1.2 Adaptation and MMN
 In addition to adaptation, another key aspect of the present study is the mismatch negativity (MMN). First described by Näätänen et al. (1978), the MMN typically occurs when a deviant stimulus (hereafter referred to as deviant) with different properties (e.g., frequency, intensity, duration, etc.) is presented within a sequence of repeated standard stimuli (hereafter standards). Extensively studied in the auditory domain, the MMN is measured by subtracting the ERP response to standards, which primarily appear near the end of a stimulus sequence, from that of deviants (Garrido et al., 2009b). The MMN typically peaks around 100–250 ms from the point of deviation (Kujala & Näätänen, 2001). The MMN is usually maximal at fronto-central areas and exhibits a polarity reversal at the mastoids (Kujala & Näätänen, 2001), while its neural source is localized in temporal and frontal areas (Alho, 1995). Importantly, attention is not required to elicit the MMN, as it is associated with pre-attentive processing in the auditory domain (Näätänen et al., 2001). 
Regarding the mechanism of MMN, two hypotheses were postulated mainly, namely the adaptation hypothesis (Jääskeläinen et al., 2004; May et al., 1999; May & Tiitinen, 2010) and the model-adjustment hypothesis (Näätänen & Alho, 1995; Näätänen et al., 2005; Näätänen & Winkler, 1999; Sussman & Winkler, 2001; Winkler et al., 1996). According to the adaptation hypothesis, the MMN arises from the attenuation and delay of the N1 component due to adaptation to the repetitive standard stimuli (May et al., 1999; May & Tiitinen, 2010). Consequently, the MMN does not reflect higher-level comparison processing or mismatch detection but rather signifies a release from stimulus-specific adaptation (Fishman, 2014). In contrast, the model-adjustment hypothesis (Näätänen & Alho, 1995; Näätänen et al., 2005; Näätänen & Winkler, 1999; Sussman & Winkler, 2001; Winkler et al., 1996) posits that the MMN represents the result of change detection between the deviants and the memory trace formed by the standards. Initially, this hypothesis applied only to repeated sounds of the same standards, but it was later expanded to encompass the detection of regularity violation to explain the presence of the MMN in experiments using standard stimuli with predictable patterns or regularities. For example, sequences of tones with increasing frequencies, with deviants that disrupt the regularities, such as repeated or decreased tones (Winkler, 2007). The occurrence of the MMN in such experimental designs supported the model-adjustment hypothesis over the adaptation hypothesis, as no repeated sound could induce adaptation in the regularities (Garrido et al., 2009b).  
 While the aforementioned studies aimed to distinguish the two postulations of MMN mechanisms, it is important to note that they are not mutually exclusive. Indeed, a predictive coding framework could reconcile the adaptation hypothesis and the model-adjustment hypothesis (Carbajal & Malmierca, 2018; Friston, 2005; Garrido et al., 2009b; Winkler, 2007). This framework suggests that our brain extracts regularities from a sequence of stimuli and forms hypotheses about upcoming stimuli based on these regularities. If a deviant occurs, the MMN is elicited because a prediction error arises from the violation of the hypothesis regarding the stimulus and regularities by the deviant. The predictive coding framework incorporates the adaptation hypothesis by proposing that when standard stimuli can be predicted more precisely by the top-down process, less weight is assigned to bottom-up influences, resulting in stronger adaptation manifested by a weaker and delayed N1 (Garrido et al., 2009b). From the predictive coding perspective, adaptation and MMN can be viewed as microscopic and macroscopic correlates of the same deviance-detection process when the repetition rule is involved (Carbajal & Malmierca, 2018). On the other hand, the predictive coding framework aligns with the model-adjustment hypothesis in suggesting that the MMN arises from a comparison between predicted input based on memory traces and actual input. A mismatch between prediction and input generates a prediction error, and the prediction model has to be adjusted. Therefore, within the predictive coding framework, the MMN signifies failures to predict bottom-up input and to suppress prediction error (Garrido et al., 2009b).
Importantly, the formation of the MMN, which is measured by the amplitude of a difference waveform generated by subtracting the average response to standards from that to deviants, is not solely attributed to the more negative amplitude in the deviant trials. It also involves more positive (or less negative) amplitudes resulting from adaptation in the standard trials. For instance, if the amplitudes of the N1 and P2 components become less negative and more positive with repetitions, respectively, the MMN will be larger, assuming that the amplitude of the deviant remains unchanged. Conversely, the MMN will be smaller if the N1 and P2 amplitudes become more negative and less positive with repetitions. This underscores the relevance of repetition positivity (RP), an increase in a slow positive wave from 50 to 250 ms post-stimulus onset with stimulus repetition (Cooper et al., 2013; Recasens et al., 2015), to MMN.
RP plays a crucial role in supporting the predictive coding mechanism of MMN. It is formed by subtracting the ERPs of standard stimuli with fewer repetitions from those with more repetitions. Importantly, the RP is not confined to the N1 time window, which poses a challenge to the adaptation hypothesis of MMN, which proposes that the N1 adaptation effect alone can fully explain MMN (Haenschel et al., 2005). In the study by Haenschel et al. (2005), RP was observed early, starting from the P1 time window, and its amplitude increased with more repetitions of standards. The researchers also found that RP could account for a large proportion of the MMN (Haenschel et al., 2005). Based on these findings, the researchers proposed that RP serves as an ERP correlate of adaptation, a mechanism involved in memory trace formation in the primary auditory cortex (A1). According to the predictive coding account, RP serves as an index of prediction error suppression resulting from the congruence between sensory input and predicted input (Baldeweg, 2007). Notably, since RP corresponds to an increase in P2 and a decrease in N1, it appears contradictory to the conventional understanding of adaptation associated with a diminishing P2. In addition, despite its significance, research on RP remains relatively limited compared to studies on MMN. Thus, further investigations are needed to fully elucidate the relationships between RP, adaptation, and MMN.