The impact of prenatal and postnatal acoustic experience to the discrimination of native versus non-native linguistic rules in the brain: insights from combined electrophysiological and vascular signals in 3-month-olds Authors: Sonja Rossi, Isabella Thurner, Annika Rackl, Jenny Tippmann, Hellmuth Obrig Published: 2026 ABSTRACT Some scaffolding of later language acquisition begins prenatally. Fetuses perceive slow acoustic features (e.g., stress patterns), while fast aspects relevant for phonetics become fully accessible only after birth. This study investigated how prenatal and postnatal acoustic experience influence 3-month-olds’ neural responses to native and non-native linguistic rules. A multi-methodological combination (electroencephalography/EEG and functional near-infrared spectroscopy/fNIRS) assessing temporally and topographically precise neural signals was used, while infants listened to native/non-native pseudowords differing in stress patterns (prenatally accessible) and phonotactics (postnatally accessible). EEG and fNIRS revealed novelty/recognition memory effects for non-native rules (EEG: Nc and positive slow wave/negative slow wave components for phonotactics and stress pattern and fNIRS: larger activations for non-native phonotactic and stress pattern rules). The fNIRS further revealed mature right-hemispheric activations for stress pattern, while a still immature bilateral activation was present for phonotactics. These results suggest a potential impact of prenatal experience for slow prosodic features such as stress pattern and an important postnatal impact for fine-grained phonological features such as phonotactics in 3-month-olds. Introduction Although neurotypical newborn infants can acquire any language spoken in their environment, they are already equipped with astonishing cognitive abilities at birth. They can discriminate speech from non-linguistic sounds (DeCasper and Spence, 1986, Vouloumanos and Werker, 2007) and between rhythmically different languages (Mehler et al., 1988, Moon et al., 1993, Nazzi et al., 1998, Nazzi and Ramus, 2003). Some of these abilities have been shown to be shaped by prenatal acoustic experience. The auditory system is functional during the last trimester of pregnancy (Hepper and Shahidullah, 1994), thus fetuses start to be exposed to some characteristics of their native language from this very early period on. During the prenatal period, however, mostly slow acoustic features are perceived as the uterus acts as a low-pass filter allowing to hear only frequencies below 300–400 Hz (Gerhardt and Abrams, 1996, Querleu et al., 1988). Such frequencies include prosodic features (i.e., the melodic and rhythmic aspects) of language and may allow for vowel discrimination (Urbanec et al., 2024). Perception of more fine-grained phonological features (i.e., the differentiation of consonants and their combination) becomes only possible from birth on, when newborns are capable to perceive the whole frequency spectrum including slow and fast acoustic features (Abrams et al., 1998, Querleu et al., 1988). In the present study, we aim at elucidating the impact of prenatal language exposure to the differentiation of native from non-native linguistic rules at 3 months of age. In particular, we are interested in neural mechanisms, assessed by an innovative simultaneous approach combining electroencephalography (EEG) with functional near-infrared spectroscopy (fNIRS). Because these two neuroscientific methods assess the electrophysiological and vascular response respectively, they show complementary sensitivity to capture maturational neural mechanisms. Vascular and neuronal responses are coupled but sensitivity to temporal and spatial characteristics differ. Therefore, we opted to combine both non-invasive and infant-suitable methods in the same subjects in the same experimental setting to get more extensive insights into early neural developmental trajectories of the differentiation between native and non-native linguistic rules. During the first year of life, infants get more sensitive to linguistic features relevant in their native language and gradually loose sensitivity to non-native features not relevant in their surroundings (Friederici and Wessels, 1993, Jusczyk et al., 1993, Mazuka et al., 2011, Werker and Tees, 1984). This perceptual specialization or ‘tuning in’ towards the native language implies sharpening of discriminatory ability between native and non-native linguistic features and is compatible with the view of critical periods during language development (Werker and Hensch, 2015). Onset of these critical periods are mainly driven by maturational constraints, while exposure to diverse experiences can influence the timeline, thus either accelerating or delaying the opening of critical periods (Werker and Hensch, 2015). One early landmark of language acquisition is acquisition of prosodic aspects especially how words are stressed. Another regularity acquired early by infants is the way phonemes are combined in a given language. The former refers to stress patterns, following either a trochaic pattern (i.e., in bi-syllabic words the first syllable is stressed) such as in German or an iambic pattern (i.e., stress on the second syllable) such as in French. The latter refers to phonotactics (e.g., in German br at the onset of a word is allowed but bz is not, while Slavic languages allow both). Both prosodic and phonotactic, prelexical cues are crucial for segmentation of the acoustic speech stream into single units, which allow for later lexico-semantic learning. While behavioral studies (Friederici and Wessels, 1993, Jusczyk et al., 1993, Werker and Tees, 1984) claimed a larger sensitivity to native linguistic rules around the end of the first year of life, neurophysiological studies (i.e., EEG and fNIRS) identified this discriminatory ability already earlier. Neuroimaging studies using fNIRS investigated the differentiation between native and non-native features in the first months after birth. One study (Vannasing et al., 2016) found different brain areas to be active when newborns listened to native and non-native stories compared to non-linguistic stimuli (i.e., native and non-native stories played backwards). Since native (French) and non-native (Arabic) language differ in their rhythmic structure (i.e., syllable-time versus stress-time) a prosodic pattern is manipulated. Native stories elicited a more left temporal, non-native stories a more right temporal, and non-linguistic stimuli a bilateral temporal activation pattern. Similar results were obtained in 4-month-olds listening to native (Japanese) and non-native (English) sentences compared to non-speech sounds like emotional sounds or monkey vocalizations (Minagawa-Kawai et al., 2011). A recent fNIRS study specifically investigated the impact of prenatal exposure to two different non-native languages in newborns a few days old (René et al., 2025). Increased activations in left temporal areas were found for both native and exposed non-native stories compared to unexposed stories. A study in premature infants showed that even before birth infants are able to discriminate between phonemes and voice-parameters (Mahmoudzadeh et al., 2013). Conversely, another fNIRS study (May et al., 2011) revealed no effect in newborns confronted with native (English) and non-native (Tagalog) sentences low-pass filtered in such a way to preserve prosodic and rhythmic information and thus mimicking language in utero. It should be noted that these studies focused on prosodic features integrated in longer-lasting contexts (i.e., stories or sentences). Other studies manipulated prelexical cues in single words, which is a putatively more challenging task. Prosodic aspects such as the duration of syllables led to a larger bilateral temporo-parietal activation for inconsistent (i.e., not corresponding to the subjects’ native language French) compared to consistent variations in newborns suggesting the need for a greater processing effort (Abboub et al., 2016). Similar findings were obtained in newborns presented with well-formed pseudowords compared to ill-formed ones. Well-formedness followed the rules of the sonority sequencing principle (Clements, 1990) of single phonemes and thus resembles phonotactic rules. Again, the most deviating pseudowords elicited the largest activation in bilateral fronto-temporo-parietal regions (Gómez et al., 2014). Despite the phonotactic rules investigated in this study, the material did not explicitly contrast native versus non-native phonotactics. Such a direct comparison was used in an fNIRS study in 6-month-olds training phonotactically native and non-native consonant clusters at the onset of pseudowords with pseudoobjects in an associative learning setting (Obrig et al., 2017). The results