Effects of Early Infant Nutrition on Aperiodic Exponent From 2 to 12 Months Using EEG Analysis

ABSTRACT

Neural development begins in gestation and rapidly accelerates throughout early life. The environmental effect of infant diet is a subject of increasing study as it is the basic nutritive sources which support biological processes essential for healthy development. Previous research in our group has found small but significant effects of nutrition on early cognitive developmental tests between children primarily fed human milk (BF), soy‐based formula (SF), and dairy‐based formula (MF) for the first 12 months of life. This investigation uses this same data set: a study throughout infancy from 2 to 12 months of life. This study seeks to compare brain maturation using EEG aperiodic exponent between infants who received breast milk feeding and those who received a dairy‐based or soy formula. High‐density electroencephalographic (EEG) recordings were taken at 2, 3, 4, 5, 6, 9, and 12 months of age using a 5‐min silent video baseline. Using Specparam (formerly fitting oscillations and one‐over‐F or “FOOOF”), aperiodic activity was computed through a power spectral density (PSD) analysis for each session, which were then averaged per dietary group over left, right, and medial frontal and parietal regions of interest (ROIs) and one occipital ROI. Aperiodic exponent is a potential marker of neuromaturation, as it is hypothesized to relate to excitatory/inhibitory (E/I) balance and structural development in GABAergic systems. We used generalized estimating equations in order to evaluate differences in aperiodic exponent across dietary grouping as well as across age. Consistent with some previous findings, aperiodic exponent was found to decrease with age; however, no significant associations were found with infant diet except at 3 months of age in which BF infants had a larger aperiodic exponent than formula fed (FF) infants.

Short abstract

Our research addresses the gap in understanding how aperiodic activity develops and changes throughout the first year of life. There are currently conflicting reports of whether the slope of the brain's aperiodic activity (the exponent) increases or decreases throughout normative infant development, and our research lends support to the idea that the exponent decreases. In addition, this is the first study to examine early nutrition's effect on aperiodic activity in infancy.

1. Introduction

The brain begins developing in gestation and enters a critical period throughout infancy. During gestation, neuronal proliferation—the creation of new neurons—continues into postnatal development through at least 4.5 weeks postpartum (Couperus and Nelson 2006), and the most prominent period of myelination continues until 2 years of age (Tierney and Nelson 2009). Developmental processes after birth are subjected to a wider range of external factors than during prenatal periods, with infant diet as a vital factor (Lucas et al. 2001). Infant nutrition directly supports these developmental processes and presents a simple yet effective mechanism to interface with these features of neural development. As such, infant nutritive content is believed to play a critical role in the structural and functional development of the brain, and it has been demonstrated that different healthy infant diets (fed human milk or commercial infant formulas) may impact neuronal development both acutely (Berger et al. 2022; Deoni et al. 2013; Ottolini et al. 2020; Cusick and Georgieff 2016) and longitudinally (Deoni 2018; Ou et al. 2015; Victora et al. 2005; Gökçay 2010).

Electroencephalography (EEG) measures brain activity with temporal resolution on the millisecond‐scale. Using a non‐invasive sensor cap, this neuroimaging modality detects changes in minute electric potentials from excitatory postsynaptic potentials at the timescale in which cognition occurs. The multiple sensors on an EEG cap collect this information as regional summations of the electric potentials (voltage) produced by excitatory post‐synaptic potentials in the brain (Manning et al. 2009). The signal produced is in the form of a continuous wave, which can be separated into periodic and aperiodic components. The periodic, or synchronous, component has previously been the primary measure of interest, with the aperiodic component simply regarded as “noise” (Voytek et al. 2015). However, traditional methods of periodic spectral analysis fail to disaggregate the periodic and aperiodic components and often have not considered how aperiodic activity affects their findings (Donoghue et al. 2022). Recent technological advancements and subsequent investigations have made it possible to fully consider the dynamics and significance of this underlying signal. To examine this signal more fully, open‐source toolboxes such as Specparam (formerly “FOOOF”) (Donoghue et al. 2020) have been developed that enable the parameterization of both the periodic and aperiodic components of EEG. Specparam analyzes a signal's power spectrum which is the power of standardized frequency bands throughout a given EEG and then parses out the oscillatory and aperiodic components. Aperiodic activity presents broadly as “1/f‐like,” in which power generally decreases as frequency increases, following a power‐law function (P α 1/f ^β in which P is power, f is frequency, and β represents power‐law exponent) (He 2014). Once oscillatory components are removed, the power of the resultant signal shows a linear relationship with frequency. Therefore, the toolkit returns two representative aperiodic values: an offset and a slope. The aperiodic offset is the y‐intercept of the best‐fit line and the aperiodic exponent details this line's (negative) slope. Specparam has been empirically evaluated and the aperiodic exponent was found to have good test–retest reliability for eyes‐open and eyes‐closed resting states (McKeown et al. 2024).

