Passive exposure to speech sounds modifies change detection brain responses in adults

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY-NC-ND 4.0 https://creativecommons.org/licenses/by-nc-nd/4.0/ Passive exposure to speech sounds modifies change detection brain responses in adults © 2018 Elsevier Inc. Accepted version (Final draft) Kurkela, Jari; Hämäläinen, Jarmo; Leppänen, Paavo H.T.; Shu, Hua; Astikainen, Piia Kurkela, J., Hämäläinen, J., Leppänen, P. H., Shu, H., & Astikainen, P. (2019). Passive exposure to speech sounds modifies change detection brain responses in adults. NeuroImage, 188, 208-216. https://doi.org/10.1016/j.neuroimage.2018.12.010 2019 Accepted Manuscript Passive exposure to speech sounds modifies change detection brain responses in adults L.O. Kurkela Jari, A. Hämäläinen Jarmo, H.T. Leppänen Paavo, Shu Hua, Astikainen Piia PII: S1053-8119(18)32154-2 DOI: https://doi.org/10.1016/j.neuroimage.2018.12.010 Reference: YNIMG 15478 To appear in: NeuroImage Received Date: 27 June 2018 Revised Date: 3 December 2018 Accepted Date: 5 December 2018 Please cite this article as: Kurkela Jari, L.O., Hämäläinen Jarmo, A., Leppänen Paavo, H.T., Hua, S., Piia, A., Passive exposure to speech sounds modifies change detection brain responses in adults, NeuroImage (2019), doi: https://doi.org/10.1016/j.neuroimage.2018.12.010. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 1 Passive exposure to speech sounds modifies change detection brain 1 responses in adults 2 3 Kurkela Jari L.O. 1* , Hämäläinen Jarmo A. 1 , Leppänen Paavo H.T. 1 , Shu Hua 2 , Astikainen 4 Piia 1 5 6 1 Department of Psychology, University of Jyvaskyla, Jyväskylä, Finland 7 2 National Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal 8 University, Beijing, China 9 10 * Corresponding author: Department of Psychology, Mattilanniemi 6-8, 40014 University 11 of Jyvaskyla, Jyväskylä, Finland; phone: + 358 408054538; e-mail: [email protected] 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 2 ABSTRACT 33 In early life auditory discrimination ability can be enhanced by passive sound exposure. In 34 contrast, in adulthood passive exposure seems to be insufficient to promote discrimination 35 ability, but this has been tested only with a single short exposure session in humans. We 36 tested whether passive exposure to unfamiliar auditory stimuli can result in enhanced 37 cortical discrimination ability and change detection in adult humans, and whether the 38 possible learning effect generalizes to different stimuli. To address these issues, we 39 exposed adult Finnish participants to Chinese lexical tones passively for 2 h per day on 4 40 consecutive days. Behavioral responses and the brain’s event-related potentials (ERPs) 41 were measured before and after the exposure for the same stimuli applied in the exposure 42 phase and to sinusoidal sounds roughly mimicking the frequency contour in speech 43 sounds. Passive exposure modulated the ERPs to speech sound changes in both ignore 44 (mismatch negativity latency, P3a amplitude and P3a latency) and attend (P3b amplitude) 45 test conditions, but not the behavioral responses. Furthermore, effect of passive exposure 46 transferred to the processing of the sinusoidal sounds as indexed by the latency of the 47 mismatch negativity. No corresponding effects in the ERPs were found in a control group 48 that participated to the test measurements, but received no exposure to the sounds. The 49 results show that passive exposure to foreign speech sounds in adulthood can enhance 50 cortical discrimination ability and attention orientation toward changes in speech sounds 51 and that the learning effect can transfer to non-speech sounds. 