Individuals with HCP were identified through the Cerebral Palsy Research Registry (Hurley et al., 2011), local clinics, and parent support groups. In brief, participants were; (1) at least 6 years of age at time of testing; and (2) had a diagnosis of CP with unilateral motor impairment of the upper extremity; full criteria have been reported previously (Hill and Dewald, 2020). A cohort of age-matched controls without known neurological impairment (typical development; TD) was recruited for comparison. This experimental protocol was approved by the Institutional Review Board of Northwestern University. All participants were minors; parents provided informed consent prior to participation and participants provided assent.

Grip strength was recorded with the participant sitting with the shoulder at 0° of abduction and elbow at 90° of flexion. A ratio was calculated of the paretic to NP hand or ND to dominant hand (Jamar Hand Dynamometer, B&L Engineering, Tustin, CA, United States). For the cohort with HCP, a number of clinical assessments were used to classify function and impairment including: the Gross Motor Function Classification Scale (GMFCS) (Palisano et al., 1997; Rosenbaum et al., 2008), the Manual Abilities Classification Scale (MACS) (Eliasson et al., 2006), the Test of Arm Selective Control (TASC) (Krosschell et al., 2015; Sukal-Moulton et al., 2018), the Fugl-Meyer Assessment-Upper Extremity (FMA-UE) (Fugl-Meyer et al., 1975; Fasoli et al., 2009), and the ABILHAND-Kids (ABL-H) (Penta et al., 2001; Arnould et al., 2004).

All experimental conditions were tested one arm at a time with the first arm tested randomized for each participant. The active arm is referred to as the “Active_Isometric Arm” for isometric tasks and “Active_Reaching Arm” for dynamic tasks. The contralateral arm was cued to rest and is referred to as the “Rest Arm.” To determine the presence/absence of mirrored muscle activation and directionality when executing a single arm task, participants completed the following set of static and dynamic tasks in each arm.

All analysis was performed in MATLAB. Raw EMG signals for each muscle were full-wave rectified, baseline corrected, and manually inspected for artifacts. Artifacts were removed prior to smoothing of data with a moving average filter (250 ms window). Signals for each muscle were normalized to the maximum value recorded for that muscle across all maximal and submaximal tasks for that participant. This normalization strategy was selected to capture the highest capacity of the muscle, regardless of whether it was a voluntary effort or an involuntary co-activation.

Statistical analysis was completed using SPSS software (version 26, SPSS Inc., Chicago, IL, United States). A value of p < 0.05 was considered statistically significant for all tests. Q-Q plots were used to inspect data distributions and detect outliers. T-tests were used to determine whether there were differences in age or grip ratio between TD and HCP groups and Fisher’s Exact tests were used to determine if there were differences in sex and writing hand.

To evaluate whether the HCP and TD groups mirrored differently during isometric efforts, mirrored ratios were analyzed with a generalized linear mixed effects model for each arm with factors of group (TD or HCP) and muscle (IDL, BIC, BRD, TRI, WFL, and WEX), and a random factor of participant. To evaluate whether there was a directionality of mirroring within the TD or HCP group, mirrored ratios were analyzed with a generalized linear mixed effects model with factors test arm (Dominant/NP or ND/Paretic) and muscle (IDL, BIC, BRD, TRI, WFL, and WEX), and a random factor of participant. For each model, post hoc comparisons with Bonferroni corrections for multiple comparisons were made for significant main effects. To evaluate the effect of muscle activation in the Active_Reaching Arm on muscle activation in the Rest Arm, EMG summations were compared in a linear mixed effects model with fixed factor of group_arm (TD_Dominant, TD_NonDominant, HCP_NonParetic, or HCP_Paretic), random factor of participant, and a covariate of Active_Reaching Arm EMG. For each group_arm combination, the parameter estimates of the group_arm by Active_Reaching Arm EMG interaction were compared using t-tests to detect differences in the relationship (slope) between activation in the Active_Reaching Arm and the Rest Arm.

