Differential Neural Dynamics in Psychomotor Retardation and Agitation of Depression

ABSTRACT

Psychomotor disturbances like agitation and retardation are key symptoms of major depressive disorder (MDD). Despite their clinical significance, the underlying neural mechanisms, for example, motor or psychomotor, remain yet elusive. This study aimed to investigate whether psychomotor agitation and retardation in MDD are associated with alterations in brain dynamics. A total of 119 patients with MDD and 94 HCs were recruited and undertaken fMRI testing. Brain dynamics was measured by the time delays, the lag propagation of global to somatomotor network (SMN) resting state functional connectivity (FC, e.g., lag propagation). Lag propagation of global to SMN FC was delayed in retarded MDD compared to both agitated MDD (t = 3.256, pFDR = 0.006) and HC (t = 2.493, pFDR = 0.041). Further, we observed a significant correlation of the severity of agitation and retardation, measured by the Hamilton depression scale, with global to local SMN's time delays, respectively (agitation: r = −0.19, p = 0.04; retardation: r = 0.32, p = 0.03). Finally, early global to SMN delays predicted a close association of agitation and anxiety levels (F = 5.18, p = 0.025). In contrast to these results in global‐to‐SMN dynamics, no significant delay changes were observed in the local intra‐network SMN dynamics. Together, our findings show distinct neural dynamics in MDD psychomotor retardation, for example, delayed, and agitation, for example, early in global to local SMN functional connectivity. This supports the psychomotor over the motor model of psychomotor retardation which carries major implications for clinical diagnosis and therapy.

This study investigates the dynamic functional connectivity underlying depression with psychomotor disturbance. Findings identified a differential lag propagation pattern between agitated and retarded depression related to the somatomotor network. Notably, the manifestation of symptom‐brain correlations in opposing directions accentuates the distinctiveness of each subtype.

Abbreviations

A‐MDD

agitated major depressive disorder

EPI

echo‐planar imaging

eTS

edge‐center timeseries

FD

frame‐wise displacement

fMRI

functional magnetic resonance imaging

FOV

field of view

HAMA

the Hamilton Anxiety Rating Scale

HAMD

Hamilton Depression Rating Scale

QC

quality control

R‐MDD

retarded major depressive disorders

SMN

somatomotor network

STPP

spatiotemporal psychopathology

TD

time delay

TE

echo time

TR

repetition time

1. Introduction

Major depressive disorder (MDD) is a highly prevalent psychiatric illness that inflicts significant social and economic consequences. Nearly 300 million individuals are afflicted by MDD globally (Evaluation 2021). Among the patients with depression, about 70% of them have psychomotor disturbances like psychomotor retardation and agitation (Ulbricht et al. 2018). These are critical in early clinical diagnosis of MDD onset (Bennabi et al. 2013). Psychomotor retardation is manifest in abnormal slowness of movements and cognition, while agitation presents with excessive movements and irritability (Parker and Paterson 2014). Considering the distinct presentation of retardation and agitation, previous studies suggest that MDD subjects can be classified into subtypes based on their psychomotor symptoms, namely agitated depression (A‐MDD) and retarded depression (R‐MDD) (Akiskal et al. 2005; Sobin and Sackeim 1997). However, the neural mechanisms underlying both psychomotor retardation and agitation in MDD remain yet unknown. Addressing this gap in our knowledge is the goal of our study.

The observation of psychomotor disturbance, characterized as a movement deficit, has sparked a debate between motor and psychomotor hypotheses (Figure 1A). The motor hypothesis posits that psychomotor disturbance originates from localized dysfunction within the somatomotor network (SMN) and can therefore be treated with single‐target SMN interventions. This view is supported by evidence from Parkinson's disease animal models (Goldberg et al. 2002; Hendrix et al. 2025) and human neural imaging studies (Wróbel et al. 2025; Xia et al. 2023). Conversely, the psychomotor hypothesis conceptualizes psychomotor disturbance as a mis‐integration of motor and non‐motor functions, including cognitive, sensory, and affective functions (Northoff, Hirjak, et al. 2020; Northoff et al. 2021). Consequently, it suggests that multi‐target approaches addressing both SMN and non‐motor regions would be more beneficial. Various studies show that psychomotor disturbances are closely associated with other brain regions outside the SMN, like the visual cortex (Liu et al. 2022), global cortical neural activity (Lu et al. 2022; Zhang et al. 2020), and subcortical midbrain activity (Conio et al. 2020; Magioncalda et al. 2020; Martino et al. 2020). However, which of the two hypotheses holds, motor vs. psychomotor, remains yet unresolved as no studies so far directly tested them against each other.

FIGURE 1.

