (Rao & Ballard, 1999; Friston, 2005a). Mecha- (2002). Impairment in frontal but not temporal components of
nism (2) might interfere (in combination with [1]) with the mismatch negativity in schizophrenia. International Journal of
Psychophysiology, 43, 111–122.
relaying of predictive information via top-down and lateral
Baldeweg, T, Klugman, A., Gruzelier, J.H., & Hirsch, S.R. (2004).
connections (David et al., 2005). It is conceivable that deficits Mismatch negativity potentials and cognitive impairment in
in RP and deviant negativity generation in patients with schiz- schizophrenia. Schizophrenia Research, 69, 203–217.
ophrenia can be partly dissociated depending on the predom- Baldeweg, T., Moelle, M., Merle, V., & Born, J. Ketamine attenu-
inant focus of cellular abnormalities. This could have impli- ates repetition suppression during human auditory perceptual
cations for the design of targeted therapeutic interventions. learning. (in preparation)
Baldeweg, T., Williams, J.D., & Gruzelier, J.H. (1999). Differential
changes in frontal and subtemporal components of mismatch
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Summary negativity. International Journal of Psychophysiology, 33,

This document is copyrighted by the American Psychological Association or one of its allied publishers.

143–148.
Here I reviewed empirical evidence on MMN generation Baldeweg, T., Wong, D., & Stephan, K.E. (2006). Nicotinic mod-
from the perspective of a predictive coding model of per- ulation of human auditory sensory memory: Evidence from
mismatch negativity potentials. International Journal of Psy-
ceptual learning (Friston, 2003) that attempts to shed light
chophysiology, 59, 49–58.
on general principles of cortical organization and function-
Baudena, P., Halgren, E., Heit, G., & Clarke, J.M. (1995). Intra-
al integration. A number of specific hypotheses for MMN cerebral potentials to rare target and distractor auditory and
generation were confirmed: (1) There is evidence for mul- visual stimuli. III. Frontal cortex. Electroencephalography and
tiple, hierarchically deployed MMN components; (2) Clinical Neurophysiology, 94, 251–264.
MMN encodes the probabilistic structure of the sound en- Blankenburg, F., Stephan, K.E., Baldeweg, T., Kilner, J., Driver,
vironment; (3) there are distinct ERP repetition effects to J., & Friston, K.J. Predictive coding and mismatch negativity:
standards – RP, a possible correlate of SSA of auditory cor- A combined on-line EEG and fMRI study. (in preparation)
tex neurons and an auditory analog of RS in the visual sys- Cowan, N., Winkler, I., Teder, W., & Näätänen, R. (1993). Memory
tem (preliminary evidence also suggests that similar RP prerequisites of the mismatch negativity of the auditory event-
correlates exist for more persistent auditory traces); and (4) related potential (ERP). Journal of Experimental Psychology:
neuromodulatory effects to NMDA receptor antagonism as Learning, Memory, and Cognition, 19, 909–921.
well as to nicotine were restricted to the RP. This is in line Creutzfeldt, O.D. (1977). Generality of the functional structure of
the neocortex. Naturwissenschaften, 64, 507–517.
with synaptic changes in modulatory backward connec-
Csepe, V., Karmos, G., & Molnar, M. (1987). Evoked potential
tions that lead to an efficient suppression of prediction error
correlates of stimulus deviance during wakefulness and sleep
(repetition suppression). A number of further hypotheses in cat: Animal model of mismatch negativity. Electroencepha-
regarding the interaction of auditory cortical regions can be lography and Clinical Neurophysiology, 66, 571–578.
tested in suitable experimental models. Furthermore, how David, O., Harrison, L., & Friston, K.J. (2005). Modelling event-
predictive codes guide auditory perception, as suggested related responses in the brain. Neuroimage, 25, 756–770.
recently for visual object recognition (Summerfield et al., Deouell, L.Y. (2007). The frontal generator of the mismatch neg-
2006), is another interesting avenue for research. ativity revisited. Journal of Psychophysiology, 21, 188–203.
Desimone, R. (1996). Neural mechanisms for visual memory and
their role in attention. Proceedings of the National Academy of
Acknowledgment Sciences USA, 93, 13494–13499.
Doeller, C.F., Opitz, B., Mecklinger, A., Krick, C., Reith, W., &
I am grateful to Karl Friston and Klaas Enno Stephan for Schroger, E. (2003). Prefrontal cortex involvement in preatten-
discussion and comments on this manuscript. tive auditory deviance detection: Neuroimaging and electro-
physiological evidence. Neuroimage, 20, 1270–1282.
Dyson, B.J., Alain, C., & He, Y. (2005). I’ve heard it all before:
Perceptual invariance represented by early cortical auditory-
evoked responses. Brain Research Cognitive Brain Research,
References 23, 457–460.
Felleman, D.J., & Van Essen, D.C. (1991). Distributed hierarchical
Alain, C., Woods, D.L., & Knight, R.T. (1998). A distributed cor- processing in the primate cerebral cortex. Cerebral Cortex, 1,
tical network for auditory sensory memory in humans. Brain 1–47.
Research, 812, 23–37. Friston, K. (2002). Functional integration and inference in the
Alho, K., Woods, D.L., Algazi, A., Knight, R.T., & Näätänen, R. brain. Progress in Neurobiology, 68, 113–143.
(1994). Lesions of frontal cortex diminish the auditory mis- Friston, K. (2003). Learning and inference in the brain. Neural
match negativity. Electroencephalography and Clinical Neu- Networks, 16, 1325–1352.
rophysiology, 91, 353–362. Friston, K., Harrison, L., & Penny, W. (2003). Dynamic causal
Baldeweg, T. (2006). Repetition effects to sounds: Evidence for modelling. Neuroimage, 19, 1273–1302.

