The aperiodic and periodic activities of EEG dynamically relate with cognitive performance in schizophrenia

Abstract

Background

Aperiodic and periodic activities of electrophysiological signals have strong correlation with various neurocognitive factors. In the current study, we aim to investigate the aperiodic exponent and periodic oscillations (alpha and beta band power) and their associations with cognitive performance in schizophrenia (SCZ).

Methods

We enrolled 32 SCZ patients and 33 healthy controls (HC) for the study. Cognitive performance was assessed using the total Brief Assessment of Cognition in Schizophrenia (BACS). Before and after the language comprehension tasks (humor, metaphor and irony), the 5-min eyes-closed and eyes-open EEG signals were collected respectively. The aperiodic exponent and periodic power were extracted and obtained according to the division of brain regions. Finally, Pearson correlation was used to examine the relationships between the EEG parameters and behavioral measures.

Results

SCZ participants exhibited higher aperiodic exponents and lower periodic oscillations compared to HC. The aperiodic exponent decreased significantly after the tasks in central location (F (56,1) = 8.93, P = 0.004, η² = 0.14) in both groups, while the periodic oscillations had no significant change. The variance of the aperiodic exponent showed significantly negative correlation with Z scores of humor comprehension tasks (r = −0.42, P = 0.027) in SCZ. Besides, the pre-tasks aperiodic exponent in posterior location positively correlated with T scores of attention and speed of information processing (r = 0.35, P = 0.048).

Conclusions

Our findings confirmed the higher aperiodic exponent with lower alpha and beta band in SCZ, along with the variability and task-responsiveness of the aperiodic exponent. The findings suggest that aperiodic exponent holds strong cognitive functional implications in SCZ.

1. Introduction

Schizophrenia (SCZ) is characterized by disruptions to multiple behavioral and cognitive domains. Increasingly, cognitive impairment has been recognized as part of core symptoms. Almost all SCZ patients (98.1 %) exhibit cognitive impairment below expectations to different degree (Keefe et al., 2005), encompassing both low-level cognition (neurocognitive impairment) (Gebreegziabhere et al., 2022) and high-level cognition (social cognition and metacognition) (Frith, 2023). These distinct cognitive domains (e.g., neurocognitive domains like working memory versus social cognition) appear to be modulated by related but distinct neural mechanisms (e.g., involving differential engagement of prefrontal cortical regions versus limbic and temporoparietal structures) and are differentially impaired in individuals with SCZ. The impact of these cognitive impairments on functional disability and related outcomes is profound, affecting employment, independence, relapse prevention, medical comorbidity and costs (Keefe and Harvey, 2012). However, the biology underlying these cognitive impairments remains elusive.

The onset and severity of cognitive impairment have been linked to electroencephalogram (EEG) activity (Uhlhaas and Singer, 2010). Power spectral density (PSD) of EEG is made up of two components: a rhythmic periodic component, and an overlapping, irregular aperiodic component. While rhythmic periodic component is concentrated at specific frequencies, which is visible as peaks on the power spectrum, the aperiodic component manifests as the background spectral trend that decays with apparent 1/f^β behavior. Traditional analyses, which ignored or treated aperiodic activities as noise, have for decades pointed to abnormal neural oscillations in the theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (>30 Hz) bands in SCZ patients compared to controls (Newson and Thiagarajan, 2018). Hirano and Uhlhaas summarized current findings and perspectives on aberrant neural oscillations in SCZ in a review (Hirano and Uhlhaas, 2021), including reductions in the amplitude and connectivity of resting-state alpha-band power, and increase in delta and theta power, aligned with alterations in core circuit motifs that have been implicated in the pathophysiology of schizophrenia. However, the robustness of these oscillatory pattern changes in the experimental literature was inconsistent (Morrow et al., 2023). The possible causes of the variance may lie in the heterogeneity of alpha oscillations with respect to their generating sources and the mixture of alpha power with broadband background activity. Oscillatory and aperiodic activities are likely to be generated by distinct neural mechanisms and play different functional roles, which strongly calls for the necessity to disentangle them. However, the aperiodic component was historically overlooked due to a lack of dedicated analytical frameworks. For example, a relatively popular method (EEGLAB toolbox (Delorme and Makeig, 2004)) used the newtimef() function to ignore the aperiodic component, via to convert the data to the decibel (dB) scale through baseline division and log transformation. Gyurkovics et al. (2021) have found that this procedure may lead to a misrepresentation and misinterpretation of physiological mechanisms and result in misinterpreting aperiodic activity as periodic activity. Therefore, there is an increasing consensus on the imperative careful parametrization of spectral features to minimize its associated bias.