Recent investigations have shown that aperiodic activity has a number of observed physiological correlates, such as age (Voytek et al. 2015; McSweeney et al. 2021; Schaworonkow and Voytek 2021; Hill et al. 2022; Wilkinson et al. 2024), drug response and receptivity (Muthukumaraswamy and Liley 2018; Liu et al. 2021), cognitive task (Voytek et al. 2015; Donoghue et al. 2020; Preston et al. 2022), and band‐specific shifts in excitatory‐inhibitory (E/I) balance in the brain (Donoghue et al. 2020; Gao, Peterson, and Voytek 2017; Chini et al. 2022). This shift in E/I balance is developmentally relevant, as it is reflective of the underlying neurophysiology known to be maturing at this time. Specifically, interneurons producing GABA (gamma‐aminobutyric acid) undergo a switch from excitatory to inhibitory signaling early in the postnatal period (Ben‐Ari 2014; Cherubini et al. 1991) which has been shown to manifest in changes in E/I balance (Ahmad et al. 2022). E/I balance has been investigated as a potential biomarker of the abnormal GABAergic systems that present in Autism Spectrum Disorder (Zhao et al. 2022; Arutiunian et al. 2024), circuit‐level models of Alzheimer's disease (Martínez‐Cañada et al. 2023), as well as in treatment responses of schizophrenia (Liu et al. 2021). Analyzing the dynamics of aperiodic activity in early developmental stages, through non‐invasive means like EEG neuroimaging, may invigorate models for further investigations in the underlying neuropathology of E/I balance. The dynamics of the aperiodic signal stand as a potential analytical addition to various neurodevelopmental investigations.

The American Academy of Pediatrics currently recommends exclusively breastfeeding newborns for the first 6 months after birth (Meek and Noble 2022), further suggesting formula if it is not possible to breastfeed. While human milk does provide early health benefits such as increased immunity to disease (Carr et al. 2021; Field 2005), lower rates of infections (Galton Bachrach et al. 2003), and lower likelihood of obesity and diabetes later in life (Arenz et al. 2004), the extent to which it provides a benefit to neurodevelopment is less well known. While a large literature exists demonstrating that human milk provides a slight cognitive advantage when assessed by neuropsychiatric exams (Prado and Dewey 2014; Walfisch et al. 2013; Eickmann et al. 2007; Kafouri et al. 2013), recent meta‐analyses have called into question the role of confounding variables (Walfisch et al. 2013) (maternal education and intelligence quotient (IQ) as well as familial socioeconomic status), and it remains unclear whether this advantage has direct or measurable neuronal correlates (Gilbreath et al. 2024). It is also important to note that when differences in mental or cognitive development between human milk or breastfed (BF) and formula fed (FF) infants are reported, often both groups are still within normal ranges (Andres et al. 2012). Although there are compositional and nutritional differences between human milk and commercial formula, there is an effort to include essential nutrients found in human milk thought to be important for cognition such as docosahexaenoic acid (DHA) (Decsi et al. 2023; Colombo et al. 2013) to formula (Pickett‐Bernard 2006). These additions to commercial formula may serve to narrow the reported gap in cognition reported between infants fed human milk and formula. Previous research in our group has used EEG to identify potential differences in the functional neuronal development between BF and formula fed (FF) infants. Overall, infants fed human milk displayed markers in their PSD that indicated a slight neuromaturational advantage over their formula‐fed counterparts at 2 months of age; however, this effect was subtle and did not persist past 3 months (Gilbreath et al. 2023). Taken together, the extent to which different healthy infant diets influence functional neuronal development is unknown.