Hogrefe & Huber Publishers Journal of Psychophysiology 2007; Vol. 21(3–4):204–213

 212 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation

Friston, K. (2005a). A theory of cortical responses. Philosophical Javitt, D.C., Grochowski, S., Shelley, A.M., & Ritter, W. (1998).
Transaction of the Royal Society London: Series B Biological Impaired mismatch negativity (MMN) generation in schizo-
Sciences, 360, 815–836. phrenia as a function of stimulus deviance, probability, and
Friston, K. (2005b). Hallucinations and perceptual inference. Be- interstimulus/interdeviant interval. Electroencephalography
havioural Brain Sciences, 28, 764–766. and Clinical Neurophysiology, 108, 143–153.
Galaburda, A.M., & Pandya, D.N. (1983). The intrinsic architec- Kaas, J.H., & Hackett, T.A. (1998). Subdivisions of auditory cor-
tonic and connectional organization of the superior temporal tex and levels of processing in primates. Audiology and Neu-
region of the rhesus monkey. Journal of Comparative Neurol- ro-otology, 3, 73–85.
ogy, 221, 169–184. Kahkonen, S. (2006). Magnetoencephalography (MEG): A non-
Garey, L.J., Ong, W.Y., Patel, T.S, Kanani, M., Mortimer, A.M., invasive tool for studying cortical effects in psychopharmacol-
Barnes, T.R.E. et al. (1998). Reduced dendritic spine density ogy. International Journal of Neuropsychopharmacology, 9,
on cerebral cortical pyramidal neurons in schizophrenia. Jour- 367–372.
nal of Neurology, Neurosurgery and Psychiatry, 65, 446–453. Koch, C., & Poggio, T. (1999). Predicting the visual world: Silence
Garrido, M.I., Kilner, J.M., Kiebel, S.J., Stephan, K.E., & Friston, is golden. Nature Neuroscience, 2, 9–10.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

K.J. (2007). Dynamic causal modelling of evoked potentials: Kropotov, J.D., Näätänen, R., Sevostianov, A.V., Alho, K., Reini-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