Normality of age and grip ratio distributions was confirmed with Shapiro–Wilk tests. There was not a significant difference in age (t₂₁ = −0.712, p = 0.484, g = −0.293) or sex (p = 1.00, phi-coefficient = −0.087) between the test groups. More individuals in the HCP group wrote with their left hand compared to the TD group (p = 0.009, phi-coefficient = 0.589). The HCP group had significantly lower grip strength ratios (t₁₇ = 5.139, p < 0.001, g = 1.765) indicating a weaker paretic hand.

Representative EMG data for one participant with HCP and one participant with TD during isometric maximum efforts can be seen in Figures 2, 3. Mirrored ratios were transformed using natural logs prior to statistical analysis to satisfy assumptions of normality. All statistical results presented are based on the transformed data while figures present untransformed data. When contracting the dominant/NP arm, there was a significant main effect of group [F₍₁, ₁₂₆₎ = 22.655, p < 0.001], muscle [F₍₅, ₁₂₆₎ = 4.370, p = 0.001], and the interaction of group by muscle [F₍₅, ₁₂₆₎ = 6.812, p < 0.001] with the HCP group demonstrating larger mirrored ratios for all muscles. Post hoc pairwise comparisons revealed a significant difference in mirrored ratios between TD and HCP for BIC (higher in HCP; t₁₂₆ = −5.619, p < 0.001, g = −2.188), WFL (higher in HCP, t₁₂₆ = −5.060, p < 0.001, g = −2.480), WEX (higher in HCP; t₁₂₆ = −2.187, p = 0.031, g = −0.905), and BRD (higher in HCP; t₁₂₆ = −4.830, p < 0.001, g = −1.576) but not TRI (t₁₂₆ = −1.270, p = 0.206) or IDL (t₁₂₆ = −1.231, p = 0.221). Additionally, there was a significant difference in mirrored ratio between BIC and WEX (higher in WEX; t₁₂₆ = −2.981, p < 0.041, g = −0.019), IDL and WFL (higher in IDL; t₁₂₆ = 3.062, p = 0.038, g = 0.031), WFL and WEX (higher in WEX; t₁₂₆ = −3.489, p = 0.010, g = −0.033), and BRD and WEX (higher in WEX; t₁₂₆ = −3.059, p = 0.038, g = −0.024). When contracting the ND/paretic arm, there was a significant main effect of muscle [F₍₅, ₁₂₆₎ = 6.843, p < 0.001] but not group [F₍₁, ₁₂₆₎ = 0.124, p = 0.725] or the interaction of group by muscle [F₍₅, ₁₂₆₎ = 1.904, p = 0.098]. Post hoc pairwise comparisons revealed a significant difference in mirrored ratio between BIC and IDL (higher in IDL; t₁₂₆ = −4.913, p < 0.001, g = −0.109), BIC and WEX (higher in WEX; t₁₂₆ = −4.312, p = 0.000, g = −0.099), IDL and BRD (higher in IDL; t₁₂₆ = 3.714, p = 0.004, g = 0.097), and BRD and WEX (higher in WEX; t₁₂₆ = −3.113, p = 0.028, g = −0.086). These results highlight a difference in mirroring between TD and HCP groups when completing the volitional task on the dominant/NP arm but not a difference between groups when completing the same task with the ND/paretic arm.