revealed a larger activation in bilateral fronto-temporal regions for native in contrast to non-native pseudowords. A study in 18-month-olds using the same linguistic material in a simple passive listening design, however, failed to find differences between these conditions (Steber and Rossi, 2020). While the question remains unresolved, at what exact time a clear differentiation between native from non-native prelexical rules arises, the hemispheric lateralization in very young infants seems to indicate a lateralized response when longer-lasting contexts are presented while the more effortful processing of words in isolation seems to indicate a more bilateral recruitment of brain regions. In adults, processing of prosodic/stress pattern recruits more right-hemispheric brain regions while phonological/phonotactic is predominantly processed in the left hemisphere (Friederici and Alter, 2004, Hickok and Poeppel, 2007, Rossi et al., 2011). During infancy, newborns and 3–4-month-olds were found to show an adult-like hemispheric asymmetry for both prosody and phonology (Dehaene-Lambertz et al., 2002, Homae et al., 2006, Peña et al., 2003, Sato et al., 2010, Telkemeyer et al., 2009) in materials using sentential or longer-lasting acoustic contexts. Whether a similar adult-like hemispheric lateralization is present very early on also if words or short stimuli are presented in isolation, is still a matter of debate (Gervain et al., 2008, Kreitewolf et al., 2014, Skeide and Friederici, 2016). The present study is intended to provide evidence that hemispheric lateralization can be one index of maturation in the neural underpinnings of language acquisition. A neuroimaging technique (fNIRS) assessing the slow vascular responses, is complemented by EEG assessment, which relies on the fast electrophysiological response. As we will analyze event-related brain potentials (ERPs) in order to identify fast processing dynamics occurring online during acoustic presentation of native and non-native pseudowords, we provide a brief overview on previous ERP studies addressing this topic. Most studies adopted an oddball paradigm, consisting of a frequent standard stimulus and a rarely occurring target stimulus (deviant). This paradigm elicits a so-called mismatch response (MMR), larger in amplitude when the rare target stimulus is successfully detected as a deviant (Kushnerenko et al., 2002). Using this rationale, a differentiation between native and non-native stress patterns was attested in 4-month-old infants (Friederici et al., 2007, Friedrich et al., 2009). The MMR was more positive-going in infants which showed poor language skills at 2.5 years, while it was more negative-going with normal language abilities at this age (Friedrich et al., 2009). Even though a differentiation between native and non-native stress patterns can be measured by an MMR during early infancy, the polarity of this ERP component changes over the course of maturation. A longitudinal MMR study (Rivera-Gaxiola et al., 2005) in 7- and 11-month-old English monolingual infants contrasted two phonetically deviant syllables (i.e., one corresponding to the native language and one to the non-native language Spanish) to a standard syllable including phonemes common to both languages. 7-month-olds showing positive MMRs at this age revealed discriminatory abilities for both the native and the non-native contrast at 11 months, however, they showed a larger negativity for the native and a larger positivity for the non-native contrast. On the other hand, 7-month-olds showing negative MMRs at this age revealed a larger negativity for both native and non-native contrasts at 11 months. This effect was larger for the native compared to the non-native contrast. Apart from the interesting qualitative differences in polarity of the MMR reflecting different developmental patterns, the larger focus to the native contrast at 11 months is interpreted as a beginning neural commitment to the native language. Apart from oddball paradigms, other ERP studies presented various stimuli diverging in different linguistic aspects equally often within one experiment. Such a design provides a higher ecological validity as it resembles a natural language environment in which several stimuli belong to one linguistic category and infants have to extract the familiar rules and differentiate these from unfamiliar ones. One ERP study (Friedrich and Friederici, 2005) acoustically presented phonotactically native and non-native pseudowords combined with visual objects in a learning study to 12 -, 19-month-olds and adults. The ERPs elicited an early (∼200–400 ms) frontal negativity, larger for native compared to non-native pseudowords in 12- and 19-month-olds. This effect is interpreted as a phonotactic familiarity effect similar to a general familiarization effect around 200–650 ms to repeated unknown words found in several studies (Junge et al., 2014, Junge et al., 2012, Kooijman et al., 2013). An ERP study in 18-month-olds, showed a larger centro-parietally distributed N400 component (500–1000 ms) for native in contrast to non-native pseudowords (Steber and Rossi, 2020) which is usually also found in adult processing (Friedrich and Friederici, 2005, Rossi et al., 2011). In adults, the N400 is largely related to lexical access and is thus considered a more mature processing of phonotactics than a frontal familiarization negativity. Two associative learning studies in 6-month-olds, in which native and non-native pseudowords were trained with pictures of pseudo- and real objects also found a clear differentiation between native and non-native phonotactics between 650 and 850 ms but in these young infants a larger positivity was found for native versus non-native rules (Obrig et al., 2017, Richter et al., 2021). This effect contrasts with the above mentioned reversed effects of larger negative ERPs for native rules. This may be due to the age of subjects mirroring maturational mechanisms or may resemble other processing mechanisms. Similar to the MMR paradigm, also in an equal distribution paradigm a more negative-going wave for native phonotactics in older infants reflects more mature processing mechanisms, while a positive-going wave resembles initial tuning into language. Interestingly, however, a discrimination between native and non-native rules can be attested quite early but neural mechanisms seem to be still under development. A larger negativity for unfamiliar (or a positivity for familiar) items was also found in studies focusing on ERP components associated with attentional and memory related functions (e.g., Nc component or positive slow waves - PSW). Experimental designs in 7-month-olds in which either speech or non-speech utterances preceded an object presentation led to a more negative fronto-central Nc component around 300–900 ms for non-speech as well as a larger fronto-central positive slow wave around 1000–2000 ms for speech (Peykarjou et al., 2024, Peykarjou et al., 2020). The authors interpret the Nc as the allocation of attention to novel/unfamiliar stimuli while the PSW is assumed to reflect the access and updating of at least partially encoded memory representations. In the same line of reasoning, a more positive PSW (500–1000 ms) was found for emotional (i.e., happy/angry) and thus more familiar compared to neutral words in 7-month-olds (Grossmann et al., 2005). In familiarization paradigms in which objects were familiarized, 9-month-old infants showed a larger Nc (350–700 ms) for new compared to familiarized objects indicating novelty detection processes while a larger PSW (700–1500 ms) was attested for familiarized compared to new objects indicating successful recognition of encoded objects (Kopp and Lindenberger, 2011). Even though these studies do not directly compare native and non-native linguistic rules, the ERP components might be relevant also for such a comparison. In the present study we tested whether the prenatal exposure to prosodic (i.e., stress pattern) but not to phonotactic rules of the native language affects early neural processing mechanisms and whether a few months of postnatal experience with native phonotactic rules lead to similar or differential neural processes. Specifically, we ask whether the discriminatory ability between native and non-native linguistic rules is present in 3-month-old infants. We contrast linguistic rules that can be perceived already in utero, that is stress patterns, to linguistic rules that can be perceived only after birth, that is phonotactics, in order to find out whether the prenatal exposure elicits more mature neural processing. We wanted to immerse our infants into a highly natural setting, thus we created a variety of well-controlled monosyllabic pseudowords either conforming to German phonotactic or stress