52 53 54 Keywords: perceptual learning, speech sounds, passive exposure, event-related potentials 55 56 57 58 59 60 61 62 63 64 65 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 3 1. Introduction 66 67 68 In early infancy, cortical discrimination ability is enhanced even by passive sound 69 exposure alone (e.g., Cheour et al., 1998; Cheour et al., 2002; Kuhl, 2004; Trainor, Lee, & 70 Bosnyak, 2011). In contrast, in adulthood passive sound exposure in absence of training 71 seems to be insufficient to affect the neural-level discrimination ability (Näätänen et al., 72 1993; Sheehan et al., 2005; Elmer et al., 2017) or behavioral discrimination performance 73 (Wright et al., 2010; 2015). Instead, effects of active discrimination training have been 74 shown in several studies by measuring the mismatch negativity (MMN) (Kraus et al., 75 1995; Tremblay et al., 1997; Tremblay et al., 1998; Tamminen et al., 2015), P3a (Atienza 76 et al., 2004; Uther et al., 2006; Seppänen et al., 2012) and P2 (Atienza et al., 2002; Reinke 77 et al., 2003; Sheehan et al., 2005) components of event-related potentials (ERPs). These 78 components reflect pre-attentive change detection (MMN) and subsequent attention 79 shifting (P3a) based on a memory trace formed by the learned sound feature (Näätänen et 80 al., 2005; Polich, 2007) and sound feature encoding and stimulus classification (P2) (for a 81 review see Crowley & Colrain, 2004). 82 Even though effects of passive exposure have been studied on brain responses 83 related to pre-attentive change detection, possible effects of passive exposure on attentive 84 change detection of sounds have not been investigated, i.e. effects on N2b and P3b 85 components. 1-hour attentive identification training with speech sounds, however, showed 86 learning-related changes in N2b and P3b (Alain et al., 2010). In another study, 87 identification training resulted in only enhanced P3b responses and no changes in N2b 88 (Ben-David et al., 2011). Similarly, attentive discrimination training with speech sounds 89 resulted in enhanced P3b-like but not N2b-like microstates in electroencephalography 90 (Giroud et al., 2017). 91 Even if previous studies have failed to demonstrate effect of passive exposure on 92 auditory change detection in adults (Näätänen et al., 1993; Sheehan et al., 2005; Elmer et 93 al., 2017; Wright et al., 2010; 2015), passive exposure to sounds seems not to be entirely 94 ineffective either. Perceptual learning on an auditory discrimination task (Wright et al., 95 2010) or on an identification task (Wright et al., 2015) that is combined with sessions of 96 passive exposure is more efficient than the active training alone as indexed by behavioral 97 responses (Wright et al., 2010; 2015). Furthermore, passive exposure to sounds increases 98 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 4 amplitude of the P2 component (Sheehan et al., 2005; Tremblay et al., 2007; Tremblay et 99 al., 2010; Ross et al., 2013). Thus, passive exposure seems to have at least facilitating 100 effect on auditory perceptual learning in adulthood. 101 One possible reason for the failure of the previous studies in demonstrating the effect 102 of passive exposure on discrimination ability can be the short, 1 - 2 hour, exposure time 103 that has been used in previous studies (Näätänen et al., 1993; Sheehan et al., 2005; Elmer 104 et al., 2017). Active training studies have provided training over several days, and this has 105 led to better discrimination ability as indexed by the enhancement of the MMN, P3a and 106 P3b responses (Kraus et al., 1995; Tremblay et al., 1997; Giroud et al., 2017). 107 Furthermore, it has been shown that sleep deprivation hinders the learning-related increase 108 in the MMN amplitude and prevents the appearance of the P3a component (Atienza et al., 109 2004). Thus, the learning-related changes in cortical responses seem to be sleep110 dependent, probably requiring memory consolidation during nocturnal sleep (Alain et al. 111 2015). Based on this assumption, it could be possible that the effects of mere passive 112 exposure emerge if the exposure is expanded on several days, allowing memory 113 consolidation. This has not yet been tested explicitly, however. 