When comparing mirroring between arms for the TD group (Figure 4A), there was a significant main effect of muscle [F₍₅, ₉₆₎ = 7.340, p < 0.001]. Post hoc pairwise comparisons between muscles revealed a significant difference in mirrored ratios between BIC and TRI (higher in TRI; t₉₆ = −3.381, p = 0.011, g = −0.103), BIC and IDL (higher in IDL; t₉₆ = −4.926, p < 0.001, g = −0.123), BIC and WEX (higher in WEX; t₉₆ = −3.640, p = 0.005, g = −0.076), IDL and WFL (higher in IDL; t₉₆ = 3.705, p = 0.001, g = 0.136), and IDL and BRD (higher in IDL; t₉₆ = 3.705, p = 0.005, g = 0.091). Other tested effects were not significant {test arm [F₍₁, ₉₆₎ = 1.555, p = 0.215], test arm by muscle [F₍₅, ₉₆₎ = 0.965, p = 0.443]}. When comparing mirroring between arms for the HCP group, there was a significant main effect of test arm [F₍₁, ₁₅₆₎ = 83.543, p < 0.001], muscle [F₍₅, ₁₅₆₎ = 4.213, p = 0.001], and the interaction of test arm by muscle [F₍₅, ₁₂₆₎ = 3.793, p = 0.003] demonstrating a significant directionality of mirroring from the NP arm to the paretic arm. Post hoc pairwise comparisons revealed a significant difference in mirrored ratios between BIC and WEX (higher in WEX; t₁₅₆ = −3.097, p = 0.030, g = 0.008), TRI and WEX (higher in WEX; t₁₅₆ = −3.774, p = 0.003, g = 0.001), and BRD and WEX (higher in WEX; t₁₅₆ = −3.544, p = 0.007, g = 0.003). Additionally, the mirrored ratio was significantly higher when the NP arm was contracting all muscles except WEX (t₁₅₆ = 1.819, p = 0.071) according to post hoc pairwise comparisons (BIC: t₁₅₆ = 6.437, p < 0.001, g = 2.077; TRI: t₁₅₆ = 3.418, p = 0.001, g = 0.961; IDL: t₁₅₆ = 2.139, p = 0.034, g = 0.768; WFL: t₁₅₆ = 2.750, p = 0.007, g = 0.805; BRD: t₁₅₆ = 5.826, p < 0.001, g = 1.433) as shown in Figure 4B. In summary, the HCP group demonstrated more mirrored muscle activation when contracting the NP arm in all muscles except the WEX whereas the TD group demonstrated small mirrored ratios in each arm for all muscles.

Multiple combinations of muscle activations can be employed to maximize the torque production in one direction. Therefore, evaluating muscle activations during a dual task at multiple joints provided an avenue to explore mirrored activation in a specific task at submaximal effort levels. Total arm muscle activity increased with shoulder abduction load level for all arms tested (Table 2). The total normalized paretic arm EMG in those with HCP was higher than the ND arm of those with TD indicating greater relative activity in those with HCP.

There was a significant main effect of Active_ReachingArm EMG on Rest Arm EMG [F₍₁, ₂₆₆₎ = 21.541, p < 0.001] and the interaction of group by Active_ReachingArm EMG [F₍₁, ₂₅₈₎ = 24.317, p < 0.001]. This significant interaction highlights that the slopes of the co-activation relationship between the Active_Reaching Arm and the Rest Arm were different depending on which arm was reaching (Figure 5). In the HCP group at timepoint 2, the relationship between Active_Reaching Arm EMG and Rest Arm EMG has a steeper slope when reaching with the NP arm compared to reaching with the paretic arm (t₂₅₇ = 8.425, p < 0.001). In the TD group, there was not a significant difference in the Active_Reaching Arm EMG vs. Rest Arm EMG relationship between reaching with the dominant arm or ND arm at timepoint 2 (t₂₄₇ = 1.158, p = 0.248). This result highlights that the Rest Arm EMG increases more proportionally to the increase in Active_Reaching Arm EMG in the HCP group when reaching with the NP arm but there is not a difference between arms for the TD group. Additional slopes and statistical comparisons from timepoints 1 and 3 are shown in Tables 3, 4 and demonstrate similar results in the HCP group.