pattern rules (i.e., the subjects’ sole native language) or to Slovak phonotactic/French stress pattern rules (i.e., non-native condition). By controlling the non-native rules, we were able to a priori exclude infants additionally confronted with Slavic languages or French in their linguistic surroundings. We simultaneously assessed EEG and fNIRS. Thus, this cutting-edge combination bears the potential to gather insights from two different physiological signals, namely the electrophysiological and vascular signals. The EEG is capable to record fast temporal dynamics and polarity shifts showing up maturational trajectories while the fNIRS can highlight whether adult-like or less mature brain regions and hemispheric lateralization effects guide early discriminatory abilities between native and non-native linguistic rules. Considering the pioneering nature of the present study, several hypotheses will be put under investigation in order to provide an extensive understanding of the impact of pre- and postnatal language acquisition (please refer to Fig. 1 for a graphical presentation of the hypotheses). If prenatal exposure is indispensable, we will expect discriminatory abilities between native and non-native stress pattern only. If, however, three months of postnatal experience is enough, discriminatory abilities should also be present for phonotactic rules. With respect to maturational trajectories of underlying brain regions, we expect a more mature right-hemispheric activation in fronto-temporo-parietal brain areas for stress pattern and a more left-hemispheric recruitment for phonotactics. Immature processing should, however be related to a more bilateral recruitment of brain areas.Fig. 1Graphical presentation of the hypotheses put under investigation in the present study. At the top expected results are shown if prenatal exposure is the driving force, while at the bottom expected results for a stronger impact of postnatal exposure is shown. A. Hypotheses for the ERP data. B. Hypotheses for the fNIRS data. Graphical presentation of the hypotheses put under investigation in the present study. At the top expected results are shown if prenatal exposure is the driving force, while at the bottom expected results for a stronger impact of postnatal exposure is shown. A. Hypotheses for the ERP data. B. Hypotheses for the fNIRS data. With respect to electrophysiological signals, we consider different mechanisms potentially being present at this young developmental stage: a familiarity effect, an attentional and memory driven effect. These different mechanisms, occurring in successive time ranges, are known to be indexed by diverse ERP components, namely a frontal negativity, an Nc component and a positive slow wave (PSW), respectively. Similar to the hypotheses for fNIRS, we expect a differentiation between native and non-native stress pattern and no differentiation between native and non-native phonotactics, if prenatal exposure is the driving force, while a differentiation in both linguistic information is expected if 3 months of postnatal experience with the native language is enough. If a familiarity effect is present a larger negativity is expected for native compared to non-native rules, while Nc and PSW components should elicit more positive courses for familiar stimuli, thus native rules. Results of the present study will have important implications e.g. for infants at risk for later language disorders such as extremely premature infants, who are confronted precociously with the whole range of acoustic frequencies when the brain is potentially not mature enough to process all these various information types. Methods Participants In total, 42 3-month-old healthy infants took part in the present experiment. All subjects were raised monolingually with German as their only native language. Data of 32 subjects entered EEG analyses, while data of 33 subjects had sufficient data to enter statistical analyses of the fNIRS recordings. Data of 26 infants overlapped in both methods. Reasons for excluding subjects from further analyses included technical problems during data acquisition (EEG: 2 excluded; fNIRS: 6 excluded), precocious termination of the experiment due to infants’ discomfort (EEG: 2 excluded; fNIRS: 2 excluded) or too many movement artifacts during measurements (EEG: 6 excluded; 1 excluded). Attrition rates in our study were 23.8% for EEG data and 21.4% for fNIRS data, which is less than in other infant EEG (Stets et al., 2012) and fNIRS studies (Baek et al., 2023, Gervain et al., 2011). Demographic data of the study cohort are reported in Table 1.Table 1Demographic data of subjects entering EEG and fNIRS data analyses.EEGfNIRSn3233sex (female)1616mean age (days)9593weeks of gestation37–4237–42mean birth length (cm)51.150.7mean birth weight (g)3553.73523.6mean head circumference at birth (cm)35.335.4mean APGAR-index after 10 min9.8 (7 missings)9.9 (5 missings)no. of breastfeed infants31 out of 3231 out of 33no. of infants with chronic diseases0 out of 31 (1 missing)0 out of 32 (1 missing)no. of infants with visual impairments0 out of 31 (1 missing)0 out of 32 (1 missing)no. of infants with auditory impairments0 out of 31 (1 missing)0 out of 32 (1 missing)education of mother12 compulsory school12 compulsory school6 high school7 high school14 university14 universityeducation of father8 compulsory school7 compulsory school5 high school5 high school19 university21 universityno. of left-handed mothers1 out of 323 out of 32 (1 missing)no. of left-handed fathers1 out of 31 (1 missing)2 out of 32 (1 missing)no. of mothers with speech disorders0 out of 321 out of 33no. of fathers with speech disorders0 out of 320 out of 33 Table 1Demographic data of subjects entering EEG and fNIRS data analyses.EEGfNIRSn3233sex (female)1616mean age (days)9593weeks of gestation37–4237–42mean birth length (cm)51.150.7mean birth weight (g)3553.73523.6mean head circumference at birth (cm)35.335.4mean APGAR-index after 10 min9.8 (7 missings)9.9 (5 missings)no. of breastfeed infants31 out of 3231 out of 33no. of infants with chronic diseases0 out of 31 (1 missing)0 out of 32 (1 missing)no. of infants with visual impairments0 out of 31 (1 missing)0 out of 32 (1 missing)no. of infants with auditory impairments0 out of 31 (1 missing)0 out of 32 (1 missing)education of mother12 compulsory school12 compulsory school6 high school7 high school14 university14 universityeducation of father8 compulsory school7 compulsory school5 high school5 high school19 university21 universityno. of left-handed mothers1 out of 323 out of 32 (1 missing)no. of left-handed fathers1 out of 31 (1 missing)2 out of 32 (1 missing)no. of mothers with speech disorders0 out of 321 out of 33no. of fathers with speech disorders0 out of 320 out of 33 Demographic data of subjects entering EEG and fNIRS data analyses. Material The material consisted of 200 bisyllabic pseudowords of a consonant-consonant-vowel-consonant-vowel (CCVCV) structure without any semantic content. Pseudowords were assigned to four experimental conditions á 50 pseudowords: native phonotactics and native stress pattern (e.g., /brópo/); native phonotactics and non-native stress pattern (e.g., /bropó/); non-native phonotactics and native stress pattern (e.g., /bzópo/); non-native phonotactics and non-native stress pattern (e.g., /bzopó/) (Fig. 2B). All onset consonant clusters (first two consonants) of the pseudowords were either phonotactically native (e.g., /bropo/) or non-native (e.g., /bzopo/) with respect to German, the subjects’ native language. In total, 20 onset consonant clusters (10 native and 10 non-native) were constructed, each used in five different pseudowords. The five different pseudowords contained five different vowels. The first and second vowel per pseudoword were identical. Phonotactically non-native onset consonant clusters were phoneme sequences which are attested in the Slovak language. Slovak is very useful in this regard, since it provides a greater variability and more combination possibilities of consonants for word onsets than German (Hanulikova, 2009). It should be noted that all phonemes also those present in non-native pseudowords are part of the German phonetic inventory, but the combination is not attested in German. The frequency of consonant clusters was controlled for both native (German) and non-native (Slovak) pseudowords by assessing the lemma frequency per million using the German Celex corpus (Baayen et al., 1995) and the Slovak National Corpus (http://korpus.juls.savba.sk/index.en.html), respectively. A paired t-test between the frequency of German and Slovak consonant clusters did not show any significant difference [t(9) = 1.604, p = .147] (German consonant clusters: mean frequency per million: 810.91, SD: 509.07; Slovak consonant clusters: mean frequency