The Spinal Cord: Laminae, Tracts, and Local Computation

01Cord gray and white matter

The spinal cord sorts its cells by lamina and its axons by destination

The spinal cord sorts its cell bodies into ten laminae and its axons into three funiculi, and both sortings predict function. Gray matter sits at the center in the shape of a butterfly, wrapped by columns of myelinated white matter. Cell bodies gather into dorsal, intermediate, and ventral horns, which Rexed divided into laminae I through X. Ascending and descending axons run outside them in the dorsal, lateral, and ventral funiculi. The dorsal horn receives sensory afferents. The ventral horn contains the alpha and gamma motor neurons that drive muscle. The intermediolateral cell column occupies the lateral gray of the thoracic segments and sends sympathetic fibers out of the cord. The gray thickens at the cervical and lumbar enlargements, exactly where the limb plexuses draw extra motor neurons into the ventral horn.

The top lamina of the dorsal horn contains a countable population. The rat L4 segment yields about 400 lamina I projection neurons on each side, and roughly 120 of them project to the periaqueductal gray Spike 2003. About 85 percent of that rat population can be labeled from the caudal ventrolateral medulla or the lateral parabrachial area on the opposite side. Many of those cells branch to more than one of those targets.

Where the spinal cord ends is a range, not a fixed point

The conus medullaris tip sits at the lower third of the L1 vertebral body in most adults, and across 1,000 consecutive MRI studies none sat above T12 Elfiky 2025. Imaging of a separate 944 adults found 18.9 percent terminating at or below the L1 to L2 interspace Paul 2025. Tuffier's line corresponded to the L4 to L5 interspace in 99.8 percent of that cohort's reviewed lumbar radiographs, 720 films in all. A needle is therefore planned against the iliac crest and not against an assumed cord level. Nearly one adult in five ends lower than the L1 to L2 line classic teaching treats as clear.

The ascent that produces those numbers happens before birth. Ultrasound placed the conus at the L2 to L3 intervertebral space in premature infants of 27 to 29 postmenstrual weeks. It reached the adult L1 to L2 level around the 40th postmenstrual week Beek 1996. Beek measured no further ascent in that cohort of 99 children. A larger series caught the last of it. In 1,077 infant ultrasound scans the conus lay caudal to L2 in 13 percent of babies at 0 to 30 days Rozzelle 2014. That fell to 11.4 percent at 31 to 60 days and 4.7 percent at 61 to 100 days. A term newborn already sits at or near the adult level, and the remaining one in eight clears over the first three months.

The termination decides which structure an injury strikes. A lesion at T12 meets cord, and a lesion at L3 meets the roots of the cauda equina. The syndromes belonging to each side of that boundary are set out in the cord and nerve root.

02Findings

What the research shows

03Descending motor tracts

Descending tracts split into a lateral system for the limbs and a medial system for posture

The lateral corticospinal tract is the largest descending pathway in the spinal cord, and roughly 90 percent of corticospinal fibers decussate at the pyramidal crossing in the lower medulla. From there they run in the lateral funiculus to contralateral limb motor neurons. The rubrospinal tract joins them from the midbrain red nucleus and biases flexor tone in the contralateral limbs.

Human material makes that crossing untidy. Three-quarters of human spinal cords are asymmetric, and in three-quarters of the asymmetric ones the right side is the larger, because more corticospinal fibers cross to the right Nathan 1990. The level of the crossing differs between the two sides of the same cord. The asymmetry bears no relation to handedness. No two cords distribute the descending load in the same arrangement.

The uncrossed remainder varies just as widely. Across 15 late-stage human fetuses of 28 to 30 weeks the anterior corticospinal tract ran from bilaterally thick, to unilaterally thick, to thin or absent Jin 2016. In the thickest group its cross-sectional area reached 40 percent of the lateral tract. In the 50 midterm fetuses of the same study the lateral corticospinal tract appeared in the cervical cord at 14 to 15 weeks of gestation. The anterior tract was found in only 3 of them. The often-quoted figure of about 8 percent descending uncrossed in the anterior corticospinal tract is an average that few individual cords hold, and a small further remainder stays uncrossed inside the lateral tract itself.

