Spinal manipulation induces somatomotor changes, that is, changes in muscle activity, apparently because of sensory input from the somatic nervous system. In asymptomatic patients, Herzog's group [279, 280] showed that PA spinal manipulative treatments applied to the cervical, thoracic, lumbar, and sacroiliac regions increased paraspinal EMG activity in a pattern related to the region of the spine that was manipulated. The EMG response latencies occur within 50 to 200 milliseconds after initiation of the manipulative thrust. Similarly, SM using an Activator-adjusting instrument applied to a transverse process elicited paraspinal EMG activity at the same segmental level but within 2 to 3 milliseconds.  This is surprisingly fast for a reflex response. Colloca and Keller  confirmed these latter findings in symptomatic patients with low back pain and, in addition, reported that the increased EMG activity, while beginning within 2 to 3 milliseconds of the manipulation, reached its peak within 50 to 100 milliseconds. Paraspinal EMG responses were greatest in magnitude when the manipulation was delivered close to the electrode site, and interestingly, the more chronic the low back pain, the less the EMG response. The EMG electrodes were not placed relative to any physical finding in the low back such as palpable muscle tension, as perceived by the practitioner or tissue tenderness as experienced by the patient.
Spinal manipulation's effect on paraspinal muscle activity is not exclusively excitatory. In 1 symptomatic patient with spontaneous muscle activity in the thoracic spine, Herzog's group  observed reduced paraspinal EMG activity within 1 second after a thoracic SM. In a case series study, DeVocht et al  collected surface EMG activity from 16 participants in 2 chiropractic offices. Electrodes were placed over 2 sites exhibiting paraspinal muscle tension determined by manual palpation. Spinal manipulation was administered to 8 participants using Activator protocol. The other 8 were treated using Diversified protocol. EMG activity was decreased after treatment by both methods by at least 25% at 24 of the 31 EMG recording sites.
The effects of SM on paraspinal EMG activity may also be associated with increases in muscle strength. Suter et al  studied symptomatic patients with sacroiliac joint dysfunction, anterior knee pain, and evidence of motor inhibition to knee extensor muscles. A side posture SM applied to the sacroiliac joint significantly decreased the inhibition of the knee extensors on the side of the body to which the manipulation was applied. Similarly, Keller and Colloca  found that erector spinae isometric strength (assessed using EMG) was increased after spinal compared with sham manipulation.
A series of studies have addressed how SM affects central processing of somatomotor information. Spinal manipulation can increase the excitability of motor pathways in the central nervous system and depress the inflow of sensory information from muscle spindles to these motor pathways. This may, in part, account for the disparate clinical findings described above. In asymptomatic patients, Dishman et al  showed that SM increased central motor excitability. EMG activity from gastrocnemius muscle, evoked by direct activation of descending corticospinal tracts using transcranial magnetic stimulation, was larger after lumbar SM compared with simply positioning the patient but not applying the manipulation. However, SM can also depress the H-reflex. Manipulation applied to the sacroiliac joint in a PA direction decreased the magnitude of the tibial nerve H-reflex for up to 15 minutes in asymptomatic humans.  Similarly, side-posture lumbar manipulation of L5–S1 joint inhibited the H-reflex from the tibial nerve.  Mobilization alone but not massage also inhibited the tibial nerve H-reflex, but the effect of manipulation tended to be greater. [288, 289] After manipulation alone, the inhibition lasted for approximately 20 seconds but lasted up to 1 minute when the SM was preceded by spinal mobilization. Similarly, SM delivered to the cervical region depressed the median nerve H-reflex.  The magnitude of the response from the lumbar manipulation was greater than the response from the cervical manipulation, suggesting that central processing of sensory inputs from a SM is different in the neck and the low back.  The depression of the H-reflex does not appear to be a global response. Instead, it appears specific to the region of the spine manipulated because cervical manipulation did not affect the tibial nerve H-reflex.  Patient positioning, which flexes the lumbar spine before the manipulation, may augment the inhibition of tibial nerve the H-reflex. 
