Gripping objects during everyday manual tasks requires the coordination of muscle contractions and muscle relaxations. The vast majority of studies have focused on muscle contractions. Although previous work has examined the motor cortex during muscle relaxation, the role of brain areas beyond motor cortex remains to be elucidated. The present study used functional magnetic resonance imaging to directly compare slow and precisely controlled force generation and force relaxation in humans. Contralateral primary motor cortex and bilateral caudate nucleus had greater activity during force generation compared with force relaxation. Conversely, right dorsolateral prefrontal cortex (DLPFC) had greater activity while relaxing force compared with generating force. Also, anterior cingulate cortex had greater deactivation while relaxing force compared with generating force. These findings were further strengthened by the fact that force output parameters such as the amplitude, rate, duration, variability, and error did not affect the brain imaging findings. These results demonstrate that the neural mechanisms underlying slow and precisely controlled force relaxation differ across prefrontal–striatal and motor cortical–striatal circuits. Moreover, this study demonstrates that the DLPFC is not only involved in slow and precisely controlled force generation, but has greater involvement in regulating slow and precisely controlled muscle relaxation.
Gripping is an important function of primate and rodent behavior (Iwaniuk and Whishaw 2000; Spinks et al. 2008). Muscles must be contracted and relaxed with precise amplitude, rate, and duration to smoothly execute every day tasks, such as eating and grooming. Yet, much of the literature on motor control has focused on muscle contraction and not muscle relaxation. In humans, muscle relaxation is clinically important because it can be associated with cortical and subcortical dysfunction. Early work in nonhuman primates demonstrates that inactivation of one of the major output structures of the basal ganglia, the globus pallidus, leads to difficulty in turning off muscle activation (Mink and Thach 1991). Studies of Parkinson's disease have shown that patients experience difficulty relaxing isometric contractions about the elbow (Corcos et al. 1996) and during precision grip (Wing 1988). Furthermore, other work has shown that elbow torque relaxation time is related to the Unified Parkinson's Disease Rating Scale motor score of Parkinson's disease patients performing an isometric force task (Robichaud et al. 2005). Despite the relationship between basal ganglia dysfunction and difficulty in muscle relaxation, the role of cortico-basal ganglia circuits in muscle relaxation remains undefined.
The literature is clear that muscles do not relax through a simple cessation of projection neuron activity in primary motor cortex (M1), but through an active process that involves complex neural circuits. A study using transcranial magnetic stimulation over human M1 shows that intracortical inhibitory circuits actively suppress corticospinal excitation during muscle relaxation of an isometric pinch grip (Buccolieri et al. 2004). Previous work using electroencephalography (EEG) (Pope et al. 2007) and functional magnetic resonance imaging (fMRI) (Toma et al. 1999) found that neural activation in M1 was greater for muscle contraction than for muscle relaxation during an isometric pinch grip task. This is consistent with studies using single motor unit recording techniques which have shown that the firing rates of motor units at the same force level are lower during slow relaxation than during contraction (De Luca et al. 1982; Denier van der Gon et al. 1985).
Other work in the brain suggests that areas outside M1, such as supplementary motor area (SMA) and presupplementary motor area (pre-SMA), may also play an active role in muscle relaxation. For instance, subdural electrical stimulation anterior to the primary motor face area in M1 and anterior to the supplementary motor face area in SMA of humans can cause a tonic muscle contraction to relax following stimulation (Luders et al. 1985, 1995). This phenomenon was termed a negative motor area because the stimulation caused the tonic contraction to relax rather than contract. Also, studies using subdural recordings and EEG in humans observed Bereitschaftpotentials in SMA (Terada et al. 1995, 1999; Yazawa et al. 1998) and pre-SMA (Yazawa et al. 1998) associated with muscle relaxation. Previous work using fMRI directly compared neural activation in SMA and pre-SMA and found that blood oxygenation level–dependent (BOLD) activation volume was greater for muscle relaxation than muscle contraction (Toma et al. 1999). In summary, previous work has shown that neural activation in M1 and SMA is different for muscle relaxation compared with muscle contraction. It remains unclear how changes in BOLD activation in other brain areas differentially mediate muscle contraction and muscle relaxation during pinch grip.
The current study used fMRI in humans to directly compare the neural processes that underlie precisely controlled force generation and precisely controlled force relaxation during pinch grip. The study had 2 main objectives. The first objective was to determine if M1, SMA, and pre-SMA have greater neural activation while slowly and precisely generating force compared with relaxing force. The second objective investigates whether neural populations in the prefrontal cortex and basal ganglia have differences in BOLD activation between the 2 grip force tasks. We measure the amplitude, rate, duration, variability, and error of grip force to ensure that differences observed between force generation and force relaxation are not due to these other parameters which are known to affect BOLD signals (Spraker et al. 2007; Prodoehl, Yu, Wasson, et al. 2008).