Hypothesis and measurement. (A) Scheme for the motor and psychomotor hypotheses. The motor hypothesis focuses on the abnormalities within somatomotor network, while the psychomotor hypothesis emphasizes the aberrant relationship between somatomotor network and other brain areas. (B) Diverging brain dynamic‐psychomotor symptom association between agitated (A‐MDD) and retarded depression (R‐MDD). The neural fluctuation in A‐MDD is earlier than that in the healthy population, while in R‐MDD it is delayed. (C) Toy example of time delay estimation. Time delay estimation was assessed through lagged cross‐covariate functions between each pair of regions, with subsequent parabolic interpolation to determine values. Two metrics were derived from the time delay matrix. First, column‐wise averaging of the time delay matrix yielded the SMN TDp, obtained by extracting and averaging values corresponding to the somatomotor network. Second, intra‐SMN TD was determined by averaging values in the lower triangle of the SMN‐SMN chunk in the time delay matrix. Methodologically, SMN TDp reflects the temporal structure of the somatomotor network concerning the rest of the brain, while intra‐SMN TD indicates neural fluctuation within the somatomotor network. ATN: attention network; DMN: default mode network; FPN: front‐parietal network; SAN: salience network; SMN: somatomotor network; VIS: visual network.

The temporal structure of the brain activity characterizes the dynamics of the neural oscillations of the brain (Buzsáki 2006; Northoff 2024). At the macroscopic level, as measured by fMRI, neural dynamics are reflected in the relative fluctuations of the BOLD signal between brain regions, representing the temporal segregation and integration of brain function (Chen et al. 2017). Clinically, aberrant brain dynamics are increasingly recognized as a potential biomarker for indicating the neural basis for mental disorders like MDD (Javaheripour et al. 2023; Wang et al. 2022). On a clinical level, the dynamics of retardation are characterized by slow movements (Goldsmith et al. 2016; Haroon et al. 2016) while agitation features abnormally fast movements (Mahlberg et al. 2008). Following the framework of spatiotemporal psychopathology (Northoff 2016a, 2016b, 2018; Northoff et al. 2023), we assume more or less analogous dynamic changes in the brain's neural activity in retarded and agitated MDD subjects, for example, being slow and delayed or being fast and rapid.

In the present study, we aimed at identifying the dynamic alterations in the brain activity of MDD subjects with psychomotor agitation or retardation. Utilizing resting‐state functional magnetic resonance imaging (fMRI), we aimed to investigate whether the time delays, as measured by lag propagation, in local SMN or global‐to‐local SMN functional connectivity differ between and relate to agitation and retardation in MDD. Two hypotheses were formulated accordingly: (1) MDD subjects with agitation would exhibit rapid early lag propagation from the whole brain to the SMN, for example, global to local SMN, compared to retarded MDD and healthy populations, while those with retardation would demonstrate slow delayed lag propagations (refer to Figure 1B). (2) the global‐to‐local SMN lag propagation patterns would prove superior to the ones within local SMN in discerning the two psychomotor disturbances in MDD including robust correlations with psychomotor symptom severity.

2. Materials and Methods

2.1. Participants

A total of 119 inpatients diagnosed as major depressive disorder according to DSM‐5 were recruited from the Department of Depression in Shenzhen Kangning Hospital (Table 1). The 17‐item Hamilton Depression Rating Scale (HAMD) and the Hamilton Anxiety Rating Scale (HAMA) were evaluated during the first week of the patient's hospitalization. Exclusion criteria comprised: (1) age below 18 years, (2) comorbidity with other mental disorders, (3) HAMD total score less than 17, (4) significant prior neurological conditions, including stroke and substantial head trauma, and (5) contraindications to MRI scanning. As a comparison, 94 adult participants without psychiatric illness or a history of mental disorder were recruited as health controls (HC).

TABLE 1.

Demographic data for participants.

	HC	A‐MDD	R‐MDD	ns‐MDD
	(N = 94)	(N = 44)	(N = 27)	(N = 48)
Gender
Female	33 (35.1%)	34 (77.3%)	21 (77.8%)	36 (75%)
Male	61 (64.9%)	10 (22.7%)	6 (22.2%)	12 (25%)
Age
Mean (SD)	34.4 (11.3)	34.9 (12.3)	31.9 (13.4)	32.9 (12.9)
HAMD
Mean (SD)	\	30.8 (6.13)	25.8 (5.19)	24.9 (5.08)
Median [Min, Max]	\	32 [17, 41]	27 [17, 34]	25 [17, 36]
HAMA
Mean (SD)	\	24.8 (7.17)	20.7 (5.68)	19.1 (6.48)
Median [Min, Max]	\	25 [7, 38]	22 [10, 31]	19 [5, 35]
Pure depression
Mean (SD)	\	15.1 (2.92)	20.7 (5.68)	19.1 (6.48)
Median [Min, Max]	\	15 [6, 21]	22 [10, 31]	19 [5, 35]

Note: In the parenthesis, the percentage represents the proportion out of the total for a specific column.

