Healthy young (n = 20, 22.5 ± 2.31 y, 9 females, two left-handed) and older (n = 20, 72.7 ± 5.73 y, 6 females, none left-handed) adults participated in this study. The sample size was based on a priori power analysis for repeated measures analysis of variance (ANOVA; effect size: 0.25; power: 0.8). Participants were recruited from Groningen and the surrounding area by local flyers and mouth-to-mouth advertising. They were included if they were free of neurological disorders or other physical and psychological disorders and did not take drugs that affect nerve conduction velocity or cognitive processes, such as antidepressants, antipsychotics or sleep medications. Participants provided written informed consent before participating in a protocol that was approved by the Medical Ethical Review Committee (NL81562.042.22) of the University Medical Center Groningen and was conducted according to the Declaration of Helsinki (2013).

This study adopted a cross-sectional design based on previous work (Marinelli et al. 2017) that was extended with synchronous EEG and EMG recordings. Baseline measures included anthropometrics, handedness [Edinburgh Handedness Inventory, (Oldfield 1971)], physical [Timed Up and GO, (Podsiadlo and Richardson 1991)] and cognitive [Montreal Cognitive Assessment, (Nasreddine et al. 2005)] function. In addition, the Paratonia Assessment Instrument was performed to subjectively assess the presence of paratonia by interpreting the extent of resistance during passive movements (Hobbelen et al. 2008).

The experimental protocol started with a seated control task after which a series of arm movements with the right arm followed while brain (EEG) and muscle (EMG) activity was recorded. During the experiment, participants sat on a chair with their knees flexed at a ninety-degree angle and their feet flat on the ground. Participants performed passive elbow flexion and extension movements with the dominant right arm at three different speeds, i.e., 20, 60 and 100 beats per minute (Fig. 1). The participants were instructed to relax their arm during the passive arm movements while the researcher moved the arm through its full range of motion thirty times (i.e., thirty flexion and thirty extension movements). This research focused on upper-limb movements allowed for a rigorous and controlled investigation of the core neural mechanisms under study and enabled direct comparison with previous studies (Marinelli et al. 2017) without substantially increasing experimental complexity and session duration associated with including lower-limb movements – factors that reduce feasibility in clinical populations. At all three speeds, the passive movement conditions were performed in a continuous (i.e., one beat of the metronome between maximal flexion and maximal extension) and discontinuous (i.e., one beat of the metronome between maximal flexion and maximal extension and one beat of the metronome of pause between the end of a flexion/extension movement and the start of the next extension/flexion movement) fashion, resulting in a total of six conditions. All arm movements were performed on the rhythm of a metronome (Tempo Lite, Frozen Ape Pte. Ltd.), which was exclusively audible to the researcher through earphones. The order of the movement type was pseudorandomized so that half of the participants started with the continuous movements and the other half with the discontinuous movements. The order of movement speed was not randomized and were performed with increasing movement speed as to prevent that hypothesized resistance to passive movements during higher movement speeds influenced conditions with slower movement speeds. Breaks were allowed depending on the participant’s need for rest. Figure 1 provides a schematic overview of the experimental design.

Overview of the experimental design. In healthy younger (n = 20) and older (n = 20) individuals, continuous and discontinuous passive movements were performed at three different speeds (20, 40 and 60 bpm) while 64-channelelectroencephalography (EEG) and biceps and triceps surface electromyography (EMG) data were acquired. bpm, beats per minute

Muscle activity from the right M. biceps brachii and M. triceps brachii were recorded using surface wireless EMG sensors (Trigno Avanti, Delsys, Natick, MA, USA; 27 × 37 × 13 mm). The EMG sensors were placed over the right biceps and triceps brachii muscle bellies in the direction of the underlying muscle fibers guided by palpation during contraction (see Fig. 1), which were determined by palpating the muscle during voluntary muscle contraction. EMG data were acquired with bipolar recordings at a sampling frequency of 1926 Hz.

Electroencephalography (EEG) data were continuously acquired with a 64-channel EEG system with Ag–AgCl electrodes placed on the scalp according to the international 10–20 system (TMSi SAGA64 + , Oldenzaal, the Netherlands). The EEG data were acquired at 2048 Hz with an average reference. The impedance was kept below 10 kΩ. To minimize noise in the EEG signals, participants were asked to not speak, to avoid head movements, to relax the jaw and neck muscles and to minimize eye blinking during the measurements. A hard-wired trigger (0–5 V ramp-up TTL trigger) ensured the synchronous acquisition of EEG and EMG data.