This investigation seeks to supplement the limited research that has been done in infant neurodevelopment using EEG, and to further investigate how aperiodic exponent, as a potential marker of neuromaturation, differs between infants who were exclusively fed human milk (BF), soy‐based formula (SF), or dairy‐based formula (MF) in the first year of life. Based on previous literature which investigated aperiodic activity from infants (Schaworonkow and Voytek 2021; Rico‐Picó et al. 2023), early‐to‐mid adolescence (McSweeney et al. 2021; Hill et al. 2022), and between young and older adults (Voytek et al. 2015), we predict that aperiodic activity will decrease as a function of age and that BF infants will have a lower aperiodic exponent at developmentally relevant timepoints as compared to FF infants (Andres et al. 2012; Gilbreath et al. 2023; Jing et al. 2010; Pivik et al. 2019). Regional effects of aging and early diet on the aperiodic exponent will also be investigated, and we hypothesize that the exponents in the frontal regions will be lower and the posterior regions will be higher in accordance with recent findings (Rico‐Picó et al. 2023).

2. Methodology

Data were collected from a longitudinal cohort study investigating the effects of infant diet on cognitive growth and development (Andres, NCT00616395). Participants were recruited from 2002 to 2010 at ages 1 to 2 months, with mothers not ingesting estrogenic compounds (such as soy products) and alcohol during pregnancy and throughout their lactation period. Through pediatric consultation, parents chose the dietary group to which their infant would be assigned, and a 1:1:1 ratio was maintained between the three groups (BF:SF:MF). FF groups were instructed to maintain their chosen diet of soy‐based (SF) or dairy‐based (MF) formula, each supplemented with DHA and arachidonic acid, from two to 12 months. The BF group was encouraged to do the same, with a choice to switch to the FF group if breastfeeding was not possible between age 6 and 12 months. All groups were allowed to introduce complementary foods after age 4 months. Exclusion criteria included weight of less than five kilograms at 3 months of age; being given complementary foods prior to 4 months or switching formula after 2 months; and being administered medications which would alter typical growth and development. EEG sessions were scheduled at 2, 3, 4, 5, 6, 9, and 12 months. Prior to any study procedures, informed consent was gathered from parents. The study and its protocols were approved by the Institutional Review Board of the University of Arkansas for Medical Sciences.

A silent video‐baseline was played during each EEG session in a dim room for approximately 5 min during which a 128‐sensor cap collected EEG data (Magstim EGI, Net Amplifier 400) with a sampling rate of 250 Hz in an audiometric EEG chamber. This video was consistent across all subjects and ages and contained a shape that moved slowly across the screen while morphing into other shapes (triangles, squares, circles, etc.). Infants were seated in the caregiver's lap or independently depending on guardian preference and infant temperament. Infant attentiveness to the video was not specifically monitored; however, if the infant began moving to the degree it greatly affected data quality, EEG acquisition was paused until movements stilled. Impedances were checked at the beginning of each EEG session with the goal of each electrode being < 100 kΩ.

The EEG data were preprocessed using the Harvard Automated Processing Pipeline for Electroencephalography version 3.3 (Gabard‐Durnam et al. 2018; Delorme and Makeig 2004). Data were band‐pass filtered from 0.5 to 45 Hz, re‐referenced using REST (Yao 2001), then segmented into 10 s epochs from which biological artifacts were removed using Morlet wavelets. Individual channels and epochs still containing artifacts were then removed. EEGs containing more than 30% rejected channels or segments were excluded from further analysis and exact numbers for excluded subjects are in Table S1. The final number of subjects for each age group is as follows: 2 months (BF = 108, MF = 100, SF = 106), 3 months (BF = 114, MF = 116, SF = 118), 4 months (BF = 111, MF = 110, SF = 121), 5 months (BF = 137, MF = 136, SF = 144), 6 months (BF = 136, MF = 143, SF = 142), 9 months (BF = 122, MF = 114, SF = 120), and 12 months (BF = 135, MF = 119, SF = 133).