The decussation is instructed by a gene

Mutations in ROBO3 remove the crossing. ROBO3 is required for the medullary decussation of the corticospinal and medial lemniscal tracts, and its loss produces horizontal gaze palsy with progressive scoliosis Sicotte 2006. Diffusion tensor imaging in a patient carrying the mutation showed the major pontine crossing tracts absent, with no decussation of the superior cerebellar peduncles. The pyramidal crossing is an event the developing cord performs on instruction, and a body built without it declares the loss in its eyes and its spine.

Fiber caliber sets what the cable can deliver

The adult human medullary pyramid packs about 66,000 fibers per square millimeter, with a standard deviation of 11,000 Graf von Keyserlingk 1984. Electron microscopy of four perfusion-fixed human pyramids sorted them by total diameter. Fibers below 4 micrometers made up 87.9 percent, fibers of 4 to 10 micrometers 10.77 percent, and fibers above 10 micrometers only 1.4 percent. Nearly nine in ten corticospinal fibers are thin and slow, so command arriving at a spinal segment is spread across a wide band of conduction times instead of landing as one synchronized volley.

Reticulospinal fibers scatter through the anterior and lateral columns

Reticulospinal fibers descend bilaterally in the human spinal cord with a preponderance of ipsilateral fibers Nathan 1996. They form no well-defined tract. They scatter through the anterior and lateral columns intermingled with descending propriospinal fibers, gathering into a compact triangle only at the periphery of the lateral column, immediately anterior to the lateral corticospinal tract. The medial system governs bilateral axial muscles, head and neck position, balance, and gait, running through the anterior corticospinal tract, the vestibulospinal and reticulospinal fibers, and the tectospinal tract. Where the lateral system writes precise distal movement, the medial system maintains the postural background that movement is performed against.

The same descending material supplies the autonomic line. A compact group of central sympathetic fibers surrounds the lateral horn and terminates in the intermediolateral and intermediomedial cell columns Nathan 1996. Somatic and autonomic control sit one lamina apart in the same segment, so a single level of injury reshapes movement and organ regulation together. What happens to that outflow after it leaves the cord is set out in the autonomic nervous system.

04Dorsal column and anterolateral

The spinal cord runs sensation in columns separate enough that damage to one predicts a specific loss

In 42 people with multiple sclerosis, spinal cord magnetization transfer imaging tied the dorsal column signal to vibration sensation at R = 0.58. The lateral column signal tracked strength instead, at R = -0.45 Zackowski 2009. One imaging session separated the columns by what their damage costs. Two pathways lie behind that separation, and they cross at different levels.

The dorsal column medial lemniscus pathway conveys fine touch, vibration, and conscious proprioception. It ascends ipsilaterally as the fasciculus gracilis from the legs and the fasciculus cuneatus from the arms, then synapses in the nuclei gracilis and cuneatus of the medulla. Only there do its second order fibers decussate, forming the medial lemniscus that reaches the ventral posterior lateral nucleus of the thalamus. This is the high fidelity proprioceptive traffic that gives the nervous system a running map of where the body is.

The anterolateral system runs the opposite discipline. The lateral spinothalamic tract conveys pain and temperature from dorsal horn neurons whose fibers cross within a segment or two through the anterior commissure. They ascend the opposite anterolateral cord to the same thalamic nucleus. Lesioning one anterolateral quadrant confirms the pathway. Among 224 cancer patients treated with CT-guided percutaneous cordotomy, 98.13 percent reported initial pain relief Kanpolat 2013. The median visual analogue score fell to 0 on the first postoperative day, down from 8.

Because the two systems cross at different levels, a hemisection dissociates them. Proprioception is lost on the side of the lesion below it, and pain and temperature are lost on the opposite side. That is the Brown-Sequard pattern, and the family of cord syndromes it belongs to is set out in the cord and nerve root.

The dorsal column also carries visceral pain

A limited midline myelotomy at T10, a lesion of the dorsal column and nothing else, relieves pelvic cancer pain in patients Willis 1999. In rats and monkeys a dorsal column lesion at the same level dramatically reduces thalamic responses to noxious colorectal distension, while cutting the spinothalamic tract does not. In anesthetized monkeys the regional blood volume response to distension was eliminated outright. In the rat the postsynaptic dorsal column neurons responsible concentrate in the central gray matter of the L6 to S1 segments. The two-pathway account describes limb sensation accurately and the viscera poorly, and pelvic pain answers to a lesion in the column assigned to touch.