To distinguish different sources of activities within EEG, the PSD has been parameterized into periodic and aperiodic components according to the “power-law” distribution. That is, the PSD typically follows a straight-line pattern when plotted in coordinates of log power vs. log frequency: log(P)∝ − β log(f) or P∝f-β (0 < β < 4) (Buzsáki et al., 2012). The slope of this line is known as “1/f slope”, i.e. aperiodic exponent, and the offset reflects the broadband shift in power across frequencies. Retrospectively, there have been several methods for deriving the “1/f slope” in the human EEG literature: power law exponents (He, 2014), the Better OSCillation Detector (BOSC) (Whitten et al., 2011), the Irregular Resampling Auto-Spectral Analysis (IRASA) (Wen and Liu, 2016) and the fitting of one-over-f (FOOOF) (Donoghue et al., 2020). This study employed FOOOF, a state-of-the-art method that enables separate quantification of both parameters and has attracted significant research attention in recent years (Stanyard et al., 2024). In principle, the offset can vary independently of the slope (Donoghue et al., 2020), though, in practice, the two measures often correlate (Euler et al., 2024). Compared to limited studies investigating the offset (Zhang et al., 2023), emerging work suggests that the spectral exponent reflects the balance between excitatory and inhibitory synaptic activity - often framed in terms of underlying synaptic time-constants - and may therefore carry functional significance. Under certain theoretical models, steeper exponents (more negative slopes) have been interpreted as reflecting GABAergic inhibition, whereas flatter exponents (fewer negative slopes) are proposed to indicate a relative glutamatergic excitatory bias; this interpretation rests on the assumption that low-frequency power is primarily driven by inhibitory currents (Gao et al., 2017; Brake et al., 2024). Moreover, several reports propose that the aperiodic exponent indexes tonic E/I balance and relates to physiological and cognitive outcomes in healthy individuals - processing speed (Ouyang et al., 2020), cognitive control (Clements et al., 2021), attention (Tagliazucchi et al., 2013), working memory (Kardan et al., 2020), motor performance (Immink et al., 2021) and academic learning (Cross et al., 2022) - although the precise relationships remain to be confirmed (Østergaard et al., 2024), indicating its critical role in information transmission and storage, dynamic range, metastable states, and computational power (Zimmern, 2020). There is a growing consensus that the aperiodic exponent is important for cognitive functioning and interregional neuronal communication.

In the context of SCZ several studies have only investigated the aperiodic exponent of resting state EEG, reporting mixed results: some observed a higher spectral slope in SCZ, which could be normalized with acute memantine administration (Molina et al., 2020), while others observed the opposite (Racz et al., 2021). The lower aperiodic exponent was reported in adults with paranoid SCZ, similar with trait anxiety (Radulescu et al., 2012). On the other hand, clinically stable patients with chronic psychotic disorders exhibited“steeper” aperiodic slopes, with greater power in theta and alpha frequency bands and less power in gamma frequencies, possibly contributed by the E/I balance disturbances. There is also evidence that the aperiodic exponent of first episode schizophrenia spectrum psychosis patients has no significant differences with healthy control groups (Earl et al., 2024). These heterogeneous results may arise from the variable course of SCZ, the heterogeneity of impairments across different dimensions, inconsistencies in analytical methods employed (IRASA or FOOOF), and diverse approaches to handling regions of interest (averaged across all EEG electrodes or clusters). To date, few studies have investigated both periodic and aperiodic activities and their correlation with cognitive performance in SCZ. Modeling studies suggest that aperiodic activity and oscillatory coupling are often mechanistically interdependent (Wang, 2010), and even dynamically related during complex higher-order cognitive tasks, including language processing (Cross et al., 2022) and comprehension (Bornkessel-Schlesewsky et al., 2022). Our study therefore focused on task-modulated dynamics during figurative language comprehension - specifically metaphor, humor, and irony processing. These tasks engage hierarchical cognitive operations at the semantic-pragmatic level, requiring initial detection of semantic incongruity and contextual reanalysis for conflict resolution, skills shown to be impaired in SCZ (Adamczyk et al., 2024). Therefore, we hypothesized that: 1) the aperiodic exponent and periodic oscillation (alpha band power, AP, and beta band power, BP) of PSD would be disrupted in patients with SCZ; and 2) both the aperiodic exponent and periodic oscillation would fluctuate with cognitive tasks, with these variations correlating with performance. Therefore, the current study aims to explore the differences on aperiodic and periodic parameters of EEG and their relationship between cognitive performance in individuals with and without SCZ.

2. Method

2.1. Participants

Using G*Power software (Faul et al., 2007), A priori power analysis based on expected large effects (α = 0.05, 1-β = 0.80) indicated a minimum requirement of 38 participants (19 per group). 33 18–55-year-old participants with stable schizophrenia (SCZ, 13 males, age = 30.09 ± 10.46) and 33 healthy controls (HC, 16 males, age = 29.76 ± 10.71) participated in this study. EEG data were missing for one participant of SCZ group (excluded for low quality of EEG, i.e. number of bad channels ≥10 %). The patients' group were diagnosed with schizophrenia according to the International Classification of Diseases (ICD)-10. Other inclusion criteria were being on stable antipsychotics at least 6 months (i.e. psychopharmacological treatment had not been changed for the recent 3 months, and total score of the Positive and Negative Syndrome Scale (PANSS) ≤ 60).The criteria of exclusion were as follows: 1) experiencing pregnancy or lactation; 2) diagnosis of any other major disorder; 3) comorbid physical, infectious, immune system illnesses, or mental retardation; 4) drug or alcohol dependence; 5) a history of neurological illness; 6) mental retardation; 7) modified electroconvulsive therapy (MECT) in recent 6 months; 8) taking benzodiazepine or anti-cholinergic drugs. Especially, schooling years, sex and age were matched between two groups.