Twelve healthy right-handed subjects with corrected or normal vision participated in the study (5 males and 7 females, ages 19–34). Prior to the experiment, each subject provided informed consent to all procedures, which were approved by the local Institutional Review Board and were in accord with the Declaration of Helsinki.
Force Data Acquisition
Subjects produced isometric force using their right hand against a custom grip device (see Fig. 1, Vaillancourt et al. 2003). The grip device was constructed from nonmetallic material (polycarbonate) to allow for its use inside the magnetic resonance environment. When prompted, subjects pinched the apparatus with their right middle finger and thumb to produce force that was transmitted to a connected water-filled plastic tube (35 ft.). The grip device was fully supported such that subjects did not support the weight of the force device. A tube ran into an Entran pressure transducer (EPX-N13-250P) located outside the fMRI environment. The pressure signal was amplified by a custom built amplifier. Data from the amplifier were digitized at 100 Hz by a PCMCI National Instruments A/D board. The output from the pressure transducer was displayed to the subject and was updated at each sampling interval. The feedback was projected via the parallax biofeedback system (Thulborn 1999) and the visual display was similar to previous experiments (Vaillancourt et al. 2003). A mirror located inside the MR environment displayed visual feedback onto a video screen located 35 cm from the subject's eyes. The force output was displayed on the screen at a resolution of 640 × 480 pixels and a refresh rate of 60 Hz.
MRI Data Acquisition
Magnetic resonance images were collected using a volume head coil inside a 3 Tesla MR Scanner (GE Healthcare, Waukesha, WI, 3T94 Excite 2.0). The subjects lay supine in the scanner while performing the task. The subject's head was stabilized using adjustable padding and then fitted with the projector-visor system for displaying visual feedback. The functional images were obtained using a T2*-sensitive, single shot, gradient-echo echo-planar pulse sequence (echo-time 25 ms; repeat-time 2500 ms; flip angle 90°; field of view 200 mm2; imaging matrix 64 × 64; 42 axial slices at 3-mm thickness; 0-mm gap between slices). The high-resolution anatomical scans were obtained using a T1-weighted fast spoiled gradient echo pulse sequence (echo-time 1.98 ms; repeat-time 9 ms; flip angle 25°; field of view 240 mm2; imaging matrix 256 × 256; 120 axial slices at 1.5-mm thickness; 0-mm gap between slices).
The task sequence for the fMRI experiment used a blocked-design paradigm completed during a single functional scan. Before scanning, each subject participated in a 1-h training session outside the scanner in order to minimize motor learning effects when inside the scanner. Once in the fMRI environment, the individual's maximum voluntary contraction (MVC) was calculated. The subjects were asked to sustain a contraction of maximum force for 3 consecutive 5-s trials. Each trial was separated by a 60-s period of rest. The MVC was calculated as the average force during the sustained maximum force contraction.
The paradigm required subjects to use a pinch grip (thumb and middle finger) with their right hand to produce force to a target magnitude (Vaillancourt et al. 2003). The target always represented 15% of the individual subject's MVC and was displayed on the screen as a fixed horizontal bar. The cursor was displayed on the screen as a moveable horizontal white bar. The vertical position of the cursor with respect to the target was directly related to the force produced by the subject.
Subjects viewed the target and cursor bar on the screen during the entire functional scan. The task consisted of alternating 30-s rest and 78-s task blocks with rest blocks positioned at the beginning and end of the sequence (Fig. 1A). During the rest blocks, the subjects fixated on a stationary red target and stationary white cursor but did not produce force. During the task blocks, the subjects completed the following sequence: a 30-s precisely controlled force generation condition, an 18-s rest condition, and a 30-s precisely controlled force relaxation condition (Fig. 1B). During the 30-s force generation condition, subjects repeated a 5 s precisely controlled force generation sequence 5 times with 1 s of rest between each repetition (Fig. 1B,C, black). Similarly, during the 30-s force relaxation condition, subjects repeated a 5-s precisely controlled force relaxation sequence 5 times with 1 s of rest between each repetition (Fig. 1B,C, gray). During the 18-s rest block, subjects behaved the same as they did during 30-s rest blocks (see above). We chose 18 s to allow the hemodynamic response to return to baseline between force generation and force relaxation conditions.