114 The evidence on generalization of the auditory learning to stimulus features not 115 encountered during training is scarce. There are some studies showing that frequency or 116 syllable discrimination training generalizes to closely similar untrained stimuli (for a 117 review see Wright & Zhang, 2009). One study applied MMN to study the generalization, 118 and showed that categorization training of labial stop consonant generalizes also to 119 alveolar stop consonant as indicated by the shortened latency and increased amplitude of 120 the MMN to non-trained stimuli (Tremblay et al., 1997). 121 In the present study, we tested two highly novel aspects of auditory perceptual 122 learning: i) Effect of passive speech sound exposure on change detection and attention 123 orienting in ignore and attend test conditions, and ii) if the effect of passive exposure is 124 observed, whether it generalizes to ignored non-speech stimuli. Adult native Finnish 125 participants were exposed to speech sounds (changes in Chinese lexical tones) for a total 126 of 8 hours over 4 days. ERPs were recorded before and after the exposure to the same 127 speech sounds and also to sinusoidal sounds roughly mimicking the pitch contours of the 128 speech sounds. A control group received no exposure but participated only in the ERP 129 recordings at the same time intervals as the experimental group. 130 We expect that the passive exposure would result in modulations in the ERPs, 131 reflecting changes in both pre-attentive and attentive change detection and attention 132 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 5 orienting toward changes (MMN, P3a, N2b, and P3b), as the exposure time is longer than 133 in the previous studies (Näätänen et al., 1993; Sheehan et al., 2005; Elmer et al., 2017) and 134 allows memory trace consolidation during the nights between the exposure periods 135 (Stickgold, 2005; Alain et al., 2015). Changes in these ERP components are assumed to 136 occur due to the formation of long-term memory representations of the sounds, making 137 change detection and attention orienting to them more efficient (as in Näätänen et al., 138 1997; Winkler et al., 1999). We also hypothesized, based on the findings on sound 139 frequency training (Wright & Zhang, 2009), that the effect of passive exposure transfers to 140 the non-speech sounds. 141 142 143 2. Material and methods 144 145 146 2.1 Participants 147 148 A total of 39 monolingual Finnish-speaking participants (mean age = 23.0 years, 149 standard deviation [SD] = 3.3 years; 32 females and 7 males) volunteered for the study. 150 They were recruited with announcements in the notice boards and e-mail lists of the 151 University of Jyväskylä. The inclusion criteria for the study were an age of 18–30 years, 152 right-handedness, normal hearing measured using audiometry, and self-reported normal 153 vision (or corrected to normal vision). The exclusion criteria for the study were 154 neurological or psychiatric disorders, including sleep problems, and exposure to or training 155 in tonal languages. However, previous exposure during trips to countries where tonal 156 languages are spoken (maximum of 2 weeks) was accepted. Written informed consent was 157 obtained from each participant before inclusion in the study. The experiment was 158 undertaken in accordance with the Declaration of Helsinki, and the ethical committee of 159 the University of Jyväskylä approved the research protocol. 160 The participants were divided to two groups, one group of participants were 161 passively exposed to speech sounds (n = 18, mean age = 21.7 years, SD ± 1.7) and the 162 other served as a control group (n = 21, mean age = 24.1 years, SD ± 3.6). Data was 163 collected in ignore and attend test conditions (described below). From both ignore and 164 attend conditions data of 3 participants were omitted from statistical analysis due to 165 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 6 extensive artifacts in the EEG. After the omission, data of 18 and 21 control group 166 participants and 18 and 15 passive exposure group participants remained for the ignore and 167 attend test conditions, respectively. In the ignore test condition data, 66.6% of the 168 participants in the exposure group and the same portion of the