Ethical approval was provided by the Ethics Committee of Peking University Sixth Hospital (2022-46 and 2023-46), and all individuals provided written informed consent before participating

2.2. Clinical and cognitive assessment

All subjects underwent baseline screening with demographic and clinical questionnaires, as well as cognitive assessments. The diagnosis of SCZ was confirmed through the MINI International Neuropsychiatric Interview and International Classification of Diseases (ICD-10), and symptom severity was assessed with Positive and Negative Symptom Scale (PANSS). Neurocognitive functioning was assessed using the Brief Assessment of Cognition in Schizophrenia (BACS), and switched to T score according to the data of a large-scale mandarin-speaking population (Wang et al., 2016).

2.3. Figurative language comprehension task

The experimental materials for the figurative language comprehension task were based on the previous experimental materials to assess comprehension of humor (Adamczyk et al., 2019), metaphor (Adamczyk et al., 2021) and irony (Del Goleto et al., 2016). These three domains of figurative language deficit manifests in SCZ (Adachi et al., 2004). Due to the absence of a well-established and universally recognized experimental paradigm, we have compiled a summary of previous materials related to the understanding of metaphorical language tasks. Considering the cultural influence on language comprehension, the task was reorganized referenced from Metaphor and Sarcasm Scenario Test (MSST) (Adachi et al., 2004) and the Chinese humor comprehension materials (Ling, 2013). The task consists of 40 humor items, 15 metaphor items and 15 irony items, employing a multiple-choice style, i.e. one was correct for one point with four incorrect answers. For humor, three possible choices are provided, including: a correct funny ending; an incorrect straightforward non-funny ending; an incorrect unrelated non-sequitur ending. For metaphor and irony, the metaphoric scenarios are odd numbered and the sarcastic scenarios even with four incorrect answers. One of the incorrect answers in each sarcastic scenario was a ‘landmine answer’.

Participants were requested to read the stories and choose the figurative one, following the instructions given to each subject as follows (English translation): “In this part, you will be asked to read the texts and choose one of the choices that you think best suits the contexts. The entire test is divided into two parts - tasks. I. HUMOR - In the first part, your task is to indicate the endings that make the story humorous, a joke, or you simply consider them to be the funniest ending. II. METAPHOR AND IRONY - In the second part, your task is to indicate endings that, in your opinion, contain metaphor or irony. These are ambiguous statements or criticisms hidden in a seemingly approving statement.” The three tasks were administered in fixed-order blocks. The answer time and correct score were recorded. The score was switched to Z scores according to HC for following analysis.

2.4. EEG data acquisition and pre-processing

EEG data were continuously recorded from 64 channels using a German Brain Products system at a sampling rate of 5000-Hz, with online referencing to the Cz electrode, which were down-sampled offline to 500-Hz. Prior to the commencement of recording, electrode impedances were checked and maintained below 20 KOhms. The experiment collected continuous 10-min EEG data (5 min of eyes-open and 5 min of eyes-closed conditions) both before and after the verbal intervention task (10 min pre-test and 10 min post-test), with subsequent data processing involving segmentation into epochs (Fig. 1). During the whole recording session, subjects were instructed to stay awake and refrain from deliberate thinking.

Fig. 1.

Schematic illustration of the measurement procedure. The resting-state EEG was measured for 10 min each with 5-min eyes closed (EC) and 5-min eyes open (EO) before (pre) and after (post) an approximately 10-min figurative language comprehensive tasks. In the subsequent processing, the two consecutive 10-min EEG data segments (pre and post) will be respectively divided into two 5-min epochs (EC and EO) according to time markers.

Matlab (MathWorks, Natick, MA) with the EEGLAB toolbox (Delorme and Makeig, 2004) was used to process all data. All data were visually inspected and bad channels (Bigdely-Shamlo et al., 2015) were interpolated using EEGLAB's spherical interpolation function. Offline, data were re-referenced to the average of all electrodes, and the bandpass filtered between 0.5 and 80 Hz (fourth-order Butterworth filter) on the basis of previous literature (Hill et al., 2022), with a notch filter between 49 and 51 Hz to remove powerline noise. After artifact removal and before spectral computation, the ten-minute continuous data were segmented into two 5-min epochs for eyes-open and eyes-closed conditions. The data were further processed using multiple Wiener filtering, followed by wavelet-enhanced independent component analysis (ICA) to remove artifacts of extraneural origin (i.e., eye movements, muscle contractions, and cardiac activity). Finally, all pre-processed data files were visually inspected again before being included in the subsequent analyses and segmented into the eyes-open and eyes-closed conditions.