per million: 375.55, SD: 632.52). In order to manipulate the stress pattern, each pseudoword was recorded either with iambic or trochaic stress pattern. The stress on the first vowel, that is trochaic, is the most common stress pattern in German, stress on the second vowel is less common in German and mimics non-native stress patterns e.g. the iambic pattern primarily used in two-syllabic French words. Different acoustic parameters confirmed this differential stress pattern (Table 2). Paired t-tests for duration (ms) and maximum pitch (Hz) between the first and second vowel revealed a longer duration and larger pitch for the first vowel of the trochaic German stress pattern while the second vowel was longer and larger in pitch for the iamb French stress pattern. Intensity (dB) revealed a louder first vowel in each condition resembling the loud-first principle of the iambic-trochaic law (Bolton, 1894, Hayes, 1995). Pseudowords were presented in a pseudorandomized manner with the following criteria: maximally three phonotactically native/non-native and maximally three stress pattern native/non-native in succession and an equal amount of each condition per experimental half. Five different pseudorandomized versions were created.Fig. 2Experimental setup A. Depiction of an infant sitting in front of a monitor and loudspeakers while simultaneous EEG and fNIRS is assessed. B. Trial sequence (top) and examples of the four experimental conditions (bottom).Table 2Acoustic parameters of experimental stimuli.DurationExperimental conditiondirection effectvowel 1 (ms)vowel 2 (ms)pstress pattern + / phonotactics +vowel 1 > vowel 2684.4382.6< .001stress pattern + / phonotactics -vowel 1 > vowel 2573.0363.1< .001stress pattern - / phonotactics +vowel 2 > vowel 1217.1694.7< .001stress pattern - / phonotactics -vowel 2 > vowel 1189.0693.3< .001Pitch max.Experimental conditiondirection effectvowel 1 (Hz)vowel 2 (Hz)pstress pattern + / phonotactics +vowel 1 > vowel 2329.9182.5< .001stress pattern + / phonotactics -vowel 1 > vowel 2336.4199.1< .001stress pattern - / phonotactics +vowel 2 > vowel 1230.1333.2< .001stress pattern - / phonotactics -vowel 2 > vowel 1268.1329.2< .001IntensityExperimental conditiondirection effectvowel 1 (dB)vowel 2 (dB)pstress pattern + / phonotactics +vowel 1 > vowel 280.671.8< .001stress pattern + / phonotactics -vowel 1 > vowel 280.871.5< .001stress pattern - / phonotactics +vowel 1 > vowel 282.678.0< .001stress pattern - / phonotactics -vowel 1 > vowel 282.577.6< .001 Experimental setup A. Depiction of an infant sitting in front of a monitor and loudspeakers while simultaneous EEG and fNIRS is assessed. B. Trial sequence (top) and examples of the four experimental conditions (bottom). Table 2Acoustic parameters of experimental stimuli.DurationExperimental conditiondirection effectvowel 1 (ms)vowel 2 (ms)pstress pattern + / phonotactics +vowel 1 > vowel 2684.4382.6< .001stress pattern + / phonotactics -vowel 1 > vowel 2573.0363.1< .001stress pattern - / phonotactics +vowel 2 > vowel 1217.1694.7< .001stress pattern - / phonotactics -vowel 2 > vowel 1189.0693.3< .001Pitch max.Experimental conditiondirection effectvowel 1 (Hz)vowel 2 (Hz)pstress pattern + / phonotactics +vowel 1 > vowel 2329.9182.5< .001stress pattern + / phonotactics -vowel 1 > vowel 2336.4199.1< .001stress pattern - / phonotactics +vowel 2 > vowel 1230.1333.2< .001stress pattern - / phonotactics -vowel 2 > vowel 1268.1329.2< .001IntensityExperimental conditiondirection effectvowel 1 (dB)vowel 2 (dB)pstress pattern + / phonotactics +vowel 1 > vowel 280.671.8< .001stress pattern + / phonotactics -vowel 1 > vowel 280.871.5< .001stress pattern - / phonotactics +vowel 1 > vowel 282.678.0< .001stress pattern - / phonotactics -vowel 1 > vowel 282.577.6< .001 Acoustic parameters of experimental stimuli. All stimuli were spoken by a German/Slovak bilingual female speaker in a soundproof chamber and recorded digitally with 16 bits and 44 kHz. In order to direct the infants’ attention to the stimuli, pseudowords were spoken in infant-directed-speech (IDS) mode based on imagined interactions with infants. IDS is a standard procedure in infant research and characterized by a higher pitch and a slower, clearer speech mode (Soderstrom, 2007). Procedure In the present study, we assessed neural activity simultaneously with electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS). The former method captures electrophysiological activity representing fast processing mechanisms in the millisecond range, whereas the latter method relying on the slower hemodynamic response is capable to identify brain areas supporting different processing mechanisms with a good topographical precision. Combining both methods provides several advantages, especially when investigating auditory fine-grained stimuli, as both methods are soundless, non-invasive and suitable for infants, do not interfere with each other, and provide a quite natural setting (Rossi et al., 2012). The study was approved by the local ethical committee (No. 030–11–24012011, 23.02.2011). All procedures were performed in compliance with relevant laws and institutional guidelines. Privacy rights of participating subjects and their parents have been observed and written informed consent was obtained from all parents prior to measurements. During the experiment, infants sat in a sound-proof booth on their parent’s lap 1 m in front of a computer monitor and stereo speakers (Fig. 2A). To enable a simultaneous measurement of both methods, infants wore elastic EEG caps (EasyCap GmbH, Herrsching, Germany) in which both EEG electrodes as well as fNIRS optodes were integrated. Pseudowords were acoustically presented at about 70 dB via loudspeakers centered in front of the infant. To enhance the infants’ attention to the experiment, a silent children’s cartoon was shown on the computer screen during experimental session. Out of the 32 infants entering EEG data analysis, 20 infants did not look at the screen at all, while only 12 infants watched the silent cartoon with a range from 1 to 13 min. (average looking time of 3.4 min.). For fNIRS data, out of the 33 analyzed data sets, 22 infants did not look at the screen at all, while only 11 infants watched the silent cartoon with a range from 1 to 13 min. (average looking time of 3.9 min.). Due to the event-related design, each pseudoword was presented in a time slot of 2000 ms followed by a variable inter-stimulus-intervals (ISIs) (mean = 6000 ms; range 4000–8000 ms) (Fig. 2B). This variable ISI allowed to adjust the experimental design to the requirements of the rather slow hemodynamic response function (hrf) measured by the fNIRS. Usually, vascular responses reach their maximum of activation at around 5 s after stimulus onset with the activation returning to baseline after 15–20 s (Huettel, 2009). Therefore, variable ISIs minimize the risk of overlapping of hemodynamic responses in event-related designs. The experiment lasted about 27 min. EEG recordings EEG was recorded from 19 AgAgCl active electrodes (BrainProducts GmbH, Gilching, Germany) fastened by an elastic EEG cap at the following positions of the 10–20 system (Jasper, 1958, Sharbrough, 1991): F7/F8, F3/F4, FT7/FT8, FC3/FC4, T7/T8, C3/C4, CP3/CP4, P3/P4, Fz, Pz, and Cz (Fig. 3). Vertical and horizontal electrooculogram were recorded from FP2 and F10. The online reference electrode was positioned at the left mastoid (TP9) and an electrode at the right mastoid (TP10) was recorded for further re-referencing during offline analyses. The ground electrode was set at AFz. Electrode impedance was kept below 10 kΩ (actiCAP Control, Brain Products GmbH, Gilching, Germany). The EEG signal was measured by means of BrainVision Recorder (Brain Products GmbH, Gilching, Germany) software with a sampling frequency of 1000 Hz (amplified between 0.016 and 450 Hz). An anti-aliasing filter with a cut-off at 250 Hz (slope: 30 dB/oct) was applied prior to analogue-to-digital conversion.Fig. 3Simultaneous EEG-fNIRS configuration. A. 24 EEG electrodes (white circles) including an online reference (TP9), an electrode (TP10) for offline re-referencing, the ground electrode (AFz), vertical (FP2) and horizontal (FP10) electrooculogram and 6 NIRS light emitters (stars) and 8 light NIRS detectors (dots) resulting in 16 channels (ellipses) are assessed. L = left; R = right; F = frontal; T = temporal; P = parietal; i = inferior; s = superior B. EEG-electrode configuration used for statistical analyses including lateral regions of interest (ROIs) and midline single electrodes. Simultaneous EEG-fNIRS configuration. A. 24 EEG electrodes (white circles) including an online reference (TP9), an electrode (TP10) for offline re-referencing, the ground electrode (AFz), vertical (FP2) and horizontal (FP10) electrooculogram and 6 NIRS