Data were loaded into Brainstorm to perform a power spectral density analysis using Welch's method (window length: 1.000 s, window overlap ratio: 50.0%), and a subsequent aperiodic analysis. Specparam was used to extract the aperiodic exponent from all participants across age groups using the following parameters: [Specparam implementation: MATLAB; frequency range for analysis: 1.0–40.0 Hz; peak model: Gaussian; Peak width limits: 0.5–12.0 Hz; Maximum number of peaks: 3; Minimum peak height: 1.0 dB; Proximity threshold: 2.0; Aperiodic mode: fixed; guess weight: none]. Data were then grouped into seven regions of interest (ROIs), following groupings from previous literature using this data set (Alatorre‐Cruz et al. 2021) and that closely mirror regions selected in another infant aperiodic study (Rico‐Picó et al. 2023). The aperiodic exponent was averaged into the following ROIs: Left Frontal (channels 19, 20, 23, F3, 27, 28, F7), Medial Frontal (5, 6, 10, Fz, 12, 16, 18), Right Frontal (3, 4, 117, 118, F4, F8, 123), Left Parietal (47, 51, P3, 53, 59, 60), Medial Parietal (61, Pz, 67, 72, 77, 78), Right Parietal (85, 86, 91, P4, 97, 98), and Occipital (O1, 71, 74, Oz, 76, 82, O2). These are visualized in Figure 1.

FIGURE 1.

Regions of interest (ROIs) overlayed on 128‐channel EEG sensor map.

Generalized estimating equation (GEE) models were performed to determine the effect of age, sex, and dietary group on the aperiodic exponent using the statistical software IBM SPSS Statistics (Version 29), and differences between groups were evaluated using post hoc t‐testing. Because maternal IQ (Walfisch et al. 2013; Gomez‐Sachiz et al. 2004; Lean et al. 2018) and gestational age—even when within normal term ranges (Husby et al. 2023; Yang et al. 2010)—are both linked to infant's later cognition, they were included in our GEE model as potential confounds. Maternal IQ scores were calculated using the Wechsler Abbreviated Scale of Intelligence (WASI) (Wechsler 1999) using an aggregate score across subtests at the infant 6 month visit and gestational age was determined through parental reports. The average maternal WASI scores and gestational age of the infants for each group and timepoint are presented in Table S2. GEE models are increasingly used in longitudinal datasets as they take into account repeat measures (multiple time points per subject) inherent in these datasets while also allowing for multivariate analysis (Hanley et al. 2003; Wang 2014). In addition, paired t‐tests were used to evaluate hemispheric differences between the FL and FR ROIs as well as to examine a potential anterior‐to‐posterior gradient using the FC, PZ, and OZ ROIs. All p‐values were Bonferroni corrected and significance was set at p < 0.05. To test whether maternal WASI and infant gestational age had a significant effect on aperiodic exponent, a separate GEE was performed without the variable of dietary group and results can be found in Table S3.

To further investigate regional differences in aperiodic exponent, the source space reconstruction algorithm sLORETA (Pascual‐Marqui 2002) was employed using age‐appropriate infant brain atlases inherent in brainstorm (Li et al. 2015). Aperiodic data were averaged across all participants at 2 and 6 months before being projected into source space to further explore localized differences in aperiodic activity.

3. Results

3.1. Age‐Related Differences

GEE testing found age to have a significant effect on the aperiodic exponent across all ROIs (p < 0.001). At ages 2, 3 and 4 months old, the aperiodic exponent was higher than all subsequent age groups (p < 0.001). This relationship is depicted in Figure 2 and the p values between age groups are presented in Table 1. Topographical representations of the aperiodic exponent are visualized in Figure 2B and specific analysis of regional variations is further explored using our ROIs.

FIGURE 2.

Average aperiodic exponent at each region of interest for all subjects at each time point (A). Average topological plot of the aperiodic exponent for all subjects at each time point (B).

TABLE 1.

GEE results of the aperiodic exponent for each region of interest.