2.5. Aperiodic and periodic parameters of the spectral data

Power spectral density (PSD) helped to decompose EEG signals into their frequency-domain components for each participant and electrode using Welch's method in Matlab (2 s Hamming window, 50 % overlap) (Donoghue et al., 2020). After this, we parameterized the spectral data through separation of the periodic and aperiodic components of the signal via the exposed FOOOF Python toolbox (version 1.0.0; https://fooof-tools.github.io/fooof/) (Donoghue et al., 2020). The line fitting with “1/f noise” is comprised of three parameters: a spectral slope (aperiodic exponent) reflecting the decay rate of the power spectrum, alpha band power (6–14 Hz) and beta band (15–30 Hz) power. Consistent with prior research (Donoghue et al., 2020), and recommendations for a broad fitting range described on the author's website, we extracted aperiodic exponents from the 2–50 Hz frequency range of each power spectrum (aperiodic_mode = ‘fixed’, peak_width_limits = (Keefe et al., 2005; Morrow et al., 2023), max_n_peaks = 8, default settings otherwise). The parametrization was applied to the whole segmented 5-min recording. Then two stages were involved in the further distribution analysis. According to a combined EEG-fMRI study (Jacob et al., 2021), the scalp was divided into three broad cortical regions using the average signal across electrode clusters covering bilateral anterior (Fp1, Fp2, AFz, AF3, AF4, Fz, F1, F2, F3, F4, F5, F6, F7, F8), central (FCz, FC1, FC2, FC3, FC4, C1, C2, C3, C4, C5, C6, CP1, CP2), and posterior (Pz, P1, P2, P3, P4, P5, P6, P7, P8, POz, PO3, PO4, Oz, O1, O2) channels.

2.6. Statistical analysis

Data were analyzed using IBM SPSS Statistics version 24. After conducting Kolmogorov-Smirnov normality tests, a mixed analysis of variance was performed to examine electrophysiological differences related to schizophrenia diagnoses. Diagnoses were treated as a between-group factor, i.e. group (SCZ and HC) * time (pre- and post- task) * recording conditions (eyes open, EO; eyes closed, EC). To control for multiple comparisons, between-group main effects were Bonferroni-corrected to p < 0.05 for the average of all electrodes, p < 0.0167 for three broad cortical regions. Finally, Pearson correlations were conducted to analyze whether medication dosage (chlorpromazine equivalents) correlates with EEG measures, and explore the ability in traditional cognitive tasks, language comprehension tasks and EEG parameters in both groups. These correlations assessed the associations between the aperiodic exponent and cognition T scores in both schizophrenia and healthy control participants. Fisher's z test using cocor (Diedenhofen and Musch, 2015) was used to compare the correlations. As suggested by (Dziego et al., 2023; Jach et al., 2020), effect size and trends were emphasized without correcting for multiple comparisons to identify potential trends. To evaluate whether the sample size of this study (32 SCZ patients and 33 healthy controls, total n = 65) was sufficient to detect meaningful differences, we conducted a post-hoc power analysis, using G*Power software with α = 0.05, total sample size = 65, number of groups = 2, and number of repeated measures = 2 (eye conditions) (Faul et al., 2007).

3. Result

3.1. Demographic and clinical characteristics

Demographic and clinical characteristics are presented in Table 1. There were no significant differences in age or gender between the SCZ and HC groups. As expected, the SCZ group exhibited significantly lower cognitive scores (P < 0.001). Age and schooling years were not included in the linear mixed effects analyses reported below on account of the fact that they made no significant contribution to any statistical model.

Table 1.

Descriptive information.

	Schizophrenia (n = 32) M(SD)	Healthy controls (n = 33) M(SD)	P
Age	30.09(10.46)	29.76(10.71)	0.89
Sex (M: F)	12:20	16:17	0.68
BMI	25.75(3.98)	21.88(3.39)	0.53
Smoker (%)	0	6.06	0.53
Drinker (%)	25.00	69.70	0.002
Schooling years	14.81(3.38)	15.85(2.15)	0.63
Treatment years	8.82(9.07)	–
CPZ equivalents	348.437(239.45)
PANSS
positive	8.91(2.78)	–
negative	13.31(5.95)	–
general	20.81(4.55)	–
total	43.03(10.60)	–
BASC (T score)
Verbal learning	53.04(8.55)	61.68(7.33)	<0.001*
Working memory	49.57(6.35)	54.43(2.98)	0.001*
Motor speed	39.79(10.72)	52.24(6.82)	<0.001*
Verbal fluency	49.29(9.56)	59.85(10.10)	0.004*
Attention and speed of information processing	48.03(7.74)	57.70(5.51)	<0.001*
Executive functions	50.92(7.94)	57.70(5.51)	0.07
Language comprehensive task (Z score)
Humor score	0.002(1.01)	−2.88(3.89)	<0.001*
Humor task answer time	16.66(7.02)	12.32(3.52)	0.002
Metaphor and irony score	24.13(6.07)	28.00(2.15)	0.001
Metaphor and irony score answer time	9.39(6.05)	6.50(2.02)	0.002

The SCZ group have received treatment and medication for an average of 8.82 years (SD = 9.07), with one participant unmedicated, 11 receiving monotherapy and 20 on a combination of two atypical antipsychotic drugs. Moreover, the symptoms experienced by SCZ participants were mild in terms of severity and mainly negative.