A reproducibility study. Neuroimage, 36, 571–580. kainen, K., & Kropotova, O.V. (1995). Mismatch Negativity to
Giard, M.H., Perrin, F., Pernier, J., & Bouchet, P. (1990). Brain auditory stimulus change recorded directly from the human
generators implicated in the processing of auditory stimulus temporal cortex. Psychophysiology, 32, 418–422.
deviance: A topographic event-related potential study. Psycho- Liasis, A., Towell, A., Alho. K., & Boyd, S. (2001). Intracranial
physiology, 27, 627–640. identification of an electric frontal-cortex response to auditory
Glantz, L.A., & Lewis, D.A. (2000). Decreased dendritic spine stimulus change: A case study. Brain Research Cognitive Brain
density on prefrontal cortical pyramidal neurons in schizophre- Research, 11, 227–233.
nia. Archives of General Psychiatry, 57, 65–73. Liegeois-Chauvel, C., Musolino, A., & Chauvel, P. (1991). Local-
Gomot, M., Giard, M.H., Roux, S., Barthelemy, C., & Bruneau, ization of the primary auditory area in man. Brain, 114,
N. (2000). Maturation of frontal and temporal components of 139–151.
mismatch negativity (MMN) in children. NeuroReport, 11, Light, G.A., & Braff, D.L. (2005). Mismatch negativity deficits
3109–3112. are associated with poor functioning in schizophrenia patients.
Grill-Spector, K., Henson, R., & Martin, A. (2005). Repetition Archives of General Psychiatry, 62, 127–136.
and the brain: Neural models of stimulus-specific effects. Leung, S., Croft, R.J., Baldeweg, T., & Nathan. P.J. (2007). Acute
Trends Cognitive Sciences, 10, 14–23. dopamine D(1) and D(2) receptor stimulation does not modu-
Gu, Q. (2002). Neuromodulatory transmitter systems in the cortex late mismatch negativity (MMN) in healthy human subjects.
and their role in cortical plasticity. Neuroscience, 111, Psychopharmacology (Berlin), 194, 443–451.
815–835. Lewis, D.A., Hashimoto, T., & Volk, D.W. (2005). Cortical inhib-
Haenschel, C., Baldeweg, T., Croft, R.J., Whittington, M., & Gru- itory neurons and schizophrenia. Nature Reviews Neurosci-
zelier, J. (2000). Gamma and beta frequency oscillations in re- ence, 6, 312–324.
sponse to novel auditory stimuli: A comparison of human EEG Malone, B.J., Scott, B.H., & Semple, M.N. (2002). Context-de-
data with in vitro models. Proceedings of the National Acade- pendent adaptive coding of interaural phase disparity in the
my of Sciences USA, 97, 7645–7650. auditory cortex of awake macaques. Journal of Neuroscience,
Haenschel, C., Vernon, D.J., Dwivedi, P., Gruzelier, J.H., & Balde- 22, 4625–4638.
weg, T. (2005). Event-related brain potential correlates of hu- May, P., Tiitinen. H., Ilmoniemi, R.J., Nyman, G., Taylor, J.G., &
man auditory sensory memory-trace formation. Journal of Näätänen, R. (1999). Frequency change detection in human
Neuroscience, 25, 10494–10501. auditory cortex. Journal of Computational Neuroscience, 6,
Halgren, E., Baudena, P., Clarke, J.M., Heit, G., Marinkovic, K., 99–120.
Devaux, B. et al. (1995). Intracerebral potentials to rare target Mueller, B.W., Jueptner, M, Jentzen, W., & Muller, S.P. (2002).
and distractor auditory and visual stimuli. II. Medial, lateral, Cortical activation to auditory mismatch elicited by frequency
and posterior temporal lobe. Electroencephalography and deviant and complex novel sounds: A PET study. Neuroimage,
Clinical Neurophysiology, 94, 229–250. 17, 231–239.
Henson, R.N. (2003). Neuroimaging studies of priming. Progress Näätänen, R. (1984). In search of a short-duration memory trace
in Neurobiology, 70, 53–81. of a stimulus in the human brain. In L. Pulkkinen & P. Lyytinen
Imada, T., Hari, R., Loveless, N., McEvoy, L., & Sams, M. (1993). (Eds.), Human action and personality. Essays in honour of
Determinants of the auditory mismatch negativity. Electroen- Martii Takala (pp. 29–43). Jyvaeskylae: University of Jy-
cephalography and Clinical Neurophysiology, 87, 144–153. vaeskylae.
Jääskeläinen, I.P., Ahveninen, J., Bonmassar, G., Dale, A.M., Il- Näätänen, R. (1992). Attention and brain function. Hillsdale, NJ:
moniemi, R.J., Levanen, S. et al. (2004). Human posterior au- Erlbaum.
ditory cortex gates novel sounds to consciousness. Proceeding Näätänen, R., Jacobsen, T., & Winkler, I. (2005). Memory-based
of the National Academy of Sciences USA, 101, 6809–6814. or afferent processes in mismatch negativity (MMN): A review
Javitt, D.C., Steinschneider, M., Schroeder, C.E., & Arezzo, J.C. of the evidence. Psychophysiology, 42, 25–32.
(1996). Role of cortical N-methyl-D-aspartate receptors in au- Näätänen, R., & Michie, P.T. (1979). Early selective-attention ef-
ditory sensory memory and mismatch negativity generation: fects on the evoked potential: A critical review and reinterpre-
Implications for schizophrenia. Proceedings of the National tation. Biological Psychology, 8, 81–136.
Academy of Science USA, 93, 11962–11967. Näätänen, R., Sams, M., Jarvilehto, T., & Soininen, K. (1983).