Region of interest	p value for ages	Wald Chi	Post hoc t test for ages	p values for sex differences	Direction	p values for dietary group differences
FL	< 0.001	535.69	2 > All, 3 > All, 4 > 6, 9, 12	0.882		0.404
FR	< 0.001	701.633	2 > All, 3 > All, 4 > All	0.848		0.169
FC	< 0.001	1132.57	2 > All, 3 > All, 4 > All, 5 > 9 and 12	0.027	M > F	0.172
PL	< 0.001	1252.36	2 > All, 3 > All, 4 > All, 5 > 6	0.318		0.42
PR	< 0.001	1006.601	2 > All, 3 > All, 4 > All, 6 < 12 (0.011)	0.023	M < F	0.121
PZ	< 0.001	948.069	2 > All, 3 > All, 4 > All, 5 > 6 (0.036), 9 (0.002), 12 (< 0.001)	0.187		0.647
OZ	< 0.001	678.383	2 > All, 3 > All, 4 > All	0.072		0.049

Note: All p values for post hoc t‐testing are < 0.001 unless otherwise indicated. A shorthand notation of “all” is used to indicate that a particular age group had a significantly larger exponent than all subsequent age groups.

Differences between the FL and FR regions were evaluated to investigate potential hemispheric asymmetries. Starting at 4 months and extending until 12, the FL region demonstrated a significantly higher aperiodic exponent (Figure 3). Anterior/Posterior differences were also evaluated, showing PZ with a higher exponent than OZ at every age (p < 0.001), and a higher exponent than FC at 3, 4, 5, 6, 9, and 12months. FC was higher than OZ at 2 months and between the ages of 4 and 12 months. These results are visualized in Figure 4.

FIGURE 3.

Hemispheric differences between the frontal left (FL) and frontal right (FR) regions for each age group across all subjects. **Indicates p < 0.01.

FIGURE 4.

Regional differences between the frontal central (FC), parietal center (PZ), and occipital center (OZ) for each group across all subjects. **Indicates p < 0.01.

To further investigate these regional differences, source space reconstructions were visualized using age‐appropriate infant atlases at 2 and 6 months. These reconstructions showed a higher distribution of aperiodic activity at 2 months and more localized aperiodic activity at 6 months (Figure 5).

FIGURE 5.

Source space reconstruction of aperiodic exponent data from all 128 channels. (A) Right sagittal view of averaged data from the two‐month‐old age group. (B) Superior view of the two‐month‐old data. (C) Right sagittal view of averaged data from the six‐month‐old age group. (D) Superior view of the six‐month‐old data.

3.2. Diet and Sex‐Related Differences

Differences between dietary groups were evaluated and showed that BF infants had a lower aperiodic exponent on average than SF infants (p = 0.049) at OZ (Figure 6). However, post hoc t‐testing showed that no significant differences were maintained between dietary groups at any specific age. This suggests that age may be driving the effect seen at OZ, indicating that this finding should be interpreted cautiously. Differences in aperiodic exponent resulting from sex were examined and localized effects were found: males had a greater aperiodic exponent at the FC (p = 0.027) while females had a greater exponent in the PR (p = 0.023).

FIGURE 6.

Group differences in the average aperiodic exponent for each time point for the occipital ROI (OZ).

4. Discussion

Relatively few studies have investigated aperiodic activity during infancy (Schaworonkow and Voytek 2021; Wilkinson et al. 2024; Chini et al. 2022; Rico‐Picó et al. 2023; Favaro et al. 2023), and among these studies, there are several discrepancies. The current study seeks to investigate aperiodic activity for the full first year of life (2 to 12 months) and is the only study to examine the effect of infant diet. In addition to expanding the observed developmental timeframe, this study is one of the largest infant EEG data sets and is one of the first to use source space reconstruction to visualize spatial differences in aperiodic activity.