3.2. Between-group differences in different neural metrics

3.2.1. Aperiodic exponent

A trend-level main effect of group was observed, with the SCZ group demonstrating a higher aperiodic exponent than the HC group in the anterior (F (56,1) = 4.15, P = 0.046, η² = 0.069, Cohen's f = 0.27, power = 0.65), central (F (56,1) = 4.27, P = 0.044, η² = 0.073, Cohen's f = 0.28, power = 0.70) and posterior locations (F (56,1) = 3.81, P = 0.056, η² = 0.066, Cohen's f = 0.26, power = 0.63). Power analysis is below the conventionally desired threshold of 0.80. This suggests that the sample size may be underpowered to reliably detect smaller effects, particularly after Bonferroni correction, where this effect was no longer significant. Eye condition also showed a main effect on aperiodic exponent in all three locations (anterior: F (56,1) = 68.96, P < 0.001, η² = 0.55; central: F (56,1) = 54.98, P < 0.001, η² = 0.50; posterior: F (56,1) = 31.01, P < 0.001, η² = 0.356), indicating that the aperiodic exponent was significantly higher in the eyes-open condition. No significant interaction between group and eye condition was observed. (see in Fig. 2).

Fig. 2.

Topographic maps showing population average power spectrum in aperiodic exponent, alpha band power and beta band power. The results suggest higher (represented by relatively warmer color) aperiodic activities and lower periodic activities for SCZ compared to HC. Note: EC: eyes-closed condition; EO: eyes-open condition; SCZ: schizophrenia; HC: healthy control; before: before the tasks; after: after the tasks.

3.2.2. Alpha band power

There was a main effect of group, indicating a significant decrease in AP for the SCZ group compared to the HC group in the posterior location (F (56,1) = 11.04, P = 0.002, η² = 0.17, Cohen's f = 0.45). The power for detecting this group main effect was approximately 0.95. This indicates that the sample size was sufficient to detect this moderate-to-large effect. No significant main effect of group in anterior (F (56,1) = 7.28, P = 0.009, η² = 0.12) and central (F (56,1) = 2.57, P = 0.12, η² = 0.12) locations following Bonferroni correction. Similarly, eye condition showed a main effect in the three locations (anterior: F (56,1) = 174.88, P < 0.001, η² = 0.76; central: F (56,1) = 162.12, P < 0.001, η² = 0.75; posterior: F (56,1) = 139.54, P < 0.001, η² = 0.72), indicating that the AP was significantly lower in the eyes-open condition. No significant interaction between group and eye condition was observed.

3.2.3. Beta band power

No significant main effect of group was found for beta band power (BP) in any of the three locations after Bonferroni correction (anterior: F (56,1) = 3.48, P = 0.068, η² = 0.07; central: F (56,1) = 5.45, P = 0.024, η² = 0.10; posterior: F (56,1) = 7.21, P = 0.010, η² = 0.14). However, eye condition showed a significant main effect in all three locations (anterior: F (56,1) = 40.37, P < 0.001, η² = 0.47; central: F (56,1) = 75.87, P < 0.001, η² = 0.61; posterior: F (56,1) = 36.69, P < 0.001, η² = 0.46), indicating that BP was significantly lower in the eyes-open condition. No significant interaction between group and eye condition was observed.

3.3. Aperiodic and oscillatory changes over time and space following language comprehension tasks

To test whether the aperiodic exponent was influenced by the tasks, we compared neural parameters before and after the language comprehension tasks (as an intragroup factor) respectively in SCZ and HC groups, using mixed-design ANOVA. The aperiodic exponent tended to decrease (indicating a flattening of the exponent) after the tasks (as shown in Fig. 2), with a significant change in the central location (F (56,1) = 8.93, P = 0.004, η² = 0.14, Cohen's f = 0.40, power = 0.90). There was no significant interaction between group and trial (F (56,1) = 0.001, P = 1.00, η² < 0.001), and no significant difference between the two groups (F (56,1) = 0.46, P = 0.50, η² = 0.008). However, this trend was not significant in the other locations.

Periodic parameters (AP and BP) remained more stable over the tasks, with no significant differences of AP and BP between pre- and post-task values.