Journal of Psychophysiology 2007; Vol. 21(3–4):204–213 Hogrefe & Huber Publishers

 T. Baldeweg: ERP Repetition Effects and Mismatch Negativity Generation 213

Probability of deviant stimulus and event-related brain poten- dependent changes in the responses of inferotemporal neurons.
tials. In R. Sinz & M.R. Rosenzweig (Eds.), Psychophysiology Network: Computational Neural Systems, 11, 169–190.
1980 (pp. 397–405). Jena/Amsterdam: VEB Gustav Fi- Stephan, K.E., Baldeweg, T., & Friston, K.J. (2006). Synaptic
scher/Elsevier. plasticity and dysconnection in schizophrenia. Biological Psy-
Näätänen, R., Schroeger, E., Karakas, S., Tervaniemi, M., & chiatry, 59, 929–939.
Paavilainen, P. (1993). Development of a memory trace for a Strous, R.D., Cowan, N., Ritter, W., & Javitt, D.C. (1995). Audi-
complex sound in the human brain. NeuroReport, 4, 503–506. tory sensory (“echoic”) memory dysfunction in schizophrenia.
Näätänen, R., Tervaniemi, M., Sussman, E., Paavilainen, P., & American Journal of Psychiatry, 152, 1517–1519.
Winkler, I. (2001). “Primitive intelligence” in the auditory cor- Summerfield, C., Egner, T., Greene, M., Koechlin, E., Mangels,
tex. Trends in Neuroscience, 24, 283–288. J., & Hirsch, J. (2006). Predictive codes for forthcoming per-
Nelken, I., & Ulanovsky, N. (2007). Mismatch negativity and ception in the frontal cortex. Science, 314, 1311–1314.
stimulus-specific adaptation in animal models. Journal of Psy- Todd, J., Michie, P.T., & Jablensky, A.V. (2003). Association be-
chophysiology, 21, 214–223. tween reduced duration mismatch negativity (MMN) and
Opitz, B., Rinne, T., Mecklinger, A., von Cramon, D.Y., & Schro- raised temporal discrimination thresholds in schizophrenia.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ger, E. (2002). Differential contribution of frontal and temporal Clinical Neurophysiology, 114, 2061–2070.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