4.1. Age‐Related Differences

Our research supports the growing body of literature demonstrating that aperiodic exponent decreases as a function of age (Voytek et al. 2015; McSweeney et al. 2021; Schaworonkow and Voytek 2021; Hill et al. 2022; Rico‐Picó et al. 2023; Favaro et al. 2023). This relationship has been observed in low‐frequency analysis of electrocorticography (ECoG) and EEG data, where it was lower in older adults (60–70 years old) compared to younger adults (20–30 years old) when participants performed a visual working memory task (Voytek et al. 2015). Studies in childhood and adolescent populations have also demonstrated these trends, with exponent decreasing from 4 to 12 years of age (Hill et al. 2022), and between 15 and 15 years of age (McSweeney et al. 2021). In infancy, there are relatively few papers mapping out aperiodic exponent across time. In the first paper to do so, Schaworonkow et al., researchers found that exponent decreased with age between a range of 38–203 days (~1.2–6.5 months) (Schaworonkow and Voytek 2021), with another group reporting that exponent decreased across the time‐points 6, 9, and 16 months (Rico‐Picó et al. 2023). Our findings mirrored this pattern, showing a decrease with early development, particularly from 2 months until 5 months of age in a healthy infant population. This progression is also seen spatially in the source space reconstructions. At 2 months of age, there was a greater global dispersion of a high aperiodic exponent (Figure 5A,B), and more localized areas of generally lower aperiodic exponent at 6 months (Figure 5C,D). As a decrease in the aperiodic exponent corresponds to a flattening of the PSD, this suggests an increasing presence of excitatory activity (relative to inhibitory activity) in the first 5 months of life. This shift could reflect the development of early GABAergic systems in the brain (Gao, Lin, et al. 2017).

Not all studies examining aperiodic activity during infancy support these results; however, two studies report the opposite relationship between aperiodic exponent and age –with the exponent increasing with increasing age (Wilkinson et al. 2024; Chini et al. 2022). In Wilkinson et al., researchers analyzed the aperiodic exponent in infants from 2 to 44 months in several large cohorts and found an increase in exponent as the infant aged (Wilkinson et al. 2024). There are two important methodological differences between Wilkinson's study and the current study: (1) because the original Specparam software did not produce sufficiently good fits (defined in the paper as an increased mean‐squared error for the 2–7 month age range) on their data, researchers modified the code to make it more applicable to their infant data set; (2) there were differences in filtering methods and frequency ranges used in each study. Importantly, Wilkinson's alterations of Specparam did produce a more accurate fit to their data by elevating the initial baseline estimate of the flattened power spectra, suggesting a more careful consideration of applying Specparam to infants in its native state. In addition, recent literature indicates the aperiodic exponent is heavily dependent on both the total frequency spectrum of the data and the frequency range of interest (Gerster et al. 2022). Gerster et al. found that a ‘spectral plateau’ which occurs in most datasets at high frequencies around 100 Hz significantly affects the aperiodic exponent, and this effect is more substantial as the upper bound of the frequency range of interest increases (i.e., there is a greater effect from 1 to 50 Hz than from 1 to 10 Hz). Wilkinson's study uses a wider band pass filter (1–100 Hz) and computed the aperiodic activity in a different frequency range (2.5–50 Hz) than used in this study. Given the similarities in our data sets, the diverging findings may result from these methodological differences and should be interpreted carefully. These conflicting results also illuminate the need for a more definitive parameterization for computing aperiodic activity. The current study does not test any other version of Specparam as the original version returned acceptable fits, defined as an R ² value greater than 0.90 and a low mean squared error (MSE) (Table S4 and Figure S1) (Ostlund et al. 2022). In another study examining aperiodic activity in infants, Chini et al. also found an increasing exponent with age in infants between 35 and 46 weeks (Chini et al. 2022). Infants in this study were asleep during EEG recordings, and previous research indicates that sleep state during later infancy and through adolescence significantly alters aperiodic exponent (Favaro et al. 2023). Therefore, differences found in this study may be due to using asleep vs. awake infants. The present study follows the general parameters of the original paper that reported aperiodic exponent in infants using Specparam (Schaworonkow and Voytek 2021) as well as the recommendations set in a recent methodological paper exploring how best to parameterize the power spectrum in neurodevelopment (Ostlund et al. 2022). Future studies should compare aperiodic exponent values from Voytek and Wilkinson's versions of Specparam to investigate potential differences due to diverging methodologies.

4.2. Region Specific Aperiodic Exponent Values

Across all age groups except for 2 months, the central‐parietal region displayed the highest aperiodic exponent when compared to frontal and occipital regions. At 2 months of age, the central‐parietal and frontal regions were not significantly different from each other, but both were significantly higher than the occipital region. In addition, at both 2 and 3 months of age, aperiodic exponent values in frontal regions were similar to the occipital regions. At all other time points, the frontal regions were significantly higher. These effects demonstrate that aperiodic activity is more distributed at these younger time points and becomes more localized to the parietal regions with increasing age. A recent study focusing on infants at 6, 9, and 16 months of age found similar results; infants consistently had larger intercepts in posterior regions (that include an analogous parietal region) across ages when compared to frontal regions (Rico‐Picó et al. 2023).