participants in the control 169 group had some musical training or had played an instrument or sang as a hobby. In the 170 attend test condition data, this was the case for 66.6% of the exposure group and 61.9% of 171 the control group participants. All the participants had studied English and Swedish as a 172 foreign language. In addition, in the ignore test condition data, 61.1% of the exposure 173 group and 88.9% of the control group participants had studied an additional language for 174 over 2 years. In the attend test condition data, this was the case for 73.3% of the exposure 175 and 81.0% of the control group participants. 176 177 2.2 Stimuli 178 179 We exposed the participants to lexical tones, since Finnish belongs to a quantitative 180 language group, and tonal changes are not part of the phonological system in this 181 language. Therefore, we expected that training effects could be observed. Because 182 discrimination threshold for the lexical tones applied in the study was not known for the 183 Finnish participants, and we did not want the participants to actively listen to the sounds, 184 two levels (large and small) of change were selected to maximize the possibility to find an 185 exposure effect. 186 The sounds were prepared so that the first phoneme /a/ was spoken by a female 187 native Chinese speaker with rising (i.e., Chinese lexical tone 2) and falling (i.e., Chinese 188 lexical tone 4) pitch contour, and they were recorded at a sampling rate of 44.1 kHz. The 189 sounds were then digitally edited using SoundForge software (SoundForge 9, Sony 190 Corporation, Japan) to modify them to have a duration of 200 ms. To isolate the lexical 191 tones and keep the rest of the acoustic features identical, pitch tier transfer was performed 192 using Praat software (Praat v5.4.06, University of Amsterdam). Pitch tier transfer 193 generated a rising tone and a falling tone, which were identical to each other, except for a 194 pitch contour difference in fundamental frequency (F0). These two tones were taken as the 195 endpoint stimuli to create a continuum of lexical tones with 10 interval steps. A morphing 196 technique was performed in MATLAB (MathWorks, Inc., MA, US), and a STRAIGHT 197 tool (Kawahara et al., 1999) was used to create the three tones applied in the experiment. 198 The repeatedly presented standard sound was the falling tone (Fig. 1A), and deviant 199 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 7 sounds were a slightly falling tone (small deviant), and a rising tone (large deviant, Fig. 200 1A) corresponding to the tones 11, 7, and 3, respectively, on the tone continuum. All 201 stimuli were normalized to have the same root mean square intensity. The detailed 202 procedure concerning how the stimuli were generated was reported previously elsewhere 203 (Xi et al., 2010). 204 The sinusoidal sounds were created using SoundForge software (SoundForge 9, 205 Sony Corporation, Japan) and they had the same duration (200 ms) and start and end F0 as 206 the corresponding speech sounds. For the standard sound, the starting frequency of F0 was 207 312 Hz and it gradually decreased to 180 Hz. The large deviant had the starting F0 at 233 208 Hz and it increased gradually to 268 Hz (Fig. 1 B). Lastly, the small deviant had a starting 209 F0 at 268 Hz and it gradually decreased to 215 Hz (Fig. 1B). 210 During the preand post-exposure electroencephalogram (EEG) recordings and 211 during the exposure, the sounds were presented in the oddball condition, where a 212 frequently occurring standard stimulus (probability of 0.80) was interspersed with two 213 deviant sounds (large or small, probability of 0.10 for each), using E-prime 1.2 214 (Psychology Software Tools Inc., Sharpsburg, USA) software, resulting in 1000 stimulus 215 presentations with a sound pressure level (SPL) of 70 dB. The inter-stimulus interval (ISI) 216 varied randomly between 440 and 520 ms (offset to onset). The stimuli were delivered in a 217 pseudorandom fashion, with the restriction that consecutive deviant sounds were separated 218 by at least two standard sounds. 