Abbreviations: A‐MDD: agitated major depressive disorder; HAMA: Hamilton Anxiety rating scale; HAMD: Hamilton depression rating scale; ns‐MDD: non‐specific major depressive disorder in this study; Pure depression: sum of HAMD item 1, 2, 7, 8, 10, 13; R‐MDD: retarded major depressive disorder; SD: standard deviation.

For the patients and HCs, additional exclusion criteria included: (1) subpar MRI image quality, and (2) excessive head motion during fMRI scanning. The assessment of head motion relied on frame‐wise displacement (FD), with mean FD values exceeding 0.25 or the occurrence of spike volumes exceeding 25% of the timeseries denoting excessive head motion. The present study was conducted under the approval by the Shenzhen Kangning Hospital Medical Ethics Committee (No. 2020‐K005‐1).

2.2. Psychomotor Subtype Identifications

The MDD patients with noticeable agitation symptoms were categorized as agitated MDD (A‐MDD), as per a previous study (Akiskal et al. 2005). Specifically, individuals scoring greater than 1 on HAMD item 9 (agitation) were identified as having A‐MDD, indicating at least “Slight agitation.” Similarly, patients with a score greater than 1 on HAMD item 8 (retardation) were categorized as retarded depression (R‐MDD). To prevent the overlap of A‐ and R‐MDD, individuals with R‐MDD were additionally required to score 0 on HAMD item 9. It is important to note that the HAMD assessment was conducted during the initial week of hospitalization, so we could exclude the confound of secondary agitation which is induced by the side effect of antidepressant (Benazzi 2004).

2.3. Resting‐State fMRI Acquisition and Preprocessing

The fMRI data were acquired using a 3.0 Tesla MRI system (Discovery MR750, General Electric, Milwaukee, WI, USA) with an eight‐channel head coil at Shenzhen Kangning Hospital. Participants were instructed to keep still with their eyes closed, remaining conscious to prevent the heightened visual network connectivity from interfering with the brain's dynamic connectivity patterns (Costumero et al. 2020; Han et al. 2023). A total of 240 whole‐brain functional images were collected using an echo‐planar imaging (EPI) sequence with 2000 ms repetition time (TR), 30 ms echo time (TE), a flip angle of 90°, in 33 axial slices, and an isotropic resolution of 3.5 mm. Additionally, high‐resolution anatomical MRI scans were conducted using a fast‐field echo (FFE) three‐dimensional T1‐weighted (3D‐T1WI) sequence at 1 mm isotropic resolution (TR/TE = 6.65/2.93 ms, flip angle = 12°) and 192 contiguous sagittal slices.

Images were preprocessed via fMRIPrep ver. 22.1.1 (Esteban et al. 2019), which included: (1) discarding the initial 10 frames, (2) slice‐timing correction, (3) head‐motion correction with six parameters, and (4) spatial normalization to MNI space. After preprocessing, timeseries of brain signal were extracted over Power‐264 atlas (Power et al. 2011), regressing out the head‐motion parameters, detrending, band‐pass for 0.01 to 0.1 Hz and smoothing for 6 mm full‐width half‐maximum kernel. The regression for the global signal was not performed to ensure the precision of calculating the dynamic metric (Raut et al. 2021) as well as the inclusion of the global functional connectivity (Scalabrini et al. 2020). Referring to the previous study (Watanabe and Rees 2017), we excluded the subcortical and cerebellar regions in Power‐264 atlas, resulting in 214 parcels. Nuisance regression and blood‐oxygen level dependent (BOLD) signal extraction was performed using a Python package Nilearn (https://nilearn.github.io/dev/index.html).

2.4. Time Delay Estimation

To assess the time shifts in somatomotor network, we calculated Time delay (TD) which exhibits stereotyped propagation patterns across the brain (Mitra et al. 2014) (Figure 2). First, the shift in TR units was computed between two time series using the cross‐variance function (CCF). Subsequently, a three‐point parabolic interpolation was applied on the shift TR, the one TR prior (TR‐1), and the one TR later (TR + 1). The TD was determined as the peak value obtained from this interpolation. This process was repeated for each pair of ROIs, culminating in a 214 × 214 TD matrix for each participant (see Supporting Information and Raut et al. (2019) for further details).

FIGURE 2.

Clinical profiles of psychomotor subtypes. (A, B) the distribution of psychomotor disturbance in all patients. (C) Group difference among all subgroups of patients in anxiety and depression severity. Correlation between agitation and anxiety in all patients. A‐MDD: agitated major depressive disorder; HAMA: the Hamilton Anxiety Rating Scale; HAMD: the 17‐item Hamilton Depression Rating Scale; ns‐MDD: non‐specific major depressive disorder; R‐MDD: retarded major depressive disorder. *p < 0.05, **p < 0.01, ***p < 0.001.