Actual force traces from a single subject for a single repetition of the precisely controlled force generation (black) and precisely controlled force relaxation (gray) sequences are depicted in Figure 1C. Both sequences were 5 s in duration, and each contained a 4-s ramp period and a 1-s hold period (Fig. 1C). The force generation sequence began as the target bar turned green and remained green for 4 s, indicating the ramp period. Subjects were trained to produce force with a consistent rate and duration so the cursor hit the target at the end of the ramp period. This period ended when the target bar turned yellow and remained yellow for 1 s, indicating the hold period. Subjects were trained to maintain isometric force at 15% MVC during the hold period. Each hold period and force generation sequence ended when the target turned red for 1 s, indicating rest. The ramp and hold periods in the precisely controlled force generation sequence are used in the precisely controlled force relaxation sequence, but the order is inverted (Fig. 1C). The force relaxation sequence began with the hold period when the target turned yellow and remained yellow for 1 s. Subjects were trained to produce force as quickly and accurately as possible to the target and maintain a contraction of 15% MVC for the duration of this hold period. Then, the target turned green and remained green for 4 s indicating the ramp period. In the force relaxation condition, subjects were instructed to decrease force with consistent rate and duration from 15% MVC so that the cursor hit the resting force level at the end of the 4-s ramp period. The sequence ended when the target bar turned red for 1 s of rest. In total, subjects completed 5 precisely controlled force generation and 5 precisely controlled force relaxation sequences per 78-s task block. This amounted to 20 sequences of each type across the functional scan.
Force Data Analysis
After the force output was collected, we performed a visual inspection of the data and marked 4 specific time points for each precisely controlled force generation and precisely controlled force relaxation sequence within each scan (Fig. 1C). The first point was marked at the onset of force. The second and third points were marked at the beginning and end of the hold period, respectively (Fig. 1C). The fourth point was marked at the offset of force. We calculated 6 main variables for the force data analysis. First, mean force amplitude was calculated as the mean force output between points 2 and 3 for both force sequences (Fig. 1C). Second, the mean duration of the ramp plus hold period for each sequence was calculated as the time difference between points 4 and 1 for both force sequences (Fig. 1C). Third, the mean duration of the ramp period was calculated as the time difference between points 2 and 1 in the force generation sequence and points 4 and 3 in the force relaxation sequence. Fourth, the rate of change of force (rate of force) was averaged across the ramp period after calculating the first derivative of force between points 1 and 2 in force generation sequences (Fig. 1C). The mean rate of force during the ramp period was calculated similarly for force relaxation sequences using the force output between points 3 and 4 (Fig. 1C). The absolute value of the rate of change of force was used so force generation and force relaxation could be directly compared. Fifth, the standard deviation of force during the ramp period of force generation and force relaxation sequences was analyzed. The linear trend in the force signal was removed and standard deviation of the force was calculated between points 1 and 2 for force generation sequences and points 3 and 4 for force relaxation sequences (Fig. 1C). Sixth, the absolute error of force during the ramp period of force generation and force relaxation was analyzed. Absolute error was calculated as the average absolute deviation of the actual force trace from a perfect ramp performance (i.e., 0% MVC to 15% MVC with constant rate of force in exactly 4 s).
Calculations were first carried out for each of the 5 individual sequences within the 4 task blocks to give 20 values for each dependent measure per subject per condition (i.e., force generation or force relaxation). These 20 values were averaged to give 6 mean dependent measures per condition per subject and then group means were calculated. All calculations were made using custom algorithms in MATLAB. Differences between the force generation and force relaxation conditions were analyzed using separate paired t-tests for each dependent measure (Statistica, Statsoft, Tulsa, OK, v6.1).
fMRI Data Analysis
Analysis of Functional NeuroImages (AFNI), the public domain software (http://afni.nimh.nih.gov/afni/), was used to process and analyze the fMRI data sets. First, we will describe methods for head motion analysis. Then, we will describe the methods used for 2 group analyses. The voxel-wise analysis was our primary analysis and the region of interest (ROI) analysis was performed in areas identified from the voxel-wise analysis to further confirm the finding.
Head Motion Analysis and Motion Correction
After importing the acquired data, motion detection, and correction functions were applied to each functional time series. This resulted in 3 time series for displacement (x, y, z) and 3 time series for rotation (roll, pitch, yaw). Head motion was quantified by calculating the absolute value of displacement and rotation at each volume and then averaging across each time series. One subject was removed from further analysis because mean head displacement exceeded the requirement of less than 1 mm in any direction. The mean head displacement across all included subjects (n = 11) was 0.29, 0.08, and 0.09 mm in the x, y, and z directions, respectively. The mean head rotation across subjects was 0.13°, 0.19°, and 0.17° for roll, pitch, and yaw, respectively.
A voxel-wise analysis was first performed on the fMRI data. Motion-corrected individual data sets were normalized by dividing the instantaneous signal in each voxel at each point in the time series by the mean signal in that voxel across each scan. After this, we applied a Gaussian filter to the resultant data sets (full-width half-maximum at 3 mm). Then, the time-series data were regressed to a simulated hemodynamic response function for the task sequence (3Ddeconvolve, AFNI). The dependent variable at this level of analysis was the estimated β-coefficient of the regressed time series and its associated t-statistic. Before group analysis, each subject's anatomical data set was manually transformed to Talairach space using AFNI. Then, each subject's individual functional data sets were transformed to Talairach space using the normalized anatomical data set as a template.