219 220 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 14 Table 1. Summary of the significant effects in the repeated measures of ANOVA for the attend condition 409 (speech sounds). * marks exposure-related effect. Degrees of freedom (df), F-values (F), P-values (P), and 410 parietal eta squared (η   ) for effect sizes are reported. 411 Component/variable Effect df F P    N2b/ amplitude Stimulus type 1,34 37.9 0.0001 0.53 Deviant type 1,34 5.2 0.030 0.13 Session 1,34 57.9 0.0001 0.63 Electrode x Group 2,33 3.5 0.043 0.17 Deviant type x Stimulus type 1,34 38.8 0.003 0.23 Stimulus type x Electrode 2,33 6.4 0.004 0.28 Session x Electrode 2,33 4.8 0.014 0.22 Stimulus type x electrode x session 2,33 4.9 0.013 0.23 N2b (deviant)/ latency Deviant type 1,34 77.36 0.0001 0.70 Session 1,34 9.46 0.004 0.22 Electrode 2,33 9.91 0.0001 0.38 Deviant type x Session 1,34 4.17 0.049 0.11 P3b/ amplitude Deviant type 1,34 32.18 0.0001 0.47 Stimulus type 1,34 97.03 0.0001 0.74 Session 1,34 4.68 0.038 0.12 Electrode 2,33 22.08 0.0001 0.57 Electrode x Group 2,33 4.26 0.023 0.21 Deviant type x Stimulus type 1,34 34.93 0.0001 0.51 Stimulus type x Session 1,34 5.37 0.027 0.14 Stimulus type x Electrode 2,33 33.64 0.0001 0.67 Session x Electrode 2,33 5.05 0.012 0.23 Stimulus type x Electrode x Group 2,33 9.36 0.001 0.36 Deviant type x Stim type x Electrode 2,33 6.42 0.004 0.28 Deviant type x Stim type x Session x Group* 1,34 6.50 0.015 0.16 P3b (deviant)/ latency Deviant type 1,34 8.99 0.005 0.21 Session 1,34 59.89 0.0001 0.64 Electrode 1,34 4.08 0.027 0.20 412 3.1.4 Correlations between ERPs in attend condition and behavioral responses 413 414 In the exposure group, there was a marginally significant correlation between the 415 post measurement P3b amplitude and reaction times for the small deviant, r = –0.497, p = 416 0.059, 99% CI [–0.80, –0.11]; the larger the response amplitude was, the faster the reaction 417 time became. Other correlations were non-significant. 418 419 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 15 420 421 Fig. 2. Grand averaged P3b responses in the exposure group, n =15 and the control group, n = 21. The 422 gray lines represent responses to deviant sounds from the pre measurement, and the black lines signify the 423 responses to deviant sounds from the post measurement. Grand averaged waveforms are presented as mean 424 values from the collapsed electrode clusters (left, middle, and right parietal; see Supplementary Fig. 1 and 3). 425 In lower panel, grand averaged scalp topographies of the deviant responses as a mean amplitude value of the 426 analyzed time window at 360–410 ms from the 128 electrodes for the P3b for the exposure group and control 427 group are shown . 428 429 430 431 432 433 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 16 434 435 Fig. 3. Passive exposure enhanced the P3b amplitude. The mean amplitude values of P3b averaged across 436 the three electrode clusters (left, middle, and right parietal) in 360–410 ms time window. The white bars 437 indicate the amplitude values at the pre measurement, and the gray bars represent those at the post 438 measurement. *** indicates a statistically significant difference (p < 0.05) and error bars indicate the 439 standard error of the mean. Post exposure responses to the deviant sounds were enhanced compared to those 440 in pre measurement. No such effect was found in the control group. 441 442 443 3.2 Ignore condition for speech sounds 444 445 3.2.1 MMN component 446 447 There was no exposure effect for the amplitude of the MMN response. However, for 448 the MMM latency, an interaction effect of session x group was found (Table 2). Separate t449 tests for the groups comparing the latencies between pre and post measurements showed 450 that the deviant response latencies became shorter from the pre to post measurements in 451 the exposure group, while no such changes in latencies were found in the control group 452 (Fig. 4 and 5). 