Two metrics, namely the SMN time delay projection map (SMN TDp) and intra‐SMN time delay (intra‐SMN TD), were derived from the TD matrix. To compute SMN TDp, we initially calculated a time delay projection map through column‐wise averaging of the TD matrix, yielding a vector with 214 elements. Subsequently, SMN TDp was obtained by averaging all parcels within SMN from the time delay projection map. For intra‐SMN TD, this metric was determined by directly averaging the lower triangle of the chunk in the TD matrix corresponding to SMN.

SMN TDp was defined as the temporal lag of the SMN relative to the rest of the brain. This metric represents a relative temporal feature that captures the relationship between the SMN and the broader brain network, encompassing both inter‐ and intra‐network connectivity. In contrast, intra‐SMN TD was defined as the relative lag structure among regions within the somatomotor network itself, signifying its internal temporal organization. Time delay estimation was performed using in‐house scripts, adapted from the methodology provided by Raut et al. (2019).

Variations in BOLD oscillations were shown to occur across distinct slow frequencies (Zuo et al. 2010). A previous study demonstrated that the SMN abnormality in different frequency bands was related to the distinction between depression and mania (Martino et al. 2016). Therefore, for further demonstrating the time lag structure of SMN in specific frequencies, we also conducted time delay analysis in the slow‐4 band (0.027–0.073 Hz) and slow‐5 band (0.01–0.027 Hz).

2.5. Statistical Analysis

To characterize the severity of depression and anxiety among MDD subtypes, we conducted a comparative analysis of the HAMD and HAMA total scores among A‐MDD, R‐MDD, and the remaining patients classified as non‐specific MDD (ns‐MDD) using one‐way ANOVA. Concurrently, to illustrate the relationship between patients' anxiety levels and agitation symptoms, we computed Pearson's correlation between the total score on the HAMA and the agitation score in the HAMD.

To examine the associations between the time lag structure of SMN and psychomotor disturbance, we initially assessed the correlation between SMN TDp and psychomotor disturbance. This involved calculating Pearson's correlation coefficients between SMN TDp and both the agitation score (HAMD item 9) and the retardation score (HAMD item 8). To mitigate the influence of agitation in the correlation between SMN TDp and the retardation score, only patients with a score of 0 on agitation were included in the SMN TDp‐retardation correlation analysis.

ANOVAs were conducted to examine mean differences in SMN TDp among the A‐MDD, R‐MDD, and HC groups, with relevant covariates controlled. In contrasting the two MDD subtypes, we controlled for depression and anxiety severity to eliminate the impact of non‐psychomotor symptoms. To control for depression level, we computed a “pure depression” score from HAMD by summing items 1, 2, 7, 8, 10, and 13, following the method outlined by Bech (2012). In contrasting between MDD and HC, age and gender were controlled. The above analytic process was conducted in the typical, slow‐4 and slow‐5 frequency band, respectively. Moreover, as a comparison, the identity analysis was also performed using intra‐SMN TD.

To illustrate the influence of SMN TDp on the interplay between anxiety and agitation, we constructed a moderation model where SMN TDp served as the moderator, with agitation score and HAMA total score as the independent and dependent variables, respectively. This moderation model was tested across all three frequency bands. Subsequently, simple slope analysis was conducted to dissect the moderated effects at various levels of SMN TDp.

Statistical significance was determined by a p‐value less than 0.05. For the group contrasts related to multi‐comparison of time delay estimation, p‐values were subjected to correction using the false discovery rate (FDR) method, and a q‐value less than 0.05 after FDR correction was considered as statistically significant. All statistics were conducted using R v4.1.2.

3. Results

3.1. Clinical Profile of Agitated and Retarded Psychomotor MDD Subtypes

According to our classification strategy (see methods), 44 patients were identified as A‐MDD while 27 as R‐MDD (Table 1). While 48 patients were categorized as non‐specific MDD (ns‐MDD), no difference was found on age, gender, and education level between the three subgroups of MDD. As shown in Figure 2C, we found that the total HAMD score was the highest in A‐MDD, intermediate in R‐MDD, and lowest in ns‐MDD (A‐MDD vs. R‐MDD: t = 3.733, pFDR < 0.001; A‐MDD vs. ns‐MDD: t = 5.191, pFDR < 0.001). However, the pure depression factor indexing the core depression symptoms (see methods) did not show significant difference between A‐ and R‐MDD (t = 0.772, pFDR = 0.442). This suggests that the nature of the psychomotor changes, for example, agitation vs. retardation, is, in part, independent of the overall depression severity.