light emitters (stars) and 8 light NIRS detectors (dots) resulting in 16 channels (ellipses) are assessed. L = left; R = right; F = frontal; T = temporal; P = parietal; i = inferior; s = superior B. EEG-electrode configuration used for statistical analyses including lateral regions of interest (ROIs) and midline single electrodes. fNIRS recordings The hemodynamic response to neural activation patterns was measured by means of fNIRS. The method assesses the response by recording changes in light attenuation which can be converted into concentration changes of oxygenated [oxy-Hb] and deoxygenated hemoglobin [deoxy-Hb] based on a modified Beer-Lambert law (Cope et al., 1989). The area sampled depends on probe arrangement. In the current setup, the area most strongly contributing to the measured changes can be assumed between each pair of light-emitting and light-detecting probes. In brief, the physiological basis underlying fNIRS is that an increase in neuronal activation is tightly coupled to an increase in regional cerebral blood flow (Logothetis and Wandell, 2004, Uludağ et al., 2004). Since the blood-flow increase is disproportionally large compared to the increase in oxygen consumption a typical response over an activated area is an increase in [oxy-Hb] as well as a decrease in [deoxy-Hb] representing focal hyperoxygenation (Fox and Raichle, 1986). We used an fNIRS system (NIRScout, NIRx Medizintechnik GmbH, Berlin) operating with wavelengths at 860 and 730 nm was used. Sampling rate was set at 12.5 Hz. 6 light emitters and 8 light detectors were positioned over bilateral fronto-temporo-parietal brain areas. Inter-optode distance was 2.5 cm. This emitter-detector configuration resulted in the assessment of 8 channels per hemisphere, which were distributed over frontal, frontotemporal, temporal, temporoparietal and parietal brain regions (Fig. 3). Positioning of fNIRS light emitters and detectors was based on the standard EEG 10–20 electrode positioning system (Jasper, 1958, Sharbrough, 1991), projected to underlying anatomical structures of interest (Kabdebon et al., 2014, Okamoto et al., 2004). Data analyses EEG data analyses EEG data was offline low-pass-filtered (high cutoff: 45 Hz: slope: 12 dB/oct; Butterworth zero-phase filter). Further, a 50 Hz notch filter was applied. Data was then segmented from −200–1300 ms with 0 ms representing pseudoword onset. Artifact-contaminated electrodes were rejected manually from each segment via visual inspection by an experienced researcher. Only infants with at least 10 artifact-free trials per condition in at least half of the electrodes were included in the final analyses (please refer to Forssman and Vallberg, 2026; Friedrich et al., 2022; Junge et al., 2014, Junge et al., 2012; Peykarjou et al., 2024 for similar procedures in infant studies). The number of rejected trials did not differed across the four conditions [F(3,93)= 2.501, p > .050] and resulted in the following numbers of retained trials: native phonotactics – native stress pattern: 23.91; native phonotactics – non-native stress pattern: 25.52; non-native phonotactics – native stress pattern: 24.79; non-native phonotactics – non-native stress pattern: 25.16. Afterwards, data was re-referenced to averaged mastoids (TP9, TP10) and a pre-stimulus baseline of 200 ms was applied. Event-related brain potentials (ERPs) were calculated by averaging artifact-free trials for each of the 4 conditions (native phonotactics – native stress pattern; native phonotactics – non-native stress pattern; non-native phonotactics – native stress pattern; non-native phonotactics – non-native stress pattern) and each subject. Subsequently, grand averages across subjects were calculated by applying an 8-Hz low-pass filter for presentation purposes only (a figure of the grand average with a 20-Hz low-pass filter is shown in Figure S1 in Supplementary Material). Time windows for final statistical analyses were selected based on cluster-based permutation tests, visual inspection of the grand averages as well as on literature. Cluster-based permutation tests were calculated with BESA Statistics 2.1 (Bornfleth et al., 2020) (within-group ANOVA with 4 levels; cluster Alpha: 0.05; number of permutations: 1000) and identified one large time window between 607 and 1211 ms (cluster value = 8699.29; p = .001). Visual inspection of grand averages indicated, however, that this time window could reflect two processes as at 800 ms the course of ERP curves changes from a clear positive course into a more negative one. Due to this reason and because we a priori aimed at testing whether an earlier Nc component (Kopp and Lindenberger, 2011, Peykarjou et al., 2024, Peykarjou et al., 2020) and a later PSW component (Grossmann et al., 2005, Kopp and Lindenberger, 2011, Peykarjou et al., 2024, Peykarjou et al., 2020) are present in our data, we split the time window into two segments (607–799 ms and 800–1211 ms). Because an earlier familiarity negativity might also underlie our processes and inspection of grand averages suggests a potential effect between 400 and 606 ms, we also investigated this earlier time window. Thus, the following time windows were statistically analyzed: 400–606 ms (frontal negativity), 607–799 ms (Nc component), 800–1211 ms (positive slow wave PSW). Because topographical localization of EEG is only rough, we decided to merge lateral electrodes to 6 regions of interest (ROIs). The following lateral ROIs were defined: left fronto-temporal (F7, FT7), right fronto-temporal (F8, FT8), left fronto-central (FC3, C3), right fronto-central (FC4, C4), left centro-parietal (C3, CP3), and right centro-parietal (C4, CP4) (Fig. 1B). We checked whether variables were normally distributed by means of Kolmogorov-Smirnov tests. Because the majority of variables were normally distributed (400–606 ms: 75.0%; 607–799 ms: 75.0%; 800–1211 ms: 61.1%), parametric tests were used for subsequent analyses. We then performed a four-way repeated-measures ANOVA for each time window with the within-subject factors Phonotactics (native vs. non-native), Stress Pattern (native vs. non-native), Region (fronto-temporal vs. fronto-central vs. centro-parietal), and Hemisphere (left vs. right) for lateral electrodes. For the midline electrodes, a three-way ANOVA (Phonotactics x Stress Pattern x Electrodes Fz, Cz, Pz) was performed. Significance level was set at p ≤ .05 and adjusted according to Bonferroni. Whenever Mauchly’s test of sphericity became significant, Greenhouse-Geisser correction (Greenhouse and Geisser, 1959) was applied. In time windows showing statistically significant effects, additionally peak-latencies were compared between conditions by means of a one-way ANOVA with the factor condition (4 experimental conditions as levels). In case of a significant effect, Bonferroni-adjusted post-hoc t-tests between conditions were performed. Significance level of p ≤ .05 is assumed. fNIRS data analyses A MATLAB (MathWorks, Inc., MA, USA, Version R2018a)-based program nilab2 (written by Stefan Paul Koch, Charité University Medicine, Berlin, Germany) was used for analyzing fNIRS data. The following preprocessing steps were applied to fNIRS data: setting the stimulus duration to 2 s; cutting pauses and segments possibly containing artifacts but no stimulation markers to obtain a stable baseline; convert light attenuation at the two wavelengths into concentration changes (mmol/l) of [oxy-Hb] and [deoxy-Hb] by means of the modified Beer Lambert function (Cope et al., 1989); artifact correction adopting a linear spline interpolation approach which has been frequently used in adult and infant research (Grossmann et al., 2010, Obrig et al., 2017) and preferred compared to methods rejecting artifact-contaminated segments (Brigadoi et al., 2014, Di Lorenzo et al., 2019) as it allows keeping a large amount of data (Brigadoi et al., 2014, Scholkmann et al., 2010); applying a 0.3 Hz low-pass filter (Butterworth, third order) to attenuate high-frequency artifacts mainly arising from heartbeat. A general-linear-model (GLM) approach was used because (1) it can better depict the complex hemodynamic response, (2) it can better handle with the partial overlap of hemodynamic responses (hrfs) resulting from the relatively short stimulation and ISIs integrated in an event-related design (de Roever et al., 2018, Issard and Gervain, 2018) and (3) resulting Beta values are better comparable to existing fMRI data where GLM usage is standard procedure. For the GLM, data were correlated with a predictor generated by convolving the boxcar function of the stimulus design with the canonical hrf (Boynton et al., 2012, Huettel, 2009) including one predictor for each of the 4 different conditions (native phonotactics – native stress pattern; native phonotactics – non-native stress