The output data were analyzed using a mixed-effect 2-way ANOVA with the sequence type (i.e., precisely controlled force generation or precisely controlled force relaxation) as a fixed factor and the subject as a random factor. The ANOVA yielded 3 data sets per voxel: 1) the estimated group mean β-value for precisely controlled force generation minus rest, 2) the estimated group mean β-value for precisely controlled force relaxation minus rest, and 3) the estimated difference in group means (force generation − force relaxation). All 3 data sets were corrected for multiple comparisons using a Monte Carlo Simulation model (AFNI alphasim, http://afni.nimh.nih.gov/afni/doc/manual/AlphaSim/). The data sets were thresholded to remove all voxels with t < 3 with an activation cluster minimum of 205 μL (P < 0.05, corrected). The corrected t-statistic associated within each voxel is displayed in the final group maps. Voxels below the threshold were considered nonsignificant.
Statistical Analysis of Regions of Interest
An ROI analysis of BOLD percent signal change was used to confirm the findings of the voxel-wise analysis. Percent signal change was acquired by first calculating the mean signal within each voxel for rest, precisely controlled force generation, and precisely controlled force relaxation blocks across each individual motion-corrected functional time series. The mean percent signal change within each voxel was calculated using the following equation:
The ROIs used for statistical analysis were the areas that had a significant difference in BOLD signal between the force generation and force relaxation tasks. These areas were detected by the voxel-wise ANOVA (see above) and then were identified using the Basal Ganglia Human Area Template (Prodoehl, Yu, Little, et al. 2008) and the Human Motor Area Template (Mayka et al. 2006). Group mean percent signal change within each ROI was analyzed using a paired t-test between the force generation and force relaxation conditions. Because the main effect of duration of the ramp plus hold period was significant, we also completed ANCOVAs for each ROI with duration of force as a continuous predictor. All tests were evaluated as significant if the P value was less than 0.05.
The study also quantified activation volume in SMA and pre-SMA in order to investigate previous observations in the literature (Toma et al., 1999). Activation volume was calculated using an ROI approach with a priori drawn ROIs in SMA and pre-SMA. Details on these ROIs have been published in the literature (Mayka et al. 2006). Left and right ROIs for SMA and pre-SMA were combined into a single, bilateral ROI for each region. The volume of activation within the SMA and pre-SMA was first quantified by counting the number of voxels within each ROI that had t > 3 on each individual activation map for the precisely controlled force generation and precisely controlled force relaxation conditions. The difference in activation volume between conditions was analyzed using a separate paired t-test for SMA and pre-SMA. The test was considered significant if the P value was less than 0.05.
The results from the force data collected during the task will first be presented. Then, the fMRI results will be presented in 2 parts. First, regions that had increased activation during the precisely controlled force generation condition compared with the precisely controlled force relaxation condition will be shown. Next, regions that had a greater change in activation during force relaxation than during force generation will be reported.
Force Output Results
The average force, duration of the ramp plus hold period, duration of the ramp period, rate of change of force during the ramp period, standard deviation of force during ramp period, and force error during ramp period, for each subject (S1-S11) and the group mean, are depicted in Figure 2A-F. The group mean force amplitude during the precisely controlled force generation and precisely controlled force relaxation condition was 15.079% MVC and 15.222% MVC (Fig. 2A). A dependent-samples t-test revealed that the mean force was not significantly different between the 2 conditions (t = −0.57, df = 10, P = 0.58). The average duration of the ramp period plus the hold period for the force generation condition was 5.15 s, whereas the average duration for the force relaxation condition was 5.01 s (Fig. 2B). A dependent-samples t-test showed that the small difference in duration between the 2 conditions was significant (t = 3.61, df = 10, P < 0.01). The average duration of the ramp period for the force generation condition and force relaxation condition was 4.17 and 4.24 s, respectively (Fig. 2C). A dependent-samples t-test confirms that the difference between conditions was not significant (t = −0.31, df = 10, P = 0.76). The mean rate of force for the force generation condition was 3.778% MVC/s and mean rate of force for the force relaxation condition was 3.745% MVC/s (Fig. 2D). A dependent-samples t-test revealed that the rate of force was not statistically different across the 2 conditions (t = 0.25, df = 10, P = 0.81). We also calculated the standard deviation of force during the ramp period of the force generation and force relaxation condition, which was 1.044 and 1.016% MVC, respectively (Fig. 2E). A dependent-samples t-test confirmed that there was no difference in standard deviation of force between the 2 conditions (t = 0.9387, df = 10, P = 0.37). Figure 2F shows that the absolute error was not different between force generation and force relaxation (t = 1.19, df = 10, P = 0.26).