Concerning anxiety, the HAMA score was the highest in A‐MDD (A‐MDD vs. R‐MDD: t = 2.581, pFDR = 0.017; A‐MDD vs. ns‐MDD: t = 4.175, pFDR < 0.001). Moreover, we observed a positive significant correlation between HAMA score and agitation severity (r = 0.36, p < 0.001, Figure 2D). This suggests that higher degrees of agitation are accompanied by higher anxiety severity. Given that both depression and anxiety severity showed differences between agitated and retarded MDD subjects, we included their measures as covariates in all subsequent analyses.

3.2. Differential Temporal Patterns in Global to SMN Functional Connectivity in Agitated and Retarded MDD

Next, we investigated whether the abnormal timing of the movement behavior in A‐MDD and R‐MDD is related to analogous timing changes in their brain's neural activity. For that purpose, we investigated the lag thread functional connectivity of the global brain with the somatomotor network. The lag thread functional connectivity measures the temporal features of the functional connectivity by probing different degrees of time lags between the source regions, for example, the global brain's cortex in our case, and the target regions, for example, the somatomotor network in our case (Mitra et al. 2023, 2014).

We first investigated the differences between A‐MDD and R‐MDD in the lag thread functional connectivity, for example, TDp, from the global brain's cortex to the SMN. This yielded significant differences between the three MDD subgroups. Specifically, Global to SMN TDp in R‐MDD was more delayed than that in both A‐MDD and HC (A‐MDD vs. R‐MDD: t = 3.256, pFDR = 0.006; HC vs. R‐MDD: t = 2.493, pFDR = 0.041, Figure 3D). Though not reaching significance, the mean values show earlier SMN TDp in A‐MDD than those in HC (t = 0.263, pFDR = 0.793), which the trend is consistent with the psychomotor hypothesis. This is well in accordance with the distributions that shows earlier TDp in A MDD, intermediate time lags in HC, and delayed TDp's in R MDD (see Figure 3D).

FIGURE 3.

Lag structure in somatomotor network and psychomotor subtypes. (A) Illustration of SMN time delay projection map for each ROI in somatomotor network for different groups. (B, C) Correlation between SMN TDp and agitation and retardation. (D) Group difference of SMN TDp for psychomotor subtypes and healthy controls. Curve indicates the distribution of SMN TDp in each group, as well as the dots. SMN TDp: Somatomotor network time delay projection map; A‐MDD: agitated major depressive disorder; HC: healthy controls; R‐MDD: retarded major depressive disorder. *p < 0.05, **p < 0.01, ***p < 0.001.

The relationship of global to SMN functional connectivity time lags with the psychomotor behavior in MDD is further supported by the correlation findings. A significantly negative correlation was found between global to SMN TDp and agitation score (r = −0.19, p = 0.04, Figure 3B). This indicates that an earlier functional connectivity from the global brain to the SMN leads to more severe agitation symptoms. Moreover, a significantly positive correlation was found between global to SMN TDp and retardation score among MDD patients without agitation (r = 0.32, p = 0.03, Figure 3C), indicating that delayed SMN TDp was linked to heightened retardation symptoms.

Finally, it shall be mentioned that global to SMN TDp in all MDD subjects or psychomotor subgroups did not correlate with HAMD (all patients: r = −0.12, p = 0.19; A‐MDD: r = 0.09, p = 0.55; R‐MDD: r = 0.01, p = 0.94) nor with HAMA (all patients: r = 0.04, p = 0.64; A‐MDD: r = 0.07, p = 0.64; R‐MDD: r = 0.09, p = 0.65). This further suggests that the observed global to SMN TDp changes are really related to the psychomotor changes themselves rather than the overall depression and anxiety severity.

Together, our findings show differential temporal changes in the global cortical functional connectivity to local SMN in agitated and retarded MDD. Specifically, functional connectivity occurs abnormally early in A‐MDD while it is delayed in R‐MDD compared to healthy subjects. This is further supported by analogous correlation findings with the psychomotor severity scores of both agitation and retardation. Finally, there is no correlation of global to local SMN TDp with overall depression and anxiety severity, indicating its specificity for psychomotor changes.

3.3. No Local Intraregional Temporal Changes in Within Network SMN in Agitated and Retarded MDD

We next tested whether this temporal pattern is specific for the global to SMN functional connectivity or, alternatively, holds locally within the SMN itself. Unlike in the global cortical to local SMN functional connectivity, we did not observe any differences in the temporal patterns of the intra‐network SMN functional connectivity. Intra‐SMN TD did not show significant differences among the three groups (A‐MDD vs. R‐MDD: t = 0.055, pFDR = 1; HC vs. R‐MDD: t = 0.879, pFDR = 1; HC vs. R‐MDD: t = 0.137, pFDR = 1, Figure SC). Moreover, no significant correlation was observed between intra‐SMN TD and agitation or retardation score (agitation: r = −0.13, p = 0.16; retardation: r = −0.27, p = 0.07, Figure S1A,B).