pattern; non-native phonotactics – native stress pattern; non-native phonotactics – non-native stress pattern). The on-condition was set at a stimulation period of 2 s (i.e., on-condition) while the off-condition (i.e., silence) represented the resting periods in between of the acoustic stimuli (i.e., ISIs). A high-pass filter of 20 s was applied to remove drifts and slow fluctuations (for similar procedures please see Altvater-Mackensen and Grossmann, 2018, Altvater-Mackensen and Grossmann, 2016; Koch et al., 2006; Obrig et al., 2017; Steber and Rossi, 2020; Telkemeyer et al., 2009) and a canonical hrf (peak at 5 s and further 15 s for return to baseline) was used to get Beta-values, which were used for statistical analyses. Both, a decrease in [deoxy-Hb] and an increase in [oxy-Hb] reflect increased activations (Lloyd-Fox et al., 2010, Obrig and Villringer, 2003). We checked whether variables were normally distributed by means of Kolmogorov-Smirnov tests. Because the majority of variables were normally distributed (oxy-Hb: 57.8%; deoxy-Hb: 54.7%), parametric tests were used for subsequent analyses. We first performed a four-way repeated-measures ANOVA similar to the EEG analyses (within-subject factors Phonotactics, Stress Pattern, Region, and Hemisphere). However, no significant results were obtained, thus, two-way repeated measures ANOVAs with the within-subject factors Phonotactics (native vs. non-native) and Stress Pattern (native vs. non-native) were performed on each fNIRS channel. Each analyses was performed for [oxy-Hb] and [deoxy-Hb], separately. We did not opt to build ROIs for fNIRS data as we wanted to keep the more accurate topographical resolution of the fNIRS methodology as fine-grained as possible. Significance level was set at p ≤ .050 and adjusted according to Bonferroni within each ANOVA and across fNIRS channels. Whenever Mauchly’s test of sphericity became significant Greenhouse-Geisser correction(Greenhouse and Geisser, 1959) was applied. Results The results detailed below show that electrophysiologically there was an effect of phonotactics over lateral electrodes between 607 and 799 ms and an effect of stress pattern over lateral and midline electrodes between 800 and 1211 ms after stimulus onset. The phonotactic effect was driven by more negative ERPs for non-native versus native phonotactic rules while the stress pattern effect elicited more positive ERPs for native trochaic compared to non-native iamb stress pattern. Analysis of the hemodynamic responses (fNIRS) showed a bilateral fronto-temporal larger activation for non-native compared to native phonotactics (in [oxy-Hb] and [deoxy-Hb]) as well as a right temporal larger activation for non-native compared to native stress patterns (in [oxy-Hb]). EEG results ERP grand average curves are depicted in Fig. 4. Additionally, bee-swarm plots depicting individual data points and means for the four experimental conditions, separately, are shown in Figure S2 in the Supplementary Material.Fig. 4ERP results. Grand averages for the four experimental conditions for all electrodes. An 8-Hz low-pass filter was applied for presentation purposes only. Yellow boxes on the exemplary electrode Cz indicate the three time windows analyzed: 400–606 ms (frontal negativity), 607–799 ms (Nc component), 800–1211 ms (positive slow wave PSW). Statistical analyses revealed a main effect for phonotactics on all lateral electrodes in the time window 607–799 ms indicating a larger Nc component for non-native rules (or a larger positivity for native) and a main effect for stress pattern on all lateral and midline electrodes in the time window 800–1211 ms indicating larger PSWs for native vs. non-native stress pattern or larger negative slow waves (NSW) for non-native stress pattern (* p < .050). Negative polarity is plotted upwards. Topographical maps are shown on the right. ERP results. Grand averages for the four experimental conditions for all electrodes. An 8-Hz low-pass filter was applied for presentation purposes only. Yellow boxes on the exemplary electrode Cz indicate the three time windows analyzed: 400–606 ms (frontal negativity), 607–799 ms (Nc component), 800–1211 ms (positive slow wave PSW). Statistical analyses revealed a main effect for phonotactics on all lateral electrodes in the time window 607–799 ms indicating a larger Nc component for non-native rules (or a larger positivity for native) and a main effect for stress pattern on all lateral and midline electrodes in the time window 800–1211 ms indicating larger PSWs for native vs. non-native stress pattern or larger negative slow waves (NSW) for non-native stress pattern (* p < .050). Negative polarity is plotted upwards. Topographical maps are shown on the right. 400–606 ms Neither the four-way ANOVA for lateral electrodes nor the three-way ANOVA for midline electrodes resulted in any significant main effect or interaction. 607–799 ms The four-way ANOVA for lateral electrodes revealed a significant main effect Phonotactics (Table 3) suggesting a larger negativity for non-native compared to native phonotactic rules. No significant effect resulted from the three-way ANOVA for midline electrodes.Table 3EEG and fNIRS results.EEG400–606 msLocationeffectdfFpηp2post-hoc direction of effect (negativity)LateralnsMidlinens607–799 msLocationeffectdfFpηp2post-hoc direction of effect (negativity)LateralP1,294.443.044.13non-native P (13.08 µV) > native P (17.55 µV)Midlinens800–1211 msLocationeffectdfFpηp2post-hoc direction of effect (positivity)LateralS1,299.011.005.24native S (14.30 µV) > non-native S (8.37 µV)MidlineS1,296.255.018.17native S (17.60 µV) > non-native S (11.26 µV)fNIRS[oxy-Hb]LocationeffectdfFpηp2post-hoc direction of effect (larger activation)L front infP x S1,3210.712.003.25nsL front supP1,324.411.044.12non-native P (−1 ×10−5 mmol/l) > native P (−3 ×10−4 mmol/l)R temp infS1,327.063.012.18non-native S (6 ×10−4 mmol/l) > native S (1 ×10−4 mmol/l)[deoxy-Hb]LocationeffectdfFpηp2post-hoc direction of effect (larger activation)R fr-temp infP1,324.804.036.13non-native P (2 ×10−4 mmol/l) > native P (6 ×10−4 mmol/l)R fr-temp supP1,324.514.041.12non-native P (−8 ×10−4 mmol/l) > native P (−3 ×10−4 mmol/l)effect indicates main effects or interaction resulting from the ANOVAs.P indicates a main effect of phonotactics.S indicates a main effect of stress pattern.P x S indicates an interaction between phonotactics and stress pattern.L = left; R = right; front = frontal; temp = temporal; fr-temp = fronto-temporal; inf = inferior; sup = superior.ηp2 indicates effect sizes in partial eta squared: small ∼.01; medium ∼.06; large ∼.14. Table 3EEG and fNIRS results.EEG400–606 msLocationeffectdfFpηp2post-hoc direction of effect (negativity)LateralnsMidlinens607–799 msLocationeffectdfFpηp2post-hoc direction of effect (negativity)LateralP1,294.443.044.13non-native P (13.08 µV) > native P (17.55 µV)Midlinens800–1211 msLocationeffectdfFpηp2post-hoc direction of effect (positivity)LateralS1,299.011.005.24native S (14.30 µV) > non-native S (8.37 µV)MidlineS1,296.255.018.17native S (17.60 µV) > non-native S (11.26 µV)fNIRS[oxy-Hb]LocationeffectdfFpηp2post-hoc direction of effect (larger activation)L front infP x S1,3210.712.003.25nsL front supP1,324.411.044.12non-native P (−1 ×10−5 mmol/l) > native P (−3 ×10−4 mmol/l)R temp infS1,327.063.012.18non-native S (6 ×10−4 mmol/l) > native S (1 ×10−4 mmol/l)[deoxy-Hb]LocationeffectdfFpηp2post-hoc direction of effect (larger activation)R fr-temp infP1,324.804.036.13non-native P (2 ×10−4 mmol/l) > native P (6 ×10−4 mmol/l)R fr-temp supP1,324.514.041.12non-native P (−8 ×10−4 mmol/l) > native P (−3 ×10−4 mmol/l)effect indicates main effects or interaction resulting from the ANOVAs.P indicates a main effect of phonotactics.S indicates a main effect of stress pattern.P x S indicates an interaction between phonotactics and stress pattern.L = left; R = right; front = frontal; temp = temporal; fr-temp = fronto-temporal; inf = inferior; sup = superior.ηp2 indicates effect sizes in partial eta squared: small ∼.01; medium ∼.06; large ∼.14. EEG and fNIRS results. effect indicates main effects or interaction resulting from the ANOVAs. P indicates a main effect of phonotactics. S indicates a main effect of stress pattern. P x S indicates an interaction between phonotactics and stress pattern. L = left; R = right; front = frontal; temp = temporal; fr-temp = fronto-temporal; inf = inferior; sup = superior. ηp2 indicates effect sizes in partial eta squared: small ∼.01; medium ∼.06; large ∼.14. Peak latencies did not show differences between the four experimental conditions [F(3,93)= 1.586, p > .050] (native phonotactics – native stress pattern: 667.56 ms; native phonotactics – non-native stress pattern: 679.88 ms; non-native phonotactics – native stress pattern: 694.22 ms; non-native phonotactics – non-native stress pattern: 702.53 ms). 