Omission mismatch negativity builds up late Ja ´ nos Horva ´th a , Dagmar Mu ¨ ller b , Annekathrin Weise b and Erich Schro ¨ ger b The mismatch negativity (MMN) event-related potential (ERP) is elicited by irregular sounds in a regular sound sequence. MMN depends on the length of the preceding regular stimulation: in repetitive sequences, changes following only two or three repetitions may elicit MMN. Thus, MMN builds up rapidly. Infrequent omissions in a repetitive sound sequence may also elicit an MMN-like waveform. In this study, the build-up of this omission-related ERP was investigated. Trains of two to nine tones with omissions at the penultimate stimulus positions were presented. Results showed no consistent omission-related ERP. In contrast, when omissions occurred with 9% probability among continuously presented tones, the omission-related waveform was elicited. Thus, the omission-related ERP builds up later than MMN. NeuroReport 21:537–541  c 2010 Wolters Kluwer Health | Lippincott Williams & Wilkins. NeuroReport 2010, 21:537–541 Keywords: event-related potentials, loudness summation, mismatch negativity, omission, rebound response, temporal window of integration a Institute for Psychology, Hungarian Academy of Sciences, Budapest, Hungary and b Institute of Psychology I, University of Leipzig, Leipzig, Germany Correspondence to Ja ´ nos Horva ´th, PhD, Institute for Psychology, Hungarian Academy of Sciences, PO Box 398, Szondi u 83/85, H-1394 Budapest, Hungary Tel: + 36 1 354 2397; fax: + 36 1 354 2416; e-mail: horvath@cogpsyphy.hu Received 12 February 2010 accepted 7 March 2010 Introduction Mismatch negativity (MMN) is an auditory event-related potential (ERP) elicited when a regularity of a stimulus sequence is violated. The elicitation of MMN by such a violation depends on the length of the preceding regular sound sequence: detecting a regularity violation requires the establishment of a regularity-representation. The build-up of MMN with an increasing number of preceding regular sounds has been interpreted to show the accumulation of evidence for the presence of the regularity [1]. For simple feature-repetition regularities such as, duration-repetition, frequency-repetition or intensity-repetition, the regularity representation is built up rapidly: changes following two repetitions (or maybe only one repetition) elicit an MMN [2–5]. Regularities may not only be broken by irregular stimuli, but also by stimulus omissions, which also elicit a characteristic negativity like the MMN, but only if stimuli are presented rapidly (see below). In this study, we investigated the build-up of this omission-related wave- form as a function of the length of the preceding repetitive sequence. We measured omission-related ERPs in two conditions: (a) the train condition, in which omissions occurred at the penultimate position of two- tone-long to nine-tone-long stimulus trains; and (b) in the continuous condition, in which omissions were randomly interspersed within a continuously presented tone sequence. This study may also shed more light on models that explain the nature of the omission-related waveform. Models that explain the omission-related waveform have to accommodate the fact that although occasionally presented, irregular stimuli generally elicit an MMN, omissions elicit the MMN-like ERP only if the stimulus presentation rate is sufficiently rapid, that is, if the onset-to-onset interstimulus interval is shorter than approximately 200 ms [6,7]. Three models have been proposed: (1) According to the temporal window of integration hypothesis, sound information is integrated within approximately 200 ms long intervals following each stimulus onset [8], with the possibility of multiple simultaneously ‘open’, overlapping integration win- dows corresponding to separate sound onsets. In sequences where the onset-to-onset interval is shorter than the temporal window of integration, an irregular omission reduces the number of stimuli contributing to the given integration interval [9]. For example, if the onset-to-onset interval is 140 ms, regularly two stimulus onsets fall within the 200 ms long window, but only one when an omission occurs, and therefore an MMN is elicited. If the onset-to- onset interval is longer than the window of integra- tion, each window receives contribution from only one stimulus (the stimulus initiating it) regardless of the occurrence of an omission, and therefore no MMN is elicited. That is, according to this model, the omission-related response is an MMN to repetition-regularity violation, which generally builds up rapidly (see above). (2) The loudness summation account [10] is based on the fact that for short sounds, duration contributes to perceived loudness: longer tones are perceived as being louder, especially for durations below 100–150 ms [11]. Using loudness-judgments in tone- pairs, Cowan [12] found that with longer intertone Cognitive neuroscience and neuropsychology 537 0959-4965  c 2010 Wolters Kluwer Health | Lippincott Williams & Wilkins DOI: 10.1097/WNR.0b013e3283398094