Force Generation Greater than Force Relaxation
Figure 3A–C depicts the fMRI results for left M1. Figure 3A demonstrates that there was robust activation in M1 for force generation compared with rest and for force relaxation compared with rest. We also compared force generation and force relaxation directly and found an area of left M1 (center of mass: x = −44.5, y = −25.0, z = 48.9) that had significantly greater activation for force generation than force relaxation (Fig. 3B, top). Many of these voxels were active during both conditions, but a greater number were active during force generation than force relaxation (Fig. 3B, bottom). Figure 3C demonstrates that there is greater group mean BOLD percent signal change within the left M1 ROI for the force generation (red) than the force relaxation (blue) condition. Statistical analysis confirms that the difference in percent signal change in the left M1 ROI was significant (t = 2.68, df = 10, P < 0.05). Additionally, an ANCOVA indicated that this result remained significant when duration of the ramp plus hold period was included as a continuous predictor (F1,9 = 9.78; P = 0.01). The interaction between task and ramp plus hold duration was not significant (F1,9 = 1.26; P = 0.30).
Figure 3D–F depicts the fMRI data for bilateral caudate nucleus. The voxel-wise ANOVA demonstrates that there was robust activation in caudate bilaterally for precisely controlled force generation compared with rest (Fig. 3D). However, there are only a few significantly activated voxels on the border of right caudate and putamen for precisely controlled force relaxation compared with rest (Fig. 3D). We identified several voxels in the head of the caudate bilaterally (center of mass [left]: x = −14.2, y = 14.1, z = 10.1; center of mass [right]: x = 14.0, y = 13.7, z = 8.8) that were significantly greater during the force generation condition compared with the force relaxation condition (Fig. 3E, top). Moreover, we found that all of these voxels are active only during the force generation condition (Fig. 3E, bottom). The ROI analysis shows that group mean BOLD signal change for the force generation condition (red) is greater than for the force relaxation condition (blue) in both left and right caudate (Fig. 3F). Separate dependent-samples t-tests verified that the difference in BOLD was significant in both left (t = 2.32, df = 10, P < 0.05) and right caudate (t = 2.33, df = 10, P < 0.05). In addition, separate ANCOVAs indicated that the difference remained significant in left caudate (F1,9 = 7.89; P < 0.05) and right caudate (F1,9 = 10.98; P < 0.01) with ramp plus hold duration as a continuous predictor. The interaction between task and ramp plus hold duration was not significant for left caudate (F1,9 = 1.34; P = 0.28) or right caudate (F1,9 = 1.38; P = 0.27).
Force Relaxation Greater than Force Generation
The voxel-wise analysis of fMRI data revealed that right dorsolateral prefrontal cortex (DLPFC) and bilateral anterior cingulate cortex (ACC) had a greater change in activation during the precisely controlled force relaxation condition than during the precisely controlled force generation condition. Figure 4A–C depicts the fMRI results for right DLPFC. The voxel-wise ANOVA demonstrates that there was robust activation in right DLPFC for the force relaxation condition compared with rest, and that this activation was greater than the force generation condition compared with rest (Fig. 4A). Figure 4B (top) displays an area in right DLPFC (center of mass: x = 47.0, y = 31.8, z = 31.9) where BOLD activation is significantly greater for the force relaxation condition than the force generation condition. All of these voxels were active only during the force relaxation condition (Fig. 4B, bottom). Figure 4C depicts group mean BOLD percent signal change within the right DLPFC ROI for the force generation (red) and force relaxation (blue) conditions. A dependent-samples t-test verified that group mean BOLD percent signal change for the force relaxation condition was significantly greater than for the force generation condition (t = −2.41, df = 10, P < 0.05). This difference was significant in an ANCOVA with ramp plus hold duration as a continuous predictor (F1,9 = 10.56; P = 0.01). The interaction between task and ramp plus hold duration was not significant (F1,9 = 1.53, P = 0.24).