Electromyography (EMG) quantification offers higher resolution than is afforded by observational rating scales alone to provide insight into how the nervous system may be activating both arms simultaneously (Green, 1967; Carr et al., 1993; Carr, 1996). Furthermore, the use of the behavioral tasks of volitional contractions and movement goes beyond probing connectivity that is elicited by TMS and enables quantifying the capability of the nervous system to engage the available descending motor pathways. Green evaluated bilateral motor unit action potentials during single joint movements at the elbow wrist, and fingers (Green, 1967). The key finding of this study was that associated movements occurred only on the paretic side when the NP side was being activated (Green, 1967). Our cohort demonstrates a similar unidirectionality in mirrored activations. We found large mirrored ratios and an increase in Rest Arm EMG as a function of Active_Reaching Arm EMG when the NP arm, but not the paretic arm, was the active arm suggesting unidirectional mirroring. Specifically, there is activation in the paretic arm when the NP arm is leading the contraction or movement whereas there is minimal activation in the NP arm when the paretic arm is leading. This direction of mirroring suggests activation of ipsilaterally terminating CST projections from the non-lesioned hemisphere (Staudt et al., 2004). In the cohort studied by Staudt et al. (2004) mirroring that occurred in the paretic limb when activating the NP limb was only seen in those with ipsilateral connectivity as detected by TMS, linking this directionality of mirroring to ipsilateral fast conducting CST pathways. In contrast, mirroring in the NP arm when leading the movement with the paretic limb is more likely a non-specific overflow seen in individuals with maintained contralateral CST from the lesioned hemisphere (Staudt et al., 2004). The directionality of mirroring found in our results could indicate ipsilaterally originating pathways as the neural mechanism of the P_Mirror hypothesis. Additional investigation with TMS in our cohort would be pertinent to confirm the cortical origins of the mirrored activations we are measuring peripherally.

Some studies have shown the opposite directionality of mirror movements (Klingels et al., 2016; Riddell et al., 2019). Riddell et al. (2019) assessed the directionality of mirror movements by aggregating the results of clinical scores from finger tapping, sequencing, and grasp tasks and TMS measures of CST organization in a cohort of participants with HCP. A key finding was that the population with ipsilateral CST arrangement showed a stronger directionality in mirroring compared to those with non-ipsilateral CST arrangements. The directionality found for those with ipsilateral CST arrangement was primarily mirroring in the NP arm during focused fine motor tasks in the paretic hand which is opposite of the results in the present study as well as previous studies (Green, 1967; Staudt et al., 2004). There is still more to be understood regarding the extent to which the directionality of mirroring is inherent in the connectivity of the CST and how much is driven by a participant employed strategy to activate the NP hand to complete the task in the paretic hand. Our methodology without feedback on muscle activation levels may be advantageous in reducing the use of a strategy to achieve a specific outcome.

Characterization of the whole limb as well as exploring the directionality of between limb muscle co-activation enables a more complete interpretation of the neural mechanisms implicated in mirror movements after early brain injury. A key focus of previous mirror movement studies has been at the hand demonstrating stronger mirroring primarily from the paretic hand to the NP hand (Woods and Teuber, 1978; Kuhtz-Buschbeck et al., 2000; Riddell et al., 2019), the opposite directionality seen in our dataset of the whole limb. A study that included the elbow found directionality primarily from the NP arm to the paretic arm (Green, 1967) which is similar to our results for the BIC, BRD, and TRI during the isometric maximum contractions. Sukal-Moulton et al. (2013) systematically investigated the effect of maximal and submaximal isometric and isokinetic efforts of the NP elbow on involuntary isometric torque production by the paretic elbow. They found that mirroring was more prevalent in individuals with lesions sustained under 6 months of age compared to those with later lesion timings and that this earlier injury timing group was less able to suppress elbow flexion mirroring into the paretic arm (Sukal-Moulton et al., 2013). While the focus of our study was on the spontaneous activity of the Rest Arm without instructions to suppress any activation, we demonstrate increased paretic arm EMG activity with load level during reaching with the NP arm which is similar to the increased paretic arm torque with NP arm effort level found previously (Sukal-Moulton et al., 2013). Interestingly, we found that while there was clear directionality in all of the proximal muscles tested, the WEX showed similar magnitude of mirroring irrespective of which arm was the Active_Isometric Arm. The similarity in mirroring between hands for this distal muscle highlights a potential difference in descending control to the hand. The necessity of the hand for fine motor tasks underscores the need for further investigation into the independent control of the hand.