3.4. Relationships between behavioral measures and neural parameters

We next examined the correlations between behavioral measures (T scores of BACS and Z scores of language comprehension tasks performance) and neural EEG measures (tables of the association between behavioral measures and neural parameters is provided in the Supplementary Materials). Before analyzing the correlations between behavioral measures and neural parameters, we validated the independence between medication (CPZ equivalents) and EEG parameters in SCZ (AP: r = 0.021, P = 0.91; BP r = 0.22, P = 0.22: aperiodic exponent: r = 0.37, P = 0.72).

3.4.1. Aperiodic exponent

For SCZ group, there was a significant positive correlation between T scores of attention and speed of information processing and pre-task aperiodic exponent in posterior location (r = 0.35, P = 0.05, power = 0.65). This correlation was not present after the task (r = 0.32, P = 0.09). The aperiodic exponent in the posterior location after tasks showed significant positive correlation with T scores of motoring speed (r = 0.34, P = 0.03, power = 0.63). As given the variability of the aperiodic exponent during tasks in 3.4, we conducted correlation analyses on the Δ-exponent (the variance of the exponent) at the two time points (before and after tasks). The Δ-exponent in central location showed significantly negative correlation with Z scores of humor comprehension tasks (r = −0.42, P = 0.03, power = 0.75), indicating better comprehension performance in SCZ group with a lower aperiodic exponent. This correlation was not found for the HC group (r = −0.32, P = 0.08, power = 0.65) (see Fig. 3).

Fig. 3.

The correlation between the variability of the aperiodic exponent and Z scores of humor comprehension tasks. Note: Δ-exponent: the variance of the exponent before and after the tasks.

3.4.2. Periodic oscillations

There was no significant correlation between AP and behavioral measures in either SCZ group or HC group. Before tasks, BP showed no correlation with behavioral measures. For SCZ, pre-task BP in posterior location showed positively correlated with humor comprehension (r = 0.55, P = 0.007, power = 0.95) and metaphor and irony comprehension (r = 0.52, P = 0.011, power = 0.92) in EC; post-task BP in the anterior location showed negative correlation with metaphor and irony comprehension (r = −0.44, P = 0.027, power = 0.78); and in the central location, the post-task BP showed negative correlation with metaphor and irony comprehension (r = −0.60, P = 0.001, power = 0.98). The variations of AP and BP showed no significant correlation with behavioral measures.

4. Discussion

This study evaluated the disparities in aperiodic exponent and periodic oscillations (AP and BP) of EEG between individuals diagnosed with SCZ and healthy controls (HC), and examined the associations between these neural parameters and behavioral measures, with three critical findings emerging: (1) The SCZ group exhibited lower AP and BP, along with a trend-level higher aperiodic exponent (although the aperiodic variance does not remain significant after Bonferroni correction), compared to the HC group; (2) The aperiodic exponent tended to decrease in central location in both groups, while periodic oscillations remained more stable after the task; (3) In the SCZ group, attention and speed of information processing, as well as motoring speed, positively correlated with the aperiodic exponent, and the variance of aperiodic exponent negatively correlated with performance of humor comprehension task. Additionally, the parameters were all observed exclusively in the eyes-open condition. Below, we discuss the results pertaining to each finding.

4.1. EEG parameter differences across groups

Our results showed a steepened PSD, i.e. higher aperiodic exponent, (though not significantly following Bonferroni corrections) with lower AP and BP in SCZ, similar to previous studies (Molina et al., 2020). Additionally, the eyes-open condition - often taken as an index of higher cortical arousal - was linked to greater levels of arousal aligned with past research (Hill et al., 2022)(human) and (Østergaard et al., 2024) (animal). Although vigilance state reliably modulates the aperiodic exponent in clinical and preclinical settings, the relatively brief five-minute recordings should maintain a stable arousal level throughout each condition.

The parameters of EEG, including the aperiodic exponent and periodic oscillation characteristics, are thought to reflect the dynamic E/I balance within cortical microcircuits (Liu et al., 2021), operating across multiple timescales and scalp locations. At a global network level, some studies have proposed that the aperiodic exponent may index E/I balance, based on computational modeling predictions (Gao et al., 2017), magnetoencephalography (MEG) evidence (Muthukumaraswamy and Liley, 2018) and EEG changes induced by Ketamine (Wang et al., 2017) - though this remains debated. As for the E/I balance of local circuits, periodic oscillations are paced by networks of inhibitory interneurons and gated by GABAA action (van Bueren et al., 2023). Recent evidence has questioned the reliability of the aperiodic exponent as a specific marker of cortical E/I ratio (Salvatore et al., 2024). Emerging research suggests the relationship between aperiodic exponent and E/I balance demonstrates complex, non-linear associations that appear to be shaped by multiple interacting factors, including synaptic kinetics, excitatory-inhibitory ratio, and network dynamics, as demonstrated through biophysical modeling and EEG studies of propofol administration in humans (Brake et al., 2024). Based on past modeling that low frequency power is dominated by inhibitory currents (Gao et al., 2017), a steeper exponent can signify greater inhibition compared to excitation, reflecting decreased asynchronous background neuronal firing, potentially correlating with the dysfunction of GABA-mediated inhibitory processes and neuroplasticity in SCZ (Chiu et al., 2018). The dysfunction of inhibitory interneurons and abnormalities in cortical connectivity integrity may be associated with lower rates of decoupling from an oscillatory carrier frequency, thereby reducing interregional oscillatory coherence, i.e. lower AP and BP, which aligned with the perspective of dynamic network communication (Østergaard et al., 2024).