Figure 4D–F depicts the fMRI data for bilateral ACC. The voxel-wise ANOVA demonstrates that there was robust deactivation in bilateral ACC for precisely controlled force relaxation compared with rest, but minimal deactivation in bilateral ACC for precisely controlled force generation compared with rest (Fig. 4D). Figure 4D only depicts voxels with BOLD deactivation, which represents a BOLD signal that is significantly lower for the task condition than the rest condition. Figure 4E (top) shows that when the 2 conditions were compared directly, we found an area in bilateral ACC (center of mass: x = 1.9, y = 34.8, z = 21.6) that had greater deactivation during force relaxation than force generation. We found that many of these voxels were deactivated only during the force relaxation condition and a small number of voxels were deactivated during both conditions (Fig. 4E, bottom). Figure 4F shows that group mean BOLD percent signal change within the bilateral ACC ROI is more negative for the force relaxation (blue) than for the force generation (red) condition. A dependent-samples t-test verified that there was significantly greater deactivation during force relaxation compared with force generation in ACC bilaterally (t = 2.55, df = 10, P < 0.05). This effect was significant in an ANCOVA with ramp plus hold duration as a continuous predictor (F1,9 = 10.97, P < 0.01). The interaction between task and ramp plus hold duration was not significant (F1,9 = 1.44, P = 0.26)
In the previous analyses, the ROI approach was only used in brain regions in which the voxel-wise analysis did identify a difference in BOLD between precisely controlled force generation and precisely controlled force relaxation. However, a previous fMRI study found differences in activation volume using a priori ROIs in SMA and pre-SMA for muscle contraction compared with muscle relaxation (Toma et al. 1999). Therefore, the current study also quantified group mean activation volume in SMA and pre-SMA using a single a priori ROI for each bilateral region. Activation volume within the SMA ROI was 11.40 and 11.76 mL for the precisely controlled force generation and precisely controlled force relaxation conditions, respectively. The volume of activation in bilateral SMA was not significant between the 2 tasks (t = −0.33, df = 10, P = 0.75). Activation volume within the bilateral pre-SMA ROI was 7.35 mL for the force generation condition and 7.42 mL for the force relaxation condition. Statistical analysis confirms that the difference in activation volume in bilateral pre-SMA was not significant (t = −0.19, df = 10, P = 0.85).
The aim of the current study was to directly compare the neural processes that mediate precisely controlled force generation and force relaxation in humans. Four brain areas had a significant difference in BOLD activation between the 2 tasks. First, there was greater BOLD activation in left M1 while precisely generating grip force than while precisely relaxing grip force. Second, there was greater BOLD activation in bilateral caudate while precisely generating grip force than relaxing grip force. Third, there was greater BOLD activation in right DLPFC while precisely relaxing grip force than while generating grip force. Fourth, there was greater BOLD deactivation in bilateral ACC while precisely relaxing grip force than while generating grip force. The findings in motor cortex are considered first and the second section discusses the findings beyond motor cortex.
Task Differences in Motor Cortex
Left M1 had greater activation during precisely controlled force generation than during force relaxation using both voxel-wise and statistical ROI analyses. The current finding in M1 extends previous work that has used EEG to investigate differences in M1 activation associated with muscle activation, holding a level of force, and muscle relaxation during isometric tasks (Rothwell et al. 1998; Pope et al. 2007) to tasks that require precisely controlled force relaxation. A consistent finding in 2 of these EEG studies is that the readiness potential in M1 is greater for muscle activation than for muscle relaxation during pinch grip. These findings are different from EEG studies that found the readiness potential in M1 is similar for muscle activation and muscle relaxation during movement (Terada et al. 1995; Pope et al. 2007). Pope et al. (2007) directly compared muscle activation and relaxation using EEG for both a movement and an isometric task and verified that the readiness potential in M1 is greater for muscle contraction than relaxation during isometric tasks, but is similar for movement tasks.
Using fMRI, Toma et al. (1999) directly compared quick muscle activation and relaxation during both movement and isometric pinch grip tasks. They found greater BOLD signal change in M1 for muscle activation compared with muscle relaxation in an isometric pinch grip task, but similar BOLD signal change in M1 for muscle activation and muscle relaxation during a wrist movement task. The isometric task used by Toma et al. (1999) required subjects to sustain a moderate level of pinch grip force and then either quickly increase force amplitude to a maximum level for contraction trials or quickly decrease force amplitude to rest for relaxation trials. Thus, there was greater force amplitude exerted during contraction trials than relaxation trials in the isometric task. The current study carefully controlled the force amplitude and other force output parameters while comparing precisely controlled force generation and force relaxation during precision grip. This extends the work of Toma et al. (1999) by demonstrating that the BOLD signal in M1 is greater when precisely generating force than precisely relaxing force when the force parameters (i.e., amplitude, rate, variability, duration, and error) are very similar between conditions.
Rothwell et al. (1998) suggest that reduced neural activation in M1 for muscle relaxation compared with muscle activation during isometric tasks may be facilitated by the withdrawal of excitatory drive from premotor cortex to M1. However, Toma et al. (1999) did not report a difference in activation in premotor cortex when comparing muscle activation and relaxation during the isometric task. Their study included 5 axial slices focused on M1 and SMA and did not include the entire premotor cortex, which extends from z = 74 to z = −2 in Talairach space (Mayka et al. 2006). The current study included the entire premotor cortex and found no difference in BOLD activation while precisely generating and relaxing grip force.