Clarke's column runs a dedicated line to the cerebellum

The dorsal spinocerebellar tract rises from the column of Clarke in lamina VII, spanning segments C8 to L2 and often cited as C8 to L3. Genetic tracing in mice showed Clarke's column mossy fiber terminals diversifying extensively in the cerebellar cortex and terminating bilaterally, with no significant axon collaterals in the spinal cord, medulla, or cerebellar nuclei Pop 2022. Neighboring Atoh1-lineage proprioceptive populations in the same mice do wire to motor neurons locally. Proprioception sent upward and proprioception used for local reflex leave the cord as separate populations.

05Local computation in the cord

The spinal cord closes its own sensorimotor loops and can generate stepping with no brain attached

Mean soleus H-reflex latency measured 30.93 ms on the right and 31.01 ms on the left in 119 healthy adult male volunteers Gupta 2024. That latency is the time a signal takes to reach the spinal cord and return to muscle. Latency tracks leg length at r = 0.55, and a side-to-side difference above 1.8 ms is abnormal. Thirty one milliseconds leaves no room in the circuit for the brain.

The loop is short by construction. A monosynaptic connection links a muscle spindle afferent directly to the homonymous alpha motor neuron, and reciprocal inhibition relaxes the antagonist through one interposed inhibitory interneuron. Withdrawal and crossed extensor reflexes coordinate whole limbs through polysynaptic chains. Human microneurography puts the resting spindle discharge at about 10 Hz, far below the cat Macefield 2018. Independent fusimotor control is present in humans and modest next to the cat's, so the magnitude of a cat result does not transfer. The receptor supplying the afferent limb, and the fusimotor traffic that sets its sensitivity, are set out in muscle spindles and proprioception.

Central pattern generators sequence stepping inside the cord

An electrical train applied over the second lumbar segment at 25 to 60 Hz and 5 to 9 V induced stepping in paraplegic subjects with complete, long-standing spinal cord injury Dimitrijevic 1998. The legs produced rhythmic alternating stance and swing phases. The drive carried no pattern. The pattern came from the cord. Central pattern generators in the intermediate gray sequence the basic rhythm of stepping, and those circuits are the substrate of the resting readiness the cord sets and the brain modulates.

One man with motor-complete paraplegia from a C7 to T1 subluxation achieved full weight-bearing standing for 4.25 minutes under epidural stimulation of the lumbosacral cord, with assistance provided only for balance Harkema 2011. Tonic stimulation and bilateral load-bearing proprioceptive input were enough to produce it in a cord receiving no clinically detectable voluntary command. Seven months after implantation he recovered supraspinal control of some leg movements, but only while the stimulation was running. The cord is the first integrator the brain delegates to, and it keeps that delegation running when nothing usable is arriving from above.

Stimulation that overwrites the afferent report stops working

Epidural stimulation blocks a significant amount of proprioceptive input in humans, though not in rats, a difference traced through computational simulation alongside matched preclinical and clinical experiments Formento 2018. That transient deafferentation prevents modulation of the reciprocal inhibitory networks locomotion runs on, and it reduces or abolishes conscious perception of leg position. An input strong enough to drive muscle destroyed the traffic the circuit was built out of. Continuous stimulation therefore helps only inside a narrow band of parameters, while burst and spatiotemporal profiles spare the afferent traffic that continuous stimulation cancels.

Stiffness after cord injury tracks something other than reflex gain

Resting tonic stretch reflex gain in 13 subjects with spinal cord injury averaged below 0.5 microvolts per degree in dorsiflexion and plantarflexion, significantly lower than in resting normal subjects Woolacott 2006. Linear regression found no relationship between that gain and resting joint stiffness, and elastic resistance at the ankle was lower after cord injury than in normal subjects, not higher. The stiffness of a spinal limb has no single reflex behind it. Spinal hypertonia is a different object from post-stroke spasticity, where the same low-frequency stretch does raise reflex gain in step with the stiffness.

06Regional selectivity

Mechanical input to the back changes what the spinal cord produces, and the change is regional and brief

Pinching the skin of the back abolished spontaneous locomotor-like activity in 5 decerebrate cats spinalized a month earlier and treated with clonidine Frigon 2012. The same pinch abolished the rhythmic clonus-like episodes above 3 Hz that appeared during reflex testing, without eliminating the reflex responses themselves. Cutaneous input to the trunk rewrote the output of a lumbar cord separated from the brain a month before.