453 454 3.2.2 P3a component 455 456 For the amplitude of the P3a component, an interaction effect for stimulus type x 457 session x group was found (Table 2). Subsequent ANOVA (stimulus type x session) 458 performed separately for each group revealed that there was an interaction effect of 459 stimulus type x session in the exposure group, F 1,17 = 5.66, p = 0.029, η   = 0.25, while no 460 session-related main or interaction effects were observed in the control group 461 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 17 (Supplementary Table 4). Separate t-tests for standard and deviant responses comparing 462 the amplitude change from pre to post measurement were conducted for the exposure 463 group. Responses to deviant sounds increased in amplitude toward a positive polarity and 464 the same was observed for the responses to the standard sounds (Fig. 4 and 5). 465 The passive exposure affected the latencies of the deviant responses in the P3a time 466 window, as indicated by the deviant type x session x group interaction effect (Table 2). 467 Subsequent ANOVA (deviant type x session) performed separately for each group 468 revealed an interaction effect of deviant type x session in the exposure group, F 1,17 = 9.01, 469 p = 0.008, η   = 0.35. No session-related main or interaction effects were found in the 470 control group. Subsequent t-tests comparing latencies separately for the deviant types in 471 the exposure group showed that after the exposure, the latency of P3a to the large change 472 was shorter (273.6 ms ± 14.9) than it was before the exposure (294.0 ms ± 24.5), t(17) = 473 2.92, p = 0.02, 95% CI [8.0, 34.0], d = 1.00 (Fig. 5). No effect was found for the small 474 change (Fig. 5). 475 476 Table 2. Summary of the significant effects in the repeated measures of ANOVA for the ignore condition 477 where the speech sounds were presented. * marks exposure-related effect. Degrees of freedom (df), F-values 478 (F), P-values (P), and parietal eta squared (    ) for effect sizes are reported. 479 480 Component/variable Effect df F P    MMN/ amplitude Stimulus type 1,34 6.73 0.014 0.17 MMN (deviant)/ latency Electrode 2,33 5.14 0.011 0.24 Deviant type 1,34 6.92 0.013 0.17 Deviant type x Session 1,34 4.17 0.049 0.11 Session x Group* 1,34 6,71 0.014 0.17 P3a/ amplitude Deviant type 1,34 4,27 0.047 0.11 Stimulus type 1,34 4.34 0.045 0.11 Session 1,34 6.91 0.013 0.17 Electrode 2,33 4.59 0.017 0.22 Stimulus type x Group 1,34 46.57 < 0.0001 0.58 Session x Group* 1,34 21.12 < 0.0001 0.38 Deviant type x Group 1,34 13.76 0.001 0.29 Dev type x Stim type x Group 1,34 12.83 0.001 0.27 Stimulus type x Session x Group* 1,34 9.90 0.003 0.23 Stimulus type x Electrode x Group 2,33 4.95 0.013 0.23 P3a (deviant)/ latency Session x Group* 1,34 4.78 0.036 0.12 Deviant type x Session x Group* 1,34 6.64 0.014 0.16 481 482 483 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 18 484 Fig. 4. Grand averaged deviant responses in the time windows of the mismatch negativity (MMN) and 485 P3a responses in the exposure group, n = 18, and in the control group, n = 18. Grand average waveforms 486 to deviant responses (small and large deviant averaged) are presented as mean values from the collapsed 487 electrode clusters (left frontal, middle frontal, right frontal; see Material and methods and Supplementary 488 Fig. 1 and 4). The gray lines represent responses in the pre measurement, while black lines represent 489 responses in the post measurement for the exposure group and the control group. The light gray bars 490 represent the time windows analyzed for the MMN and P3a responses (190–240 ms and 250–300 ms post 491 stimulus onset, respectively). In lower panel, the grand averaged scalp topographies of the responses to 492 deviant sounds as a mean amplitude value of the analyzed time window from the 128 electrodes for the 493 MMN and P3a for the exposure group and the control group are shown. 