Similar to patterns seen in the anterior–posterior regions across ages, hemispheric differences between frontal left and right regions emerged at 4 months. Left regions exhibited a higher exponent from 4 to 12 months, with no differences found at 2 and 3 months. Frontal regions were specifically identified and chosen for investigation due to the extended literature showing frontal asymmetry in traditional periodic analyses (Fox 1991; Peltola et al. 2014; Smith et al. 2017; Saby and Marshall 2012). A 2022 meta‐analysis of the infant functional asymmetry literature found that during resting state, infants tend to have a right hemispheric dominance (Bisiacchi and Cainelli 2022). This supports a long‐standing idea that the right side of the brain tends to develop faster across infancy (Draper 1988). If lower aperiodic exponent is indicative of greater maturation as our age‐related data suggests, these reduced values in the right hemisphere in relation to the left directly support this hypothesis. Furthermore, at 2 and 3 months, the lack of frontal exponent asymmetry as well as the reduced differences between anterior/posterior regions is consistent with literature suggesting that in early infancy, neuronal activity is more distributed (Haartsen et al. 2016). As infants age, networks become increasingly localized (Fan et al. 2011), indicating a higher degree of network segregation. Source localization was utilized in this study to verify these trends in our scalp‐level data. At 2 months of age, aperiodic exponent values in our source model are clearly distributed across the cortex and at 6 months there is a greater degree of localization in the parietal region. This supports our scalp‐level analysis as well as evidence from graph theory that supports the idea that the brain segregates into more discrete networks that correlate to increased age during infancy (Zhao et al. 2019).

4.3. Diet‐Related Differences

Differences in aperiodic activity due to diet were only evident in our central occipital region. BF infants were found to have a lower exponent than both MF and SF infants on average across all time‐points. However, when evaluating these differences by month (e.g., 2‐month‐old BF infants vs. 2‐month‐old MF infants), significance was not maintained. Because age, dietary group, and sex were all put into the GEE model as independent variables, there is a chance that the overwhelming effect of age could be partially driving the significance found at OZ. Our results indicate that if dietary group has an effect on aperiodic activity, it is both regional and small. These results echo our previous findings, with BF infants having only slightly higher measures of neuromaturation at specific time‐points during infancy in both EEG (Gilbreath et al. 2023; Pivik et al. 2019) and on the Bayley scale of infant development (Andres et al. 2012). Changes in aperiodic exponent were reasonably expected to result from distinct early infant diets due to the reported effects of human milk on neuromaturation (Cusick and Georgieff 2016; Deoni 2018; Belfort et al. 2016; Georgieff et al. 2018). Specifically, a BF diet is associated with increases in gamma activity (Gilbreath et al. 2023; Pivik et al. 2019) that is inherently mediated by interneurons containing γ‐aminobutyric acid (GABA) (Buzśaki and Wang 2012). GABA‐ergic activity in part drives the E/I balance that the aperiodic exponent is hypothesized to measure. A lack of significance in our findings could indicate that GABA‐ergic development is minimally affected by early infant diet, or that aperiodic exponent is not a sensitive measure of E/I balance in cases where differences are subtle. In the seminal work relating E/I balance to aperiodic exponent, Gao et al. acknowledge that slope‐inferred E/I ratio is a simplification of the underlying cortical circuits involved (Gao, Peterson, and Voytek 2017); the multitude of dynamic changes (synaptogenesis, migration of GABA‐ergic interneurons, increases in myelination) that are known to occur during infancy may be difficult to fully encapsulate through this particular method. Furthermore, a recent study using pharmacological manipulations in mice to modulate known contributors to the E/I ratio (GABA_a receptors) found that aperiodic exponent was not a reliable marker of cortical excitation/inhibition (Salvatore et al. 2024). While aperiodic exponent could serve as a proxy of the E/I ratio, conflicting evidence exists, and the extent to which exponent is indicative of the known GABA‐ergic processes that occur either in infancy or as a result of early diet should not be overstated. The infant diet literature is relatively sparse in relating how normative, healthy diets affect functional neurodevelopmental outcomes measured through neuroimaging (Gilbreath et al. 2024), and this present work supports the growing idea that human milk's effect on functional neurodevelopment may be more nuanced than once thought (Walfisch et al. 2013). A limitation of this study is that complementary foods were allowed to be introduced to all infants after 4 months of age, and this additional nutritive source could potentially dilute the effects of the initial choice of diet (BF vs. FF). However, this is likely not a major concern as aperiodic exponent did not significantly differ for more than one region before the introduction of complimentary foods.