A possible mechanism contributing to SM's inhibitory effects on the H-reflex and on spontaneous paraspinal EMG activity is suggested by recent experiments. Sensory input from tissues of the facet joint elicited by SM might reflexively decrease paraspinal muscle activity. Indahl et al  elicited reflex longissimus and multifidus EMG activity by electrically stimulating the intervertebral disk in a porcine preparation. Stretching the facet joint by injecting 1 mL of physiologic saline abolished the EMG activity.
Haldeman's group [238, 294] has shown that SM can also affect higher centers in the brain. Using magnetic stimulation, Zhu et al  stimulated lumbar paraspinal muscles and recorded the evoked cerebral potentials. Stimulation of paraspinal muscle spindles using vibration reduced the magnitude of the cerebral potentials. Similarly, muscle spasm in human patients reduced the magnitude of the paraspinal muscle evoked cerebral potentials. Spinal manipulation reversed these effects, reducing muscle spasm and restoring the magnitude of the evoked cerebral potentials. 
There is reason to believe that stretching the facet joint capsule and surrounding tissues likely occurs during SM, although this has received little study.  Furthermore, there may be reason to believe that the mechanically sensitive primary afferents could be stimulated beyond the short duration of an SM. Using MRI scans in human subjects, Cramer et al [10, 11] showed that a side-posture SM accompanied by cavitation gaps the facet joints. The synovial space of the lumbar facet joints increased in width an average of 2.2 mm in subjects who were positioned in side posture and received a side posture spinal adjustment. By comparison, the joint space widened by only 1.5 mm (a difference of 0.7 mm) in subjects who were positioned in side posture but did not receive a manipulation. The MRI scan was performed immediately after manipulation and lasted 20 minutes. Although not studied directly, it seems likely, based upon data from the laboratory of Khalsa,  that joint separations of these magnitudes are sufficient to load the facet joint tissues. If so, this raises the possibility that tissues surrounding the facet joint could be stretched for periods longer than the duration of the manipulation itself. Sensory input from tissues surrounding the facet joint that is graded with direction of facet movement  could elicit reflex muscle responses similar to those measured by Indahl et al. 
Direct evidence from 1 of the experimental models described at the outset of this section  shows that the impulse load of an SM activates a variety of low-threshold mechanoreceptors in paraspinal muscles and that abrupt changes occur in the discharge from their parent afferent neurons [251, 295] as the speed of delivery approaches that used in clinical practice. [296–299] Pickar and Wheeler  recorded afferent activity from muscle spindle and Golgi tendon organ afferents having receptive fields in the lumbar multifidus and longissimus muscles while applying a spinal manipulative-like load to a lumbar vertebra. Muscle spindle afferents from lumbar multifidus and longissimus muscles were stimulated more by the impulse of an SM than by the load preparatory to the impulse (200% compared with 30%). Another type of low-threshold mechanoreceptor, a presumed Pacinian corpuscle, uniquely responded to the impulse of a manipulative-like load, that is, it did not respond to loads with a slower force-time profile. When an SM's duration was varied between 25 and 800 milliseconds, durations shorter than 400 milliseconds produced abrupt increases in discharge rates from 6 low-threshold mechanoreceptive afferents innervating the lumbar multifidus and longissimus muscles.  An increase in loading magnitude did not appear to systematically affect the discharge from these 6 low-threshold mechanoreceptors. Interestingly, Gillette et al [263, 265] showed that both weak and strong mechanical stimuli applied to paraspinal tissues can suppress spinal cord neurons that receive noxious input from the low back. In an anesthetized human patient undergoing an L4–L5 laminectomy, SM of the lumbosacral region evoked multiunit activity from the intact S1 nerve root.  This neural discharge measured in a clinical setting may be analogous, at least in part, to the discharge of low-threshold mechanoreceptors measured in an animal model.