The EEG data were pre-processed in MATLAB R2022b (MathWorks), using scripts based on FieldTrip (Oostenveld et al. 2011). Because the frequencies ranges of interest in this study were alpha (8–12 Hz) and beta (13–30 Hz), the data were filtered with a 3 Hz high-pass filter (6th order Butterworth) and a 70 Hz low-pass filter (6th order Butterworth). This way, we removed low-frequency drifts and high-frequency noise (for example from EMG) while sufficient data was retained for adequate rejection of residual artefacts with independent component analyses (see below). To remove residual electromagnetic line noise, a notch filter with a frequency of 50 Hz and its second and third harmonic (100 and 150 Hz) was also applied. After resampling the data to 1024 Hz with a piecewise cubic hermite interpolating polynomial using Fieldtrip’s ft_resampledata function, EEG channels were visually inspected and empty channels or channels with substantial artifacts were removed. Next, an independent component analysis was performed to remove artifacts caused by eye movements and muscle contractions based on time series and topographical distributions of power. This process was repeated for each condition and consistently performed by the same researcher (AHMV). On average, three components per trial were removed. The preprocessing pipeline concluded with referencing the data to an average reference.

The EMG data was pre-processed in MATLAB R2022b (MathWorks). First, the EMG data were restructured in the FieldTrip format to enable the preprocessing of the EMG data using similar procedures as for the EEG data. The EMG data were bandpass filtered using a 4rd order Butterworth filter between 4 and 100 Hz to remove low-frequency drifts and high-frequency noise while retaining the frequencies of interest (13 – 30 Hz). Additionally, electromagnetic line noise was removed with a 4rd order Butterworth notch filter of 50 Hz and its second and third harmonic (100 and 150 Hz). The EMG data were full-wave rectified using an absolute Hilbert transform and resampled to 1024 Hz, again with a piecewise cubic hermite interpolating polynomial using Fieldtrip’s ft_resampledata function.

The preprocessed EEG time series were epoched into non-overlapping 1-s-long epochs before being transformed to the frequency domain using a multitaper Fast Fourier Transformation with a Hanning window leading to a 1 Hz frequence resolution. Power in the alpha frequency range (8–12 Hz) was averaged for five frontal electrodes [F3, Fz, F4, FC3 and FC4; (Gompf et al. 2017)].

To quantify brain-muscle interactions, EEG and EMG data were concatenated and epoched in non-overlapping 1-s-long segments before being transformed to the frequency domain between 1 and 40 Hz using a multitaper Fast Fourier Transformation method with 5 Hz smoothing. Subsequently, corticomuscular coherence was calculated using the auto- and cross spectra (Halliday et al. 1995). Then, beta-range (13 – 30 Hz) corticomuscular coherence was estimated between the pre-processed EEG timeseries in the electrode positioned over the upper arm region of the primary motor cortex contralateral to the right arm, represented by the C3, and the pre-processed EMG timeseries recorded from the biceps and triceps muscles. For each condition, the level of corticomuscular coherence was quantified as the area under the curve between the corticomuscular coherence estimates and significance line of the amplitude spectrum in the frequency domain (Amjad et al. 1997).

Statistical analyses were performed in SPSS (Version 26, IBM, Chicago, IL, USA). The Shapiro–Wilk test revealed that all data were normally distributed. Descriptive statistics (in means and standard deviations) were presented for baseline age, physical and cognitive performance, and handedness by age group. To examine whether age, movement speed and movement type impacted alpha power data and beta corticomuscular coherence, data were analysed with repeated measures ANOVA for each hypothesis. Movement speed (20, 60 and 100 bpm) and type of movement (continuous and discontinuous) were included as within-subject factors and age (healthy younger and older) was included as between-subjects factor. The significance level of the Box’s M test was set at p < 0.001, as this test is considered highly sensitive (Jiamwattanapong and Ingadapa 2021). When the assumption of sphericity was violated as evidenced by the Mauchly’s test, a Greenhouse–Geisser correction of degrees of freedom was used. Partial eta square (ηp2) was used as a measure of effect size where ≥ 0.2, ≥ 0.5 and ≥ 0.8 were considered as small, medium and large effects, respectively (Cohen 2013). For all analyses the level of significance was set at p < 0.05.