494 495 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 19 496 497 Fig. 5. Mismatch negativity latency and P3a latency and amplitude are enhanced after the passive 498 exposure. The white and gray bars represent the mean values from the pre and post measurements, 499 respectively. Error bars indicate the standard error of the mean. *** indicates a statistically significant (p 500 < .05) difference. The latency of the response to the deviant sounds in the MMN time window shortened 501 from the pre measurement (222.45 ms ± 19.61)) to the post measurement (194.57 ms ± 24.81), t(17) = 4.79, 502 p = 0.00, 95% CI [17.44, 39.78], d = 1.25), in the exposure group, while there were no changes in the 503 control group. In addition, in the P3a time window, deviant response latencies became significantly shorter 504 for the large deviant after the exposure, but not for the small deviant, and there were no changes in the 505 control group. Standard response amplitudes were more positive after the exposure (–0.05 µV ± 0.64) 506 compared to the pre measurement (–0.53 µV ± 0.55), t17 = 3.18, p = 0.026, 95% CI [–0.79, –0.19], d = 0.80. 507 Also deviant response amplitudes became more positive after the exposure (0.57 µV ± 0.57) compared to the 508 pre measurement (–0.30 µV ± 0.74), t(17) = 4.70, p = 0.002, 95% CI [–1.23, –0.53], d = 1.32. 509 510 511 512 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 20 3.3 Ignore condition for sinusoidal sounds 513 The transfer effect to non-exposed sound features was tested by presenting 514 sinusoidal sounds roughly mimicking the pitch contours of the speech sounds in a passive 515 oddball condition (Fig. 1C). The transfer effect was investigated for the components and 516 variables showing group x session interaction effects in the ignore condition where speech 517 sounds were presented, i.e. for the MMN latency, P3a amplitude and P3a latency. 518 3.3.1 MMN component 519 For the MMN latency, there was a significant interaction effect of electrode cluster x 520 session x group (Table 3). Subsequent ANOVA (electrode cluster x session) performed 521 separately for each group revealed an significant main effect session for both groups F 1,17 = 522 15.03, p = 0.001, η   = 0.47; F 1,17 = 5.90, p = 0.027, η   = 0.26 (Supplementary table 5). 523 Post hoc paired samples t-tests comparing latencies between pre and post measurements 524 separately for groups showed that in the exposure group the latencies got significantly 525 shorter from pre measurement (234.9 ms ± 10.58) to post measurement (215.27 ms ± 526 21.44), t(17) = 3.88, p = 0.014, 95% CI [10.43, 29.74], d = 1.16). There were no changes 527 in the latencies in the control group. 528 Table 3. Summary of the significant effects in the repeated measures of ANOVA for the sinusoidal sounds in 529 the ignore condition. * marks exposure-related effect. Degrees of freedom (df), F-values (F), P-values (P), 530 and Partial eta squared (η   ) for effect size are reported. 531 Component/variable Effect df F P    MMN (deviant)/ Latency Electrode 2,33 6.89 0.003 0.29 Session x Group* 1,34 19.57 < 0.001 0.37 Session x Electrode x Group* 2,33 4.52 0.018 0.22 P3a/ Stimulus 1,34 37.9 < 0.001 0.53 Amplitude Electrode 2,33 8.4 0.001 0.34 Stimulus x Group 1,34 12.4 0.001 0.27 Stimulus x Session 1,34 4.2 0.047 0.11 Stimulus x Session x Electrode x Group* 2.33 4.0 0.027 0.20 P3a (deviant)/ latency Deviant type 1,34 7.36 0.01 0.18 Deviant type x Group 1,34 5.65 0.023 0.14 532 3.3.2 P3a component 533 For the P3a amplitude, there was an interaction effect of stimulus type x electrode 534 cluster x session x group (Table 3). The following ANOVAs (electrode cluster x session x 535 group) for the deviant responses (Supplementary table 6) or (electrode cluster x session x 536 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 21 group) performed separately for each electrode cluster revealed no session x group 537 interaction effects (Supplementary table 7) Subsequent ANOVA (stimulus type x electrode 538 cluster x session) performed separately for each group revealed significant interaction 539 effect of stimulus type x electrode cluster x session in the control group, while there was 540 no session-related effects in the exposure group (Supplementary Table 8). T-tests 541 comparing deviant and standard responses separately from pre measurement to post 542 measurement in each electrode cluster was performed in the control group. They did not 543 reveal any significant results. 