800–1211 ms The four-way ANOVA for lateral electrodes revealed a significant main effect Stress Pattern (Table 3) suggesting a larger positivity for native compared to non-native stress pattern. A similar effect (i.e., native more positive than non-native) was also achieved by the three-way ANOVA for midline electrodes indicated by a significant main effect Stress Pattern (Table 3). Peak latencies did not show differences between the four experimental conditions [F(3,93)= 2.567, p > .050] (native phonotactics – native stress pattern: 991.72 ms; native phonotactics – non-native stress pattern: 909.60 ms; non-native phonotactics – native stress pattern: 939.50 ms; non-native phonotactics – non-native stress pattern: 991.81 ms). fNIRS results fNIRS results are depicted in Fig. 5. Additionally, bee-swarm plots depicting individual data points and means for the four experimental conditions, separately, are shown in Figure S3 in the Supplementary Material.Fig. 5fNIRS results ([oxy-Hb] and [deoxy-Hb]) for significant channels showing mean values and standard errors of mean. Colored ellipses in the fNIRS configuration at the bottom left of the figure indicate channels with significant main effects Phonotactics (dark blue) and Stress Pattern (light blue) in the ANOVAs. A larger activation is plotted upwards. plus indicates native rules; minus indicates non-native rules; LFs = left superior frontal; RFTi = right inferior fronto-temporal; RFTs = right superior fronto-temporal; RTi = right inferior temporal. * indicates significant effects (p < .050). fNIRS results ([oxy-Hb] and [deoxy-Hb]) for significant channels showing mean values and standard errors of mean. Colored ellipses in the fNIRS configuration at the bottom left of the figure indicate channels with significant main effects Phonotactics (dark blue) and Stress Pattern (light blue) in the ANOVAs. A larger activation is plotted upwards. plus indicates native rules; minus indicates non-native rules; LFs = left superior frontal; RFTi = right inferior fronto-temporal; RFTs = right superior fronto-temporal; RTi = right inferior temporal. * indicates significant effects (p < .050). The two-way ANOVA for [oxy-Hb] revealed a significant interaction Phonotactics x Stress Pattern at left frontal inferior regions, a significant main effect Phonotactics at left superior frontal regions and a significant main effect Stress Pattern at right inferior temporal regions (Table 3). The posthoc t-tests for the interaction did not reach significance after Bonferroni correction. The main effect Phonotactics revealed a larger left superior frontal activation for pseudowords with non-native compared to a native phonotactic rules, while the main effect Stress Pattern revealed a reliable larger activation for pseudowords with a non-native compared to a native stress pattern at right inferior temporal sites. The two-way ANOVA for [deoxy-Hb] revealed a significant main effect Phonotactics at right inferior and superior fronto-temporal sites indicating larger activations for non-native compared to native phonotactic rules (Table 3). Discussion The present study investigated how prenatal and postnatal acoustic experience affects the perception of native and non-native linguistic regularities in 3-month-old infants with German as their only native language. Two manipulations of native vs. non-native regularities were tested: (1) stress pattern variations in which besides the typical German trochaic, an iambic stress pattern (i.e., stress on the second syllable in two-syllabic words) conforming to the non-native language French was used and (2) phonotactic variations in which the combination of onset consonant clusters did or did not correspond to the native language German with non-native rules conforming to Slovak phonotactic rules. Native stress pattern is already perceived in the last trimester of pregnancy (i.e., prenatal) as fetuses at that time have functioning auditory abilities and slow acoustic variations can also be perceived in the low-pass filtering environment of the uterus. On the contrary, phonotactic features can be learned only after birth (i.e., postnatal) when infants perceive the whole frequency spectrum including fast acoustic transitions such as the transition between single phonemes. A multimethodological approach was adopted, assessing electrophysiological and vascular signals by means of EEG and fNIRS, respectively. EEG results revealed a clear differentiation between native and non-native rules, evidenced by large effect sizes. In particular, a larger negativity for phonotactically non-native compared to native pseudowords was found between 606 and 799 ms. From 800 ms onwards, a larger positivity for native pseudowords differed from more negative going non-native stress pattern manipulations. This late effect can, however, also be considered as a larger negativity for non-native compared to native stress pattern rules. The latter consideration would resemble a similar direction of effects as observed for the earlier phonotactics effect. Importantly, 3-month-olds can differentiate native from non-native rules, which are putatively acquired either starting in the prenatal (i.e. since ∼6 months) or postnatal period (i.e. since ∼3 months). The question arises whether the processes underlying the phonotactic and stress pattern effect resemble similar or diverse mechanisms. From a timing perspective, the two processes occur sequentially in time suggesting two different processes. Several ERP studies contrasting deviant/novel to standard/familiar stimuli in early infancy, found fronto-centrally distributed negativities, larger in amplitude for the most deviant/novel/unfamiliar stimulus. This component was differently labelled as Nc (negative central) (Kopp and Lindenberger, 2011, Peykarjou et al., 2024, Peykarjou et al., 2020, Richards, 2003), RON (reorienting negativity) (Schröger and Wolff, 1998), LDN (late discriminative negativity) (Cheour et al., 2001), LN (late frontal negativity) (Dehaene-Lambertz and Dehaene, 1994, Kushnerenko et al., 2007, Kushnerenko et al., 2002) or NSW (negative slow wave) (deRegnier et al., 2002, Mai et al., 2012). Common to all these different terminologies is that they describe a negative ERP component between 400 and 1000 ms with a fronto-central topography. Especially relevant for our study are findings in newborns and 2-month-olds (deRegnier et al., 2002, Mai et al., 2012) showing a more negative NSW for a stranger’s voice producing the word “baby” compared to the familiar maternal voice. Authors interpreted this component as the detection of novelty and auditory recognition memory. Further, two previous studies comparing native and non-native phonotactic rules in 6-month-olds in a pseudoword-object training, also found a larger negativity for non-native rules (Obrig et al., 2017, Richter et al., 2021) in a similar time window. Thus, these findings point to a top-down attentional orienting and perceptual processing of novel stimuli. Interestingly, the Nc peak was found to decrease during the first 6 months of life (from ∼ 1000 ms in newborns to 700 ms in 6-month-olds). This would additionally align with our early ERP findings for phonotactics starting to deviate from around 600 ms onwards. The late stress pattern effect from 800 ms onwards might be interpreted either as a positive slow wave (PSW) with larger positive amplitudes for native rules, reflecting successful recognition of encoded rules or as a negative slow wave (NSW) with larger negative amplitudes for non-native rules, suggesting a late novelty effect, similar to the earlier effect for phonotactics. The former interpretation is in line with previous studies revealing a larger PSW for emotional compared to neutral stimuli (Grossmann et al., 2005) or for familiarized compared to new objects (Kopp and Lindenberger, 2011). The latter interpretation is however also plausible as previous studies in newborns and 2-month-olds (deRegnier et al., 2002, Mai et al., 2012) found similar late effects indicating novelty effects for unfamiliar compared to familiar voices. Considering a similar fronto-central distribution of the early phonotactic and late stress pattern effect as can be seen from the topographical maps in Fig. 4, also similar novelty and memory-related process could be assumed. Indeed, without the capacity to identify novel patterns, the recognition of familiar already encoded patterns is not possible. Thus, attention-driven novelty detection and recognition memory work in concert. The later onset of the stress pattern effect compared to the phonotactics effect is explainable with the later identification in the pseudowords. While phonotactic manipulations occur in the first two phonemes of a pseudowords, the identification of