Together, we show that the temporal pattern holds only on a global cortical to local SMN level but not locally intra‐network within the SMN itself as the latter analyses neither yielded subgroup differences nor correlations. This suggests a global to local rather than purely local source of the changes in temporal patterns in the psychomotor subtypes.

3.4. Temporal Patterns of Functional Connectivity in Slower and Faster Sub‐Frequency Bands in Agitated and Retarded MDD

We next raised the question whether the changes in the temporal patterns of global to SMN functional connectivity are different for slow and faster frequency bands in the infra‐slow range of fMRI, for example, slow 5 (0.01 to 0.027 Hz) and slow 4 (0.027 to 0.073 Hz). Overall, we observed that the time delay pattern was more blurred in slow‐5 than that in the slow‐4 band with group differences and correlation being observed only in the latter (Figure 4A). For the slow‐4 band, SMN TDp in R‐MDD was more delayed than that in both A‐MDD and HC (A‐MDD vs. R‐MDD: t = 2.793, pFDR = 0.014; HC vs. R‐MDD: t = 3.494, pFDR = 0.003, Figure 4C) while being non‐significant between A‐MDD and HC (t = −0.622, pFDR = 0.535). This is further supported by the finding that Global to SMN TDp in slow 4 was significantly correlated in a positive way with the retardation score (r = 0.44, p < 0.001, Figure 4B), but not with the agitation score (r = −0.09, p = 0.32).

FIGURE 4.

Lag structure in varying frequencies. (A) Illustration of time delay matrix in typical, slow‐4 and slow‐5 bands, corresponding to psychomotor subtypes. (B) Correlation between SMN TDp and retardation score in slow‐4 band. (C, D) SMN TDp in different groups for slow‐4 and slow‐5 bands. A‐MDD: agitated major depressive disorder; HC: healthy controls; R‐MDD: retarded major depressive disorder; SMN TDp: somatomotor network time delay projection map. *p < 0.05.

Together, we observe that the group differences and correlations occur mainly in the faster frequency band of the infra‐slow frequency range, for example, slow 4, rather than in the slower band, for example, slow 5. This suggests that the temporal changes are mainly present in the faster frequency of slow 4 in MDD as retarded MDD subjects show stronger delays in slow 4 than HC and agitated subjects.

3.5. Temporal Patterns of Global to SMN Modulate the Relation of Agitation and Anxiety in MDD

In a final step, we raised the question whether the relation of the temporal patterns of functional connectivity with psychomotor behavior relates to anxiety and depression. This was motivated by the fact that we observed a significant relationship of especially psychomotor agitation with the severity of anxiety in our data (see above). For that purpose, we calculated a moderation model for probing the modulation of the correlation between agitation and HAMA by the global to SMN TDp.

We found a significant moderation effect of SMN TDp on the association between anxiety and agitation (F = 5.18, p = 0.025). Simple slope analysis showed that the correlation between agitation and HAMA score was significant only when SMN TDp exhibited early and average temporal patterns whereas it was non‐significant when SMN TDp was delayed (−1 SD: t = 4.786, p < 0.001; Mean: t = 4.562, p < 0.001; +1 SD: t = 1.539, p = 0.127, Figure 5A,B). In contrast to the global to SMN TDp, the moderation effect of intra‐SMN TD was not significant (F = 1.39, p = 0.241).

FIGURE 5.

Moderation effect of time delay. Panels show the relationship between SMN TDp and anxiety within respect to specific agitation level (dot size) for typical band (A), slow‐4 band (C) and slow‐5 band (E). Simple slope analysis after moderation model is shown to indicate the correlation between anxiety and agitation on different levels of SMN TDp corresponding to typical band (B), slow‐4 band (D) and slow‐5 band (F). The red line represents 1 SD early SMN TDp, the blue line represents the average level of SMN TDp, and the green line represents 1 SD delayed SMN TDp HAMA: the Hamilton Anxiety Rating Scale; SMN TDp: somatomotor network time delay projection map.

The significant moderation effect of global to SMN TDp was also found in slow‐4 band (F = 13.68, p < 0.001). Simple slope analysis revealed that the agitation‐HAMA correlation was significant in early and average global to SMN TDp, but not in the delayed global to SMN TDp (−1 SD: t = 5.778, p < 0.001; Mean: t = 4.765, p < 0.001; +1 SD: t = 0.76 = 51, p = 0.454, Figure 5C,D). Unlike in slow 4, the moderation effect of SMN TDp was non‐significant in slow‐5 band (F = 0.39, p = 0.532, Figure 5E,F).