4.2. Aperiodic activities are modulated by task

Neurophysiological signals are nonstationary, showing dynamic changes over time as a function of endogenous and exogenous factors. Our study observed a reduction in the aperiodic exponent following language comprehension tasks, aligning with previous research (Kosciessa et al., 2021), while periodic activity (both AP and BP) remained robust. The tasks, which involved humorous, metaphorical, and ironic texts, required participants to discern meanings beyond the literal, thus engaging higher-level cognitive resources (Mo et al., 2008). As exposure to the tasks, participants may have become more adept at allocating attention to cues relevant for successful sentence interpretation. Increased attentional modulation in accordance with comprehension may have been accompanied by increased E/I ratio, which reflects a flattened aperiodic exponent involved in processing task-relevant information. This transient E/I shift may facilitate neural ensemble flexibility for generating alternative semantic interpretations. Simultaneously, the figurative language tasks required sustained allocation of top-down attention to resolve semantic ambiguities. Flattening of the aperiodic exponent may index increased neural gain - referring to the higher transition entropies and less constrained future values (Dziego et al., 2023) - within task-relevant networks. It reveals fine-grained temporal dynamics underlying varying cognitive workload (CWL), which was defined as the amount of brain resources required for an individual to complete a task (i.e., cognitive activities requiring the achievement of a particular goal) (Chikhi et al., 2022). On the other hand, the stability of periodic components may reflect the sustained nature of top-down attentional control required throughout these complex language tasks, potentially involving different neurophysiological mechanisms than the transient aperiodic changes linked to moment-to-moment signal gain.

4.3. Associations between behavioral measures and EEG parameters

Growing evidence indicates that the aperiodic exponent is a physiologically distinct component that coexists with periodic oscillations in neural signals and may underpin a range of cognitive and behavioral states in SCZ (Molina et al., 2020). The alpha band has been proposed as a neural marker of general cognitive ability (Woodman et al., 2022). However, our study found no significant correlation between AP and behavioral measures in either group after isolating alpha band activity from the influence of the aperiodic exponent using FOOOF, and found varying correlations between BP and comprehensive task performances in different scalp locations within the SCZ group, which aligned with the role of BP in complex cognitive tasks like language processing (Weiss and Mueller, 2012). Whereas motor-related beta changes mostly have been found at central sites, language-related changes were predominantly found at left frontal and parietal sites. However, considering previous analysis methods on periodic oscillations (see introduction), the characteristics should be for reference only. The lack of spatial specificity in the present study may imply that the functional characteristic of BP is largely global in SCZ, but it could also be due to insufficient spatial resolution of EEG measurement. The spatial pattern of such periodic oscillations and cognitive behavior relationship may be rather investigated with a neuroimaging approach characterized by higher spatial resolution (e.g., high density EEG, fMRI), which is a relevant topic for future research.

We found a significant positive correlation between the aperiodic exponent and attention and information processing speed in the SCZ group, i.e. steeper PSD for SCZ group with quicker processing. This seems to contrast with the finding that aperiodic exponent in SCZ is higher than HC. However, it has been suggested that the aperiodic exponent may differentially relate to cognitive abilities at different cognitive levels (Pei et al., 2023). It is assumed that more variable and adaptive “noise” leads to a steeper 1/f slope, which is associated with faster processing speed (relying on the speed of propagation) but also exhibits low transition entropies, more redundant activity patterns, and more constrained future values (Medel et al., 2020). In other words, a steeper PSD may enhance performance on simple cognitive tasks, while a flatter PSD may be beneficial for learning under more dynamic conditions. Furthermore, the aperiodic exponent also serves as a sensitive indicator of high-level cognitive impairment (the defining distinction between high- and low-level processes (Evans and Stanovich, 2013)). This perspective could also explain our finding that the variance of aperiodic exponent negatively correlated with humor comprehension task performance. That is, the decrease in the spectral exponent after tasks may reflect increased CWL in SCZ, acting as a maker of “neural variability” that enables the brain to dynamically adjust its neural activity to meet the demands of a given task or situation. The larger variance of aperiodic exponent, the greater neural variability, which may be associated with unstable representations of goal-information. The absence of this correlation in healthy controls likely stems from preserved neurochemical flexibility: Efficient E/I modulation enables intact individuals to maintain optimal aperiodic dynamics without overt behavioral correlations, as these adjustments occur within typical physiological ranges (Sohal and Rubenstein, 2019). Additionally, threshold effects might also be a contributing factor. The complexity level of the task was insufficiently high (i.e., the tasks were relatively simple for HC) to induce a significant change in the aperiodic exponent unless the neural circuitry had to mobilize additional resources to accomplish the task.