An alternative hypothesis, supported by a transcranial magnetic stimulation (TMS) study (Buccolieri et al. 2004), is that the reduction in M1 activation associated with muscle relaxation during isometric tasks may be due to an increase in excitation of M1 intracortical inhibitory circuits. Further support of this hypothesis comes from studies of slow muscle relaxation in motor unit recordings and those using TMS. First, studies using single motor unit recording techniques have shown that the firing rates of motor units at the same force level are lower during slow relaxation than during contraction in the elbow flexor (Denier van der Gon et al. 1985), the wrist extensor (Romaiguere et al. 1993), the plantar flexor and extensor (Gorassini et al. 2002), and the first dorsal interosseus muscle (De Luca et al. 1982). Second, a TMS study demonstrated hysteresis in corticospinal excitability in that the motor evoked potential was greater for contraction than relaxation at the same force level in the biceps brachii and brachioradialis muscles in humans (Kimura et al. 2003). In summary, our study further supports the idea of Buccolieri et al. (2004) that precisely controlled force relaxation during isometric pinch grip tasks is facilitated by an increase in the excitability of intracortical inhibitory circuits in M1, and not a decrease in excitatory drive from premotor cortex to M1.
An fMRI study using a voxel-wise approach identified activation foci in SMA associated with a quick isometric muscle contraction and relaxation (Toma et al. 1999). This study found that the average percent increase in BOLD signal intensity was similar for contraction and relaxation within the local maxima of SMA (i.e., voxel with maximum Z-score in the SMA focus of activation). These authors also analyzed activation volume within a priori drawn ROIs for SMA and pre-SMA and found more extensive activation for voluntary relaxation than for voluntary contraction. Similar to the study by Toma et al. (1999), the current voxel-wise analysis did not identify a difference in BOLD activation in SMA or pre-SMA for precisely controlled force generation compared with force relaxation. The current study also quantified volume of activation within a priori ROIs for SMA and pre-SMA and found no difference between the 2 tasks, which is in contrast to the findings of Toma. The task used by Toma et al. (1999) required subjects to make quick voluntary contractions and relaxations which were short in duration. The task used in the current study required subjects to make slow and precisely controlled contractions and relaxations that were long in duration. Therefore, the activation volume in SMA and pre-SMA may be greater for fast muscle relaxation than fast muscle contraction in tasks of short duration, whereas activation volume in SMA and pre-SMA may be similar for precisely controlled muscle relaxation compared with precisely controlled muscle contraction in tasks of long duration. Previous work has shown that increasing the duration of grip force causes a robust increase in activation volume in SMA (Vaillancourt et al. 2004). Further studies are required to determine the extent to which the SMA plays a differential role in controlling fast versus slow voluntary muscle relaxations.
In summary, the current study found greater BOLD percent signal change in M1 for precisely controlled force generation compared with precisely controlled force relaxation. Also, there was no difference in BOLD percent signal change in premotor cortex, SMA and pre-SMA between the 2 tasks. These findings support the hypothesis that precisely controlled relaxation during pinch grip force is mediated by mechanisms within M1, such as intracortical inhibition. It is important to note that during precisely controlled force generation there was a quick release component and during precisely controlled force relaxation there was a quick generation component (Fig. 1C). If one were able to tease these apart, it is most likely that motor cortex would have greater activation for the quick generation component of the force relaxation condition compared with the quick release component of the force generation condition. Because we found that precisely controlled force generation had greater activity in the motor cortex than precisely controlled force relaxation, this suggests that this issue makes it harder to detect an effect and that our findings are robust to this issue. The discussion will next focus on task differences in areas beyond motor cortex.
Task Differences beyond Motor Cortex
Beyond motor cortex, we found that right DLPFC had greater activation during precisely controlled force relaxation compared with force generation. One interpretation of this finding is that right DLPFC is involved in reducing the level of force, because the force relaxation task is more difficult than the force generation task. If force relaxation was a more difficult motor task than force generation, then force error could be greater. Indeed, previous TMS and fMRI studies have shown that when the precision grip force requirements increase, this leads to an increase in corticospinal excitability and an increased BOLD signal in the motor cortex (Gallea et al. 2005; Bonnard et al. 2007). However, in the current study grip force error was the same between tasks. Indeed, neural activation could be related to task difficulty even though force error was the same. However, previous studies have shown that increased difficulty of a motor task is associated with increased activation in pre-SMA (Picard and Strick 1996; Toma et al. 1999; Lehericy et al. 2006), and in the current study there were no difference in pre-SMA between the 2 tasks.