Finally, to disentangle the effects of psychomotor agitation and anxiety on the temporal patterns of global to local SMN TDp, we divided our sample into four subgroups. These included subjects with high anxiety combined with either high or low agitation as distinguished from those with low anxiety associated with high or low agitation. This yielded an earlier but not significant global to local SMN TDp in the two high agitation groups whereas the high anxiety subgroups exhibited intermediate TDp values (Figure S2).

Taken together, the temporal patterns of global to SMN TDp moderated the relation of anxiety and agitation in especially those subjects showing early and average TDp. This further supports the observed behavioral relationship of agitation and anxiety on neuronal grounds, including its strongly temporal shaping. While our subgroup analysis indicated that the early global to SMN TDp is related to agitation rather than anxiety.

4. Discussion

In this study, we investigated the changes in the temporal patterns of global to local SMN functional connectivity in agitated and retarded MDD. First, using time delay estimation to evaluate brain dynamics, we observed that the global‐to‐local SMN dynamics are earlier and thus faster in the agitated subgroup, which also correlated with the agitation score. While the retarded MDD subjects showed delayed and slower global to local SMN dynamics, including the latter's correlation with the severity of psychomotor retardation. Second, this was further supported by analogous findings in especially the faster frequency band, for example, slow 4, of the infra‐slow frequency range. Third, in contrast to the global to local SMN dynamics, we did not find any such changes in the more local within SMN dynamics. Fourth, we observed that the relationship of anxiety and agitation was moderated by the global to local SMN dynamics, with especially faster timing strengthening their relationship.

Temporal structure alterations vary among psychomotor subgroups. Noticeably, in our study, time delay estimation in retarded MDD was later and thus more delayed than that in HC, while the agitated MDD exhibited earlier, more rapid estimates. Our findings suggest that neural dynamics hold the promise in detecting abnormal patterns of brain activity corresponding to specific psychomotor presentations. The psychomotor syndrome could be used to determine the agitated and retarded subgroups of MDD patients based on their symptomatology (Leventhal et al. 2011). Thereby, dynamics in the lag propagation of global to SMN functional connectivity may be key in identifying and differentiating the two subgroups. Finally, our results highlight the necessity for symptom‐based and individual‐specific treatment tailored to each patient's psychomotor symptoms as indexed by their dynamics, for example, temporal patterns in both brain and behavior (Bennabi et al. 2013; Feczko et al. 2019).

Our findings support the psychomotor hypothesis over the motor hypothesis. Our results show the imbalanced temporal relationship between somatomotor areas and other brain regions (Northoff et al. 2021; Yan 2024). The SMN projection map, estimating the global‐to‐local SMN time delay by comparing the temporal structure of SMN to the rest of the brain (Raut et al. 2019), revealed significant brain‐symptom associations and group differences in brain dynamics. Our analysis also revealed a significant frequency‐dependent effect in the global‐to‐local SMN time delay. This finding supports the view that functional brain architecture is specialized across distinct frequency bands, which has relevant clinical implications (Liu et al. 2014; Zhu et al. 2015). Importantly, our results advance this understanding by demonstrating that psychomotor‐related neural dynamics are likewise frequency‐specific. Notably, the main findings in the present study, including the brain‐symptom association and group differences of brain dynamics, were obtained by the global‐to‐local SMN while they did not relate to the intra‐SMN time delay. This favors the psychomotor over the motor hypothesis.

The psychomotor hypothesis delineates how the primary motor function is modulated by non‐motor function and neural sources outside the motor cortex; hence the prefix “psycho” in psychomotor disturbance (Northoff et al. 2021). This modulation may originate from subcortical areas, such as the basal ganglion and thalamus (Martino et al. 2020), as well as other cortical areas like the default mode network (Conio et al. 2020). The imbalanced function between the SMN and DMN may originate from abnormal interactions between dopamine‐driven and serotonin‐related circuits (Liang et al. 2025; Northoff et al. 2021; Song et al. 2021). Future studies could utilize multimodal neuroimaging tools, such as magnetic resonance spectroscopy or positron emission tomography, to identify the precise neurotransmitter sources involved (Gao et al. 2025).

Together, we demonstrate differential neural dynamics in retarded and agitated MDD. Interestingly, the changes in neural dynamics seem to correspond to those on the behavioral level. While retarded MDD subjects show abnormal slowness on both neural and behavioral levels, the agitated MDD subjects exhibit the opposite, namely abnormal fastness. This is well in line with psychomotor retardation and agitation being primarily dynamic or temporal disturbances as suggested in Spatiotemporal psychopathology (Northoff 2016a, 2016b, 2018). Moreover, our findings support the assumption of analogous and seemingly shared temporal changes on both neural and behavioral levels, for example, slower or faster speed, as their “common currency” (Northoff, Hirjak, et al. 2020; Northoff, Wainio‐Theberge, and Evers 2020).