5. Conclusion

Based on resting‐state EEG data, our analysis shows changes in passive aperiodic activity in the infant brain across the first year of life. Because of the study's exclusion criteria and robust population size, our findings provide a potential baseline of healthy development of aperiodic activity over the first year of life. Aperiodic activity showed an increasing flattening of the power spectral density globally over the first 5 months, after which it plateaus. This pattern was consistent in the majority of our ROIs, indicating that while having a slight regional effect, the decrease in aperiodic exponent across infancy is global. Previous research has found structural differences between infants fed human milk and infants fed formula in the first year of life (Berger et al. 2022; Deoni et al. 2013; Deoni et al. 2018; Schneider et al. 2023; Sullivan et al. 2023; Zhang et al. 2022). However, our analysis of aperiodic activity showed a very limited difference between infants fed human milk, dairy‐based formula, or soy‐based formula. This indicates that infant diet may not ultimately have a significant effect on markers of E/I balance; moreover, this study lends itself to a growing body of literature demonstrating that different healthy diets during infancy may not have a substantial effect on functional neurodevelopment.

Author Contributions

Dylan Gilbreath: investigation, methodology, writing – original draft, visualization, writing – review and editing, software, formal analysis, data curation, conceptualization. Adam Andrews: writing – original draft, investigation, writing – review and editing. Darcy Hagood: writing – review and editing. Aline Andres: writing – review and editing, funding acquisition, project administration. Linda J. Larson‐Prior: conceptualization, investigation, funding acquisition, writing – review and editing, project administration.

Funding

This work was supported by the National Institutes of Health (P20 GM103429) and U.S. Department of Agriculture (6026‐10700‐007‐00D).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Global R ² values by age in months for the fit of the aperiodic exponent.

Table S1: Number of total and excluded subjects across timepoints. Some subjects were withdrawn from the study after the EEG visit occurred; these subjects were excluded from further analysis. If EEG data failed either the pipeline QC (< 0.2 post/pre waveletting values for any of the frequencies used in the study [1–45 Hz]) or the data QC (fewer than 70% good channels or segments), data were excluded from subsequent analysis.

Table S2: Average maternal WASI scores and average gestational age of the infants at each timepoint and dietary group. The numbers of subjects per dietary group by age are also presented.

Table S3: Generalized estimating equation (GEE) results testing age‐specific associations between maternal WASI, gestational age, and aperiodic exponent values across cortical regions. For each region, GEE models included Maternal WASI × Age and Gestational Age × Age interaction terms, with Sex and Group entered as covariates and subject ID modeled as the clustering variable (exchangeable correlation structure, Gaussian family). The table reports the unstandardized coefficients (B), standard errors (SE), and 95% confidence intervals (CI) for all Maternal WASI–Age and Gestational Age–Age interaction terms. Uncorrected p‐values and Bonferroni‐corrected p‐values (corrected across all regions and all interaction tests) are provided.

Table S4: R ² values for the fit of the aperiodic exponent across all ages and regions of interest. Global R ² and Mean squared error (MSE) are also included. ROIs match Figure 1 and are as follows: Frontal Left (FL), Frontal Right (FR), Frontal Central (FC), Parietal Left (PL), Parietal Right (PR), Parietal Central (PZ), and Occipital Central (OZ).

Gilbreath, D. , Andrews A., Hagood D., Andres A., and Larson‐Prior L. J.. 2026. “Effects of Early Infant Nutrition on Aperiodic Exponent From 2 to 12 Months Using EEG Analysis.” Psychophysiology 63, no. 2: e70247. 10.1111/psyp.70247.

Data Availability Statement

Data will be made available on publication.