544 545 546 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 22 547 Fig. 6. Grand averaged MMN and P3a responses to sinusoidal sounds. Grand averaged waveforms of 548 responses reflecting MMN and P3a in the exposure group (n = 18) and the control group (n = 18). The gray 549 lines represent responses to deviant sounds (large and small deviant types averaged) from the pre 550 measurement, while black lines represent responses to deviant sounds from the post measurement for the 551 exposure group and the control group. The light gray bars represent the time windows applied in the analysis 552 of for the P3a amplitude (mean amplitude value between 300–350 ms). The mean scalp topographies of the 553 differential response (standard subtracted from the deviant, deviant types averaged) from the 128 electrodes 554 for the P3a for the exposure group and the control group from the analyzed time window. Although there 555 was a significant stimulus type x electrode cluster x session x group effect for the P3a amplitude, post-hoc 556 tests did not reveal any generalization of the exposure effect. Please note that the MMN amplitude was not 557 investigated since there was no exposure effect for it. 558 559 MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 23 4. Discussion 560 561 562 Here we show in adult humans that passive exposure to foreign speech sounds for 4 563 consecutive days, 2 h per day, enhanced the neural discrimination ability and attention 564 orientation toward changes in the speech sounds as indexed by ERPs recorded in ignore 565 and attend test conditions. The effect of passive exposure to auditory change detection 566 mechanism has earlier been found only in infants (Cheour et al., 1998; Cheour et al., 2002; 567 Kuhl, 2004; Trainor, Lee, & Bosnyak, 2011). In the attend test condition, effect of passive 568 exposure was demonstrated as enhanced P3b amplitude. In the ignore test condition, 569 effects of passive exposure were demonstrated as shortened latency of the MMN and 570 enhanced amplitude and shortened latency of the P3a. 571 The learning effect generalized to some extend to novel sounds: the latency of the 572 MMN shortened to the sinusoidal sounds not encountered during the exposure phase. This 573 effect was demonstrated only in the exposure group, not in the control group. 574 Effects of auditory perceptual learning have rarely been tested for attentive change 575 detection. Here we showed that passive exposure enhanced the amplitude of the P3b and 576 there was trend towards significant correlation between the enhanced P3b amplitude and 577 shortened behavioral reaction times to small deviant. Previously, it has been shown that 578 when the perceptual task becomes easier, the P3b amplitude increases (Isreal et al., 1985). 579 Our results are also in line with one previous study which showed that active training to 580 discriminate speech sounds enhances the microstates related to the P3b component 581 accompanied by improvements in behavioral reaction times (Giroud et al., 2017). In the 582 light of the context-updating model (Polich, 2007), passive exposure seems to ease 583 comparison process between the representation of the standard sound in memory and the 584 deviant sound input, which is also reflected as shortened reaction times. 585 N2b was the other component that was investigated in the attend test condition. 586 Here, the amplitude of the N2b was not enhanced, nor was its latency shortened due to 587 passive exposure. In prior studies applying attentive training, in line with our findings, 588 N2b was not enhanced during identification task (Ben-David et al. 2011) or during 589 discrimination task (Giroud et al. 2017). However, a study that had longer practice period 590 than in study by Ben-David et al. (2011) reported that enhancement in ability to identify 591 speech sounds were followed by increased N2b (Alain et al. 2010). It remains thus unclear 592