a native or non-native stress pattern requires listening to the whole pseudoword because the non-native iambic stress pattern is characterized by the stress on the last vowel of the pseudoword. In our early time window between 400 and 606 ms, no significant differences were found. Several infant studies found familiarity effects indexed by negativities/N200–500 (Friedrich and Friederici, 2005, Junge et al., 2014, Junge et al., 2012, Kooijman et al., 2013) in a similar time range, resulting in a larger negativity for the more familiar stimuli. Usually, infants of these studies were older (7–19 months), thus, such a familiarity effect seems not to be present in infants aged 3 months. Summarizing the ERP results, our findings show processes driven by attentional and memory-related mechanisms. The fact that a clear differentiation between native and non-native rules is present for both phonotactics and stress pattern underlines the impact of postnatal acoustic exposure. fNIRS results show a reliable difference between native and non-native phonotactics as well as stress patterns, supported by large effect sizes. More specifically, a larger activation was observed for non-native compared to native phonotactic rules in bilateral fronto-temporal areas, while a larger activation for non-native compared to native stress patterns was attested in right temporal areas. Similar to the EEG results, both effects seem to be driven by novelty detection and recognition memory, in that the most unfamiliar condition leads to the largest activation. This pattern of results is in accord with previous fNIRS studies in newborns and 3-month-olds showing larger activations, primarily in prefrontal regions, for unfamiliar/new stimuli compared to familiar ones (Nakano et al., 2009, René et al., 2025). The right temporal location of the stress pattern effect fits with studies finding non-native prosodic aspects to be more activated in this hemisphere while native cues tended to activate more left-hemispheric areas (René et al., 2025). Interestingly, previous studies finding similar lateralized results already in newborn infants used longer-lasting auditory contexts such as sentences or whole stories (Martinez-Alvarez et al., 2023, Minagawa-Kawai et al., 2011, Vannasing et al., 2016) (for fNIRS evidence on a wide-spread bilateral activation, larger for native than non-native prosodic aspects in sentences in newborns please refer to Ma et al., 2025), while studies contrasting single words revealed more bilateral activations due to the more difficult processing context (Abboub et al., 2016). The present study provides evidence, that also in a more complex single word context not only a differentiation between native and non-native stress pattern rules is possible but that the recruited brain areas show a right-lateralized and thus mature adult-like response. The right hemispheric lateralization fits with the rationale of a right-hemispheric dominance for slow prosodic features, however, usually the most familiar prosodic pattern elicits such a right hemispheric activation (Homae et al., 2006, Poeppel, 2003, Telkemeyer et al., 2009). Because in our study the right-hemispheric activation was more prominent for the most unfamiliar (i.e., non-native) rules, we opt for an interpretation similar to the ERP results. Thus, we suggest that the response indicates a novelty effect with attentional and memory retrieval mechanisms. Studies in adults but also infants found a larger activation in right frontal and parietal areas when information was retrieved from memory and new information was thus identified (Benavides-Varela et al., 2012, Fletcher and Henson, 2001, Rugg et al., 1996, Shallice et al., 1994). With respect to phonotactic rules a similar novelty effect indexed by larger bilateral fronto-temporal activations was found in the present study. This is in line with an fNIRS study in newborns finding larger bilateral activations for phonotactically ill-formed compared to well-formed pseudowords (Gómez et al., 2014). Interestingly, a study in 6-month-olds revealed larger bilateral activations for native phonotactic rules (Obrig et al., 2017), suggesting a shift from an attention-driven novelty effect in younger to a more linguistic process in older infants. The bilateral activation associated with phonotactic processing resembles an immature process, in contrast to an adult-like left-hemispheric activation (Rossi et al., 2011). The increased activation for non-native phonotactic and stress pattern rules might alternatively indicate more processing effort needed for unfamiliar/novel rules. Because native rules elicit a reduced activation, a habituation effect may be causal, that is, less effort is needed for rules that were already often heard. This corresponds to a repetition suppression effect (Friston, 2005, Grill-Spector et al., 2006) found in adults but also in infants below 6 months of age (e.g., Dehaene-Lambertz et al., 2010; Lloyd-Fox et al., 2019). However, a classical repetition suppression occurs within several repetitions, while in our study infants seem to be able to transfer what they repeatedly heard in their native language in their family surroundings to a completely new experimental setting with pseudowords they have never heard. Thus, this indicates a much more sophisticated processing mechanism because they have to apply a learned rule to a new context. We think that the above interpretations are not mutually exclusive but complement each other because infants have to identify familiar rules to be able to transfer these to a new auditory context and differentiate these from new unfamiliar rules of an unknown language. Summarizing the fNIRS results, our findings show processes driven by attentional and memory-related mechanisms, thus novelty effects for non-native rules. The fact that a clear differentiation between native and non-native rules is present for both phonotactics and stress pattern underlines the impact of postnatal acoustic exposure. The more mature right-hemispheric lateralization for stress pattern rules might be interpreted in terms of an impact of prenatal exposure or a rapid learning during the first three months after birth. Because other studies identified a similar right-hemispheric lateralization already in newborns, we tend to assume a prominent rule of prenatal exposure. However, we cannot completely rule out an additional influence of rapid postnatal learning. Future studies could better clarify this issue by investigating phonotactic and stress pattern conditions in newborn infants. These results emphasize the importance of a multi-methodological approach, as both neuroscientific methods indexed similar but also complementary processing mechanisms. Both methods highlight that 3 months of postnatal experience with phonotactic rules of the native language suffice to reliably be differentiated from non-native rules. ERP and fNIRS effects indicate similar differentiation between native and non-native rules, most likely driven by attentional and memory-related processes. Lateralization effects indexed by fNIRS emphasize a more mature recruitment of brain regions for stress pattern, maybe because of the prolonged exposure to these linguistic feature already during the prenatal period. Summarizing, the present results are important as they show for the first time an impact of both prenatal and postnatal experience on neural mechanisms in 3-month-olds assessed with a naturalistic design in which (1) several items per condition are presented and (2) each condition is presented equally often. Such a design was only rarely adopted in infants of this young age. Usually, an MMR design was used, in which a short memory trace of one single stimulus is created and the brain’s reaction to one or two deviant stimuli is tested. The findings of the present study are crucial as they bear the potential to identify neural correlates of mature or immature processing mechanisms in pathological subject groups such as preterm infants or infants with a familial risk for language disorders very early during development, much earlier than standard developmental assessments can currently provide. CRediT authorship contribution statement Annika Rackl: Writing – review & editing, Formal analysis. Jenny Tippmann: Writing – review & editing, Investigation, Formal analysis. Sonja Rossi: Writing – review & editing, Writing – original draft, Visualization, Supervision, Project administration, Investigation, Formal analysis, Data curation, Conceptualization. Isabella Thurner: Writing – review & editing, Formal analysis. Hellmuth Obrig: Writing – review & editing, Supervision, Conceptualization. Funding sources This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Declaration of Competing Interest We declare no competing interests.