Accumulating evidence underscores a robust association between figurative language deficits and impairments in socio-cognitive functioning across psychotic disorders (Adamczyk et al., 2024). This established link advocates for the systematic integration of pragmatic communication training into existing rehabilitation protocols, particularly as a dedicated module within metacognitive and social skills interventions. Notably, our findings extend this framework by revealing task-dependent aperiodic parameter alterations during figurative language processing. By quantifying interindividual variability in neurophysiological responsivity, aperiodic measures may guide personalized remediation and objectify treatment monitoring.

5. Conclusion

In the present work, we investigate the aperiodic and periodic activities of EEG in individuals with and without SCZ before and after the language comprehension tasks. Our analysis of aperiodic exponent modulations - specifically their task-dependent flattening during cognitive engagement and post-task recovery - suggests that the brain dynamically adjusts its neural activity to meet situational demands. This contributes to the expanding body of literature that underscores the significance of aperiodic activity in cognitive processes. Furthermore, our findings highlight the functional significance of separable aperiodic components in higher-order cognition, as evidence by the correlations between the variance of the exponent and comprehension performance in SCZ group. It is hoped that our results will serve as a foundation for future research, focusing on the application of aperiodic and periodic EEG parameters in the study of cognitive impairment in SCZ, thereby advancing our understanding of this complex condition. Building upon the mechanism, the next step is to investigate the potential effectiveness of the parameters used as a clinical tool for predicting cognitive risk and monitoring the reaction to cognitive treatment.

6. Limitations

It is important to address the limitations of this study alongside its future perspectives. First, the cross-sectional design precludes causal interpretations of observed effects and limits our ability to track disease progression trajectories. Future longitudinal studies should incorporate repeated measures across illness stages (e.g., from prodromal to chronic phases) to determine whether the observed EEG parameters predict clinical transitions or treatment responsiveness. Second, for smaller effects (e.g., aperiodic exponent differences) and moderate correlations, the study may be underpowered. The observed effects could represent Type I errors in exploratory analyses, or conversely, Type II errors in underpowered dimensions, suggesting that future research with larger samples could improve the detection of these effects. Third, schizophrenia exhibits great variability in disease phenotypes and treatments, which could be another potential confounding factor. In order to control the influences of disease duration, we limited the participants to stable phase. It can reduce the generalization of the conclusion, so future research is hoped to build upon more phases of SCZ, such as acute phase of first episode and those at ultra high-risk (UHR) or clinical high-risk (CHR). Although cognitive levels and illness severity were controlled for in the inclusion criteria, purifying the symptom dimension and removing the drug effects will be considered later. Future studies should further assess whether these parameters are useful in assessing objective neural changes following treatment. In addition, of our three dimensions of language comprehension, the tasks may not have been complex enough to fully elicit dynamic changes in oscillatory activity, and the static EEG data collected before and after tasks may have obscured these dynamics. Due to the lack of an experimental paradigm and validated scales, the assessment validity of our experimental materials for the evaluation of language comprehension may be subject to bias. To mitigate this limitation, we conducted intra-group heterogeneity testing. Future experiments necessitate the development of validated scales or paradigms that have undergone rigorous reliability and validity testing. Cross et al. observed dynamic changes in both aperiodic and oscillatory activity during language grammar learning tasks, these interactions between oscillatory and aperiodic activity during the learning task predicted subsequent behavioral performance (Cross et al., 2022). From this perspective, as the cognitive tasks get more dynamic and complex, it is likely that periodic oscillations will exhibit more pronounced changes in interaction with aperiodic activities. From the perspective of EEG parameters, the resting-state EEG contains more unknown information, such as aperiodic offset and microstates, future study is hoped to explore further mechanism of these parameters. This explanation remains tentative at present and requires more systematic examination in future research. Finally, future research can be integrated with other modalities (e.g., fMRI, neurostimulation) to enhance spatial resolution or mechanistic insight.

CRediT authorship contribution statement

Kexin Zhang: Writing – original draft, Methodology, Formal analysis, Data curation. Yunfei Ji: Data curation. Xiaodong Guo: Data curation. Tianqi Gao: Data curation. Xuemin Zhang: Data curation. Xin Yu: Visualization, Supervision, Funding acquisition, Conceptualization. Jing Wang: Writing – review & editing, Funding acquisition, Conceptualization.

Ethics approval statement

Ethical approval was provided by the Ethics Committee of Peking University Sixth Hospital (2022-46 and 2023-46).

Funding

This work was supported by the National Natural Science Foundation of China (No. 82171500 and No. 32070589), and Peking University Medicine Fund of Fostering Young Scholars' Scientific & Technological Innovation (BMU2022PY019).

Declaration of competing interest

The authors would like to thank the seniors who helped with data collection. The authors have no conflicts of interest to report.

Appendix A. Supplementary data

Supplementary tables