cortices to auditory change detection: fMRI and ERP results. Ulanovsky, N., Las, L., & Nelken, I. (2003). Processing of low-
Neuroimage, 15, 167–174. probability sounds by cortical neurons. Nature Neuroscience,
Paavilainen, P., Alho, K., Reinikainen, K., Sams, M., & Näätänen, 6, 391–398.
R. (1991). Right hemisphere dominance of different mismatch Ulanovsky, N., Las, L., Farkas, D., & Nelken, I. (2004). Multiple
negativities. Electroencephalography and Clinical Neurophys- time scales of adaptation in auditory cortex neurons. Journal
iology, 78, 466–479. of Neuroscience, 24, 10440–10453.
Pincze, Z., Lakatos, P., Rajkai, C., Ulbert, I., & Karmos, G.
Umbricht, D., & Krljes, S. (2005). Mismatch negativity in schiz-
(2001). Separation of mismatch negativity and the N1 wave in
ophrenia: A meta-analysis. Schizophrenia Research, 76, 1–23.
the auditory cortex of the cat: A topographic study. Clinical
Umbricht, D., Schmid, L., Koller, R., Vollenweider, F.X., Hell, D.,
Neurophysiology, 112, 778–784.
& Javitt, D.C. (2000). Ketamine-induced deficits in mismatch
Pincze, Z., Lakatos, P., Rajkai, C., Ulbert, I., & Karmos, G.
negativity generation and AX-continuous performance test in
(2002). Effect of deviant probability and interstimulus/interde-
healthy volunteers: Implications for models of cognitive defi-
viant interval on the auditory N1 and mismatch negativity in
cits in schizophrenia. Archives of General Psychiatry, 57,
the cat auditory cortex. Brain Research Cognitive Brain Re-
1139–1147.
search, 13, 249–253.
Villa, A.E., Rouiller, E.M, Simm, G.M., Zurita, P., de Ribaupierre,
Rao, R.P.N., & Ballard, D.H. (1999). Predictive coding in the vis-
Y., & de Ribaupierre, F. (1991). Corticofugal modulation of
ual cortex: A functional interpretation of some extra-classical
the information processing in the auditory thalamus of the cat.
receptive-field effects. Nature Neuroscience, 2, 79–87.
Experimental Brain Research, 86, 506–517.
Rinne, T., Alho, K., Ilmoniemi, R., Virtanen, J., & Näätänen, R.
(2000). Separate time behaviors of the temporal and frontal Villa, A.E., Tetko, I.V., Dutoit, P., de Ribaupierre, Y., & de Ribau-
mismatch negativity sources. Neuroimage, 12, 14–19. pierre, F. (1999). Corticofugal modulation of functional con-
Rosburg, T., Trautner, P., Dietl, T., Korzyukov, O.A., Boutros, nectivity within the auditory thalamus of rat, guinea pig, and
N.N., Schaller, C. et al. (2005). Subdural recordings of the mis- cat revealed by cooling deactivation. Journal of Neuroscience
match negativity (MMN) in patients with focal epilepsy. Brain, Methods, 86, 161–178.
128, 819–828. Winkler, I. (2007). Interpreting the mismatch negativity. Journal
Sams, M., Alho, K., & Näätänen, R. (1983). Sequential effects on of Psychophysiology, 21, 147–163.
the ERP in discriminating two stimuli. Biological Psychology, Winkler, I., Karmos, G., & Näätänen, R. (1996). Adaptive mod-
17, 41–58. eling of the unattended acoustic environment reflected in the
Sato, Y., Yabe, H., Hiruma, T., Sutoh, T., Shinozaki, N., Nashida, mismatch negativity event-related potential. Brain Research,
T. et al. (2000). The effect of deviant stimulus probability on 742, 239–252.
the human mismatch process. NeuroReport, 11, 3703–3708. Yu, A.J., & Dayan, P. (2002). Acetylcholine in cortical inference.
Sato, Y., Yabe, H., Todd, J., Michie, P., Shinozaki, N., Sutoh, T. et Neural Networks, 15, 719–730.
al. (2003). Impairment in activation of a frontal attention-
switch mechanism in schizophrenic patients. Biological Psy- Accepted for publication: July 16, 2007
chology, 62, 49–63.
Scherg, M., Vajsar, J., & Picton, T.W. (1989). A source analysis of
the late human auditory evoked potentials. Journal of Cogni-
tive Neuroscience, 1, 336–355. Torsten Baldeweg
Shalgi, S., & Deouell, L.Y. (2007). Direct evidence for differential
roles of temporal and frontal components of auditory change UCL Institute of Child Health
detection. Neuropsychologia, 45, 1878–1888. University College London (UCL)
Shelley, A.M., Ward, P., Catts, S.V., Michie, P.T., Andrews, S., & 30 Guilford Street
McConaghy, N. (1991). Mismatch negativity: An index of a London WC1N 1EH
preattentive deficit in schizophrenia. Biological Psychiatry, 30, UK
1059–1062. Tel. +44 20 7905-2756
Sohal, V.S., & Hasselmo, M.E. (2000). A model for experience- E-mail [email protected]

Hogrefe & Huber Publishers Journal of Psychophysiology 2007; Vol. 21(3–4):204–213