The present study points to two key clinical implications. First, the finding of dysregulated connectivity between the SMN and other networks provides a rationale for pursuing multi‐target neuromodulation therapies (e.g., multi‐target TMS), which could prove more effective than single‐target methods. Second, the observation of accelerated neural dynamics in A‐MDD suggests that therapeutic interventions aiming to directly normalize this specific dynamic abnormality may represent a more efficient path to alleviating psychomotor symptoms.

4.1. Methodological Strengths and Limitations

The present study shows major strength consisting in the large sample size and the different methods of analysis. This allows us to draw direct relationships of neural activity changes with specific forms of behavior, for example, psychomotor retardation and agitation, as distinguished from others like overall depressive symptom severity and anxiety. Moreover, our analysis incorporated the global signal (GS) of the BOLD data, a choice that safeguards meaningful information about global brain dynamics. A growing body of literature suggests that these global waves are integral to network synchronization (Raut et al. 2021; Zhang et al. 2020) and may hold significant value as clinical biomarkers (Lu et al. 2022). Hence, our approach allows us to draw more granular and fine‐grained brain‐behavior relationships beyond just distinguishing traditional disease categories like MDD from healthy subjects. This, as paradigmatically exemplified in our last analyses on relating anxiety and agitation, puts us one step closer toward characterizing neurobehavioral patterns on an individual level as required for clinical diagnosis.

Some limitations shall be mentioned. First, psychomotor disturbance was assessed using a 5‐point Likert scale item from HAMD, which provides a basic, rather than nuanced, evaluation of the symptom. Future research should consider employing more comprehensive assessment tools, such as the questionnaire developed by Parker and McCraw (2017) or the motor‐speed targeting behavioral task, like the nine‐hole task (Lefebvre et al. 2024). Second, it is important to note that while time delay estimation can enhance the temporal resolution of fMRI (Mitra et al. 2014; Raut et al. 2019), it remains a hemodynamic proxy. Due to the physiological limitations of the BOLD signal, it does not provide a precise measurement of the true neural activity lag. Future studies might explore the use of high‐temporal‐resolution devices, such as electroencephalogram or magnetoencephalogram, to examine the relative speed of neural fluctuations. Lastly, medication use during hospitalization may have influenced PMR and brain function. Although we minimized this confounding effect by collecting data during the first week of hospitalization, future studies should aim to replicate these findings in a large sample of drug‐naïve or first‐episode patients with MDD.

5. Conclusions

The present study investigated changes in brain dynamics in MDD subjects with psychomotor retardation or agitation. The lag propagation in global‐to‐local SMN resting state functional connectivity was delayed in retarded MDD compared to that in agitated MDD and HC. Further, the global‐to‐local SMN dynamics were tightly associated with the severity of psychomotor retardation and agitation, respectively. Finally, the link between anxiety and psychomotor agitation was moderated by the level of global‐to‐local SMN dynamics. Notably, these changes were only observed in the global to SMN dynamics but not within SMN itself. This supports the psychomotor over the motor model (Northoff et al. 2021).

Together, our findings show that psychomotor retardation and agitation relate to differential brain dynamics in global to SMN functional connectivity. Remarkably, one can see similar dynamics on both neural and behavioral levels, that is, slow and delayed in retardation while they are fast and rapid in agitation. Although tentatively, our findings suggest similar, if not shared, dynamic features of both brain and behavior as their “common currency” (Northoff, Wainio‐Theberge, and Evers 2020). This is well in line with a global spatiotemporal shaping of psychomotor symptoms as postulated in spatiotemporal psychopathology (Northoff et al. 2023, 2018; Northoff and Duncan 2016; Northoff and Hirjak 2023).

Funding

The work was supported by funding from Guangdong Basic and Applied Basic Research Foundation (2022A1515012503, 2024A1515013203), Nanshan District health system science and technology major project (NSZD2023021), China and Medicine Plus Program of Shenzhen University (2024YG008), Shenzhen Science and Technology Program (JCYJ20240813112900002, JCY20240813114519026) and Medical Scientific Research Foundation of Guangdong Province (A2025094).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: Supporting Information.

Liang, Q. , Xu Z., Chen S., et al. 2026. “Differential Neural Dynamics in Psychomotor Retardation and Agitation of Depression.” Human Brain Mapping 47, no. 2: e70453. 10.1002/hbm.70453.

Data Availability Statement

The codes used in the current study are available through an online repository (https://github.com/QunjunLIANG/psychomotor_timedelay). To replicate the main findings in our study, the time delay metrics were provided. The raw images used in the present study are available from the corresponding author upon reasonable request.