The presence of mirrored muscle activation during NP arm contraction found in the HCP group of our study would suggest the hypothesized of maintenance of simultaneous control of both arms from one hemisphere after early lesions to the brain. Possible mechanisms of descending pathway reorganization after a unilateral lesion have been explored in previous studies. Staudt (2010) suggests that the interruption of the typical pruning process of the CST after a unilateral lesion enables the non-lesioned hemisphere to maintain direct CST projections to motor neuron pools innervating both arms. This mechanism could coincide with the branched projection hypothesis put forth by Carr (1996) and could be the underlying mechanism in our key findings. Hawe et al. (2013) suggest instead that the integrity and volume of the white matter in the corpus callosum may be implicated in the inability to suppress mirror movement in individuals with early unilateral lesions. In a comparative literature review evaluating the ipsilateral CST projection from the non-lesioned hemisphere hypothesis against the insufficient interhemispheric inhibition hypothesis, Kuo et al. (2018) assert that the presence of mirror movements is most indicative of an ipsilateral CST reorganization. Based on the directionality of mirroring that we measured in our cohort, we also postulate that the presence of these mirrored activations is likely due to maintained ipsilaterally terminating CST projections from the non-lesioned hemisphere. In contrast to maintained ipsilateral CST projections for upper limb control, the corticoreticulospinal tract is implicated in the abnormal flexion synergies of the upper extremity in adults after stroke (Schwerin et al., 2008; McPherson et al., 2018). Pronounced flexion synergies driven by shoulder abduction during reaching are seen in adult-onset hemiplegia, whereas our previous work in pediatric-onset hemiplegia largely showed an absence of this synergy pattern in those with early lesions while completing a ballistic reaching task (Hill and Dewald, 2020). This provides evidence against the activation of brainstem mediated pathways during volitional movements and would support the hypothesis that mirrored activation could be driven by ipsilateral CST connections. Interestingly, our results show greater total arm activation in the paretic arm of those with HCP compared to the ND arm of those with TD (Table 2). Sukal-Moulton et al. (2014) detected an overflow in within limb activation during distal joint MVTs in those with earlier injuries. Specifically, they demonstrated significant coupling between elbow flexion and wrist/finger flexion compared to those without neurological impairment pointing out within limb coupling that is distinct from the proximally driven flexion synergy pattern seen in adults post stroke.

We collected data for only six muscles due to setup time and surface area available on participants’ arms, however, this snapshot did include at least one muscle crossing each joint of the upper limb. While we cued participants to relax the Rest Arm during active arm efforts, we did not provide feedback that would encourage suppression of mirrored activations as that was out of the scope of this particular study. In this study we have recorded solely muscle activation from the periphery and have not taken any measurements directly from the cortex. In order to confirm the cortical component at the origin of mirrored muscle activations, neuroimaging such as TMS, fMRI, or calculating EEG/EMG wavelet coherence (Xi et al., 2020) would be required in combination with our peripheral measures.

Tasks of daily life require the ability to independently control joints within a limb as well as the coordinated control of the hands in concert. Bimanual tasks often require one arm/hand to stabilize and the other to engage in manipulation (Johansson et al., 2006; Wang and Sainburg, 2007). Individuals with HCP frequently have difficulty with unimanual tasks involving the paretic hand and bimanual tasks involving both hands (Holmefur et al., 2010). The presence of an alternative wiring pattern of the CST may be responsible for the persistence of mirror movements seen in this population, and is unlikely to be reduced with the use of unilateral training. The use of EMG could offer an additional measurement of neural output that could be incorporated into interventions purposed to decrease mirrored movements and improve function. Unlike other neuroimaging approaches which are more complex and possibly contraindicated for this population, EMG recording can be used in a clinical setting alongside of other clinical assessments to investigate with finer resolution where the issues of co-activation are present in a similar manner as the use of the Windmill-task (Zielinski et al., 2018). EMG also has the potential to facilitate more rapid diagnosis of CST connectivity that could help guide the development of interventions though additional investigation would be needed for validation with the gold standard of TMS.