We suggest that there are 2 potential explanations for the findings in DLPFC. First, right DLPFC could be actively involved in reducing the level of force during precision grip. Second, right DLPFC could be inhibiting the prepotent response of relaxing the grip quickly. Either of these interpretations is supported by the literature that has examined response inhibition using Go/No-go tasks. Go/No-go tasks require subjects to make a movement in response to a visual stimulus (Go trials) or require subjects to inhibit the movement in response to a different stimulus (No-go trials). Electrophysiological work in monkeys identified prefrontal cortical neurons that responded preferentially to the No-go trials (Gemba and Sasaki 1989; Watanabe 1986a, 1986b). Other work has shown that electrical stimulation of No-go responsive areas in prefrontal cortex either canceled or delayed the trained movement associated with Go trials (Sasaki et al. 1989). Human EEG studies have identified a prefrontal event-related potential change that is thought to be mediating response inhibition because it has greater amplitude during No-go trials than during Go trials (Kok 1986). Also, the event-related potential increases in amplitude as it becomes more difficult to successfully inhibit a motor response (Jodo and Kayama 1992; Eimer 1993). Human neuroimaging work using positron emission tomography (Kawashima et al. 1996) and fMRI (Garavan et al. 1999; Watanabe et al. 2002; Nakata et al. 2008) have found increased regional cerebral blood flow or BOLD activation in DLPFC associated with response inhibition in Go/No-go tasks. Taken together, these studies suggest that populations of prefrontal cortical neurons become activated when inhibiting an anticipated motor response during No-go trials. It is also important to note that our finding for DLPFC was specific to the right hemisphere, consistent with the hypothesis that inhibition may be more dominant on the right compared with the left hemisphere (Kawashima et al. 1996; Garavan et al. 1999). The current study extends these findings during No-go tasks by demonstrating that right DLPFC has greater activation when precisely relaxing grip force.
Although DLPFC had greater activation during the precise relaxation of grip force, the current study found that the activation in bilateral caudate nucleus was greater while generating grip force than while relaxing grip force. This finding is different to other studies examining inhibition of a motor response prior to it beginning. For instance, an fMRI study that used a stop-signal paradigm in humans found greater activation in bilateral caudate nucleus for stop trials where the subject successfully inhibited the motor response compared with stop trials in which the subject failed to inhibit the motor response (Vink et al. 2005, 2006). Moreover, the level of BOLD activation in bilateral caudate increases parametrically when the probability for a stop signal is increased (Vink et al. 2005). Another stop-signal fMRI study showed that the level of activation in bilateral caudate is greater for subjects that respond quickly to the stop-signal compared with subjects that respond slowly to the stop signal (Ray Li et al. 2008). The main difference between stop-signal paradigms and the current paradigm is that we examined the reduction of grip force from a 15% target, whereas stop-signal paradigms examine the inhibition of an impending motor response prior to its beginning. As such, the neural activity in caudate nucleus may be important in determining how and when motor output is reduced.
It is important to point out that this study examined the differences between force generation and force relaxation. Previous studies have shown that a lesion to the GPi can lead to difficulty in turning off muscle activation (Mink and Thach 1991). This could raise the question as to why did we find no difference between force generation and force relaxation in the globus pallidus? For the most part, force generation and force relaxation share a common network, and areas active during force generation are also active during force relaxation. This is true for the globus pallidus, as this structure was active during both tasks but there was not a difference between tasks. Thus, if a structure such as the globus pallidus is active during force relaxation, and there is a lesion placed in this structure, it follows that force relaxation could be impaired.
In conclusion, precisely controlled force generation activates a similar network to precisely controlled force relaxation, yet there are also fundamental differences in how the brain regulates these 2 force tasks. It was found that the caudate and M1 have greater activity during precisely controlled force generation compared with precisely controlled force relaxation, whereas the DLPFC has greater activity during force relaxation compared with force generation. The current study also found that there was greater deactivation in bilateral ACC for precisely controlled force relaxation than precisely controlled force generation. In general, the findings are important because they are both consistent with (M1 and DLPFC) and different to (caudate) the literature which has examined motor inhibition before the contraction takes place. This suggests that the brain does not always use the same mechanism to turn off muscle, and that it depends on when and how quickly the task constraints determine the nature of the inhibitory process. In addition, it is interesting to note that activation in DLPFC, ACC, and caudate has consistently been found during tasks that include conflict resolution, cognitive control, and selection (Rowe et al. 2000; Anderson et al. 2007; Sohn et al. 2007; Vaillancourt et al. 2007). It remains unclear to what degree the current task of increasing and decreasing force relies upon the same circuits as tasks that are considered cognitive. Further studies that manipulate cognitive and motor tasks may further segregate the processes that are regulated by these structures.
National Institutes of Health (R01-NS-52318, R01-NS-58487, R01-NS-40902, R01-NS-28127).
We thank Dr Sue Leurgans at Rush University Medical Center for providing consultation on the statistical design. Conflict of Interest: None declared.