Lumbar spinal manipulation significantly attenuated the tibial nerve H-reflex in 36 subjects without low back pain, and the motoneuron pool returned to baseline within 60 seconds Dishman 2002. Cervical manipulation in the same subjects produced no significant effect on lumbar motor neurons. Mechanical input to one spinal level reaches the alpha motoneuron pool at that level, for about a minute, without spreading along the cord.

Levels do not weigh equally. Segments C1 to C3 supply about 45 percent of all spinothalamic neurons in the cat Mouton 2005. The same three segments supply about 30 percent of all spinal neurons projecting to the periaqueductal gray, counted from more than 2,400 labeled cells. Those same segments are fed by upper neck muscle, tendon, and facet joint input.

A spinal segment answers from the state it is already holding

Tone shows up in the cord as the state one segment is already holding when traffic arrives. Neurophysiology already names the running sum of excitatory and inhibitory influence converging on one neuron and calls it the cell's central integrative state, the baseline deciding how that cell answers the next signal. A spinal segment holds a great many of those sums at once, command and sensation interleaved in one column.

Axial tension, not compression, sets the neurophysiological consequence, because even a compressive lesion stretches the cord above and below it. Alf Breig showed this in cadaveric and operative material in Adverse Mechanical Tension in the Central Nervous System. The body sets that tension through the dentate ligaments, the dura, and the muscles that move the canal, so the mechanical state of the cord is something it regulates rather than merely suffers. The Unified Model of Tone therefore counts strain as part of the signaling rather than a condition on it. Change the tension on a cord, and the fidelity, the timing, and the rate of what passes through it change with it.

The model predicts what follows. An input matched to the excitability a segment is already set at changes what that segment transmits, and the change appears where the segmental map sends it and nowhere else. That prediction is regional selectivity, and it is measurable at the motoneuron pool. Lumbar manipulation attenuated the tibial nerve H-reflex, while cervical manipulation left the same pool untouched Dishman 2002.

A segment already holds a pattern before anything arrives, and a chiropractic adjustment meets that pattern with graded mechanical input at one level of the column. The cord answers in reflex amplitude and in seconds. The reply is not a movement but a change in what the motoneuron pool at that level will do next, and in a healthy segment it is over inside a minute. Frigon's cats show the same class of input working on a cord that has nothing above it, which is where the reply is easiest to see.

Health at this level is a band and not a value. The soleus H-reflex keeps its latency inside a side-to-side tolerance of 1.8 ms, and a working cord keeps each segment inside a comparable window. The pathology is a segment that no longer has a band. One kind is stuck at a single output, as in the spinal limb whose stiffness no longer tracks any reflex. The other swings past the edges of its range, as in the clonus that runs when the descending brakes are gone.

07Reading a cord level

A spinal cord lesion declares its level and its tract, which is why the cord can be localized before an image exists

A complete transverse lesion of the spinal cord abolishes everything below it. Selective tract involvement writes a legible signature instead. Dorsal column loss strips vibration and position sense. Anterolateral loss removes pain and temperature. Ventral horn or root damage produces the lower motor neuron weakness and wasting seen in conditions such as cervical spondylotic amyotrophy. Damage to the lateral corticospinal tract gives the upper motor neuron picture below the level, with spasticity, hyperreflexia, and a Babinski sign.

That sign weighs less than its reputation. Across 375 participants its sensitivity was 49.6 percent and its specificity 85.8 percent, with interobserver agreement of kappa 0.45 against kappa 0.83 for finger and foot tapping Appasamy 2018. The Babinski sign misses half of corticospinal lesions and seldom appears without one. A level is read from the whole pattern of spared and lost tracts, with one reflex serving as evidence inside it.

Bony level and neurological level are not the same level

Vertebral and cord segments fall out of register during growth, so a sensory level read on the trunk maps to a cord segment several vertebrae higher. The cord ends near L1 to L2, so an injury below that point strikes the cauda equina and carries a root level prognosis rather than a cord one.

Blood supply to the cord follows one variable feeder

The anterior spinal artery feeds the ventral two thirds of the cord, including the corticospinal tracts. MDCT angiography found the artery of Adamkiewicz in 52 of 73 patients, 71.2 percent, delineating 64 arteries in all Sukeeyamanon 2010. Of those 64 vessels, 89.1 percent arose between T9 and L2 and 41 originated on the left. Nine of the 73 patients had two feeders. The dominant supply to the lower cord is a single variable vessel that imaging fails to find in nearly a third of patients, so cord infarction is level-specific and hard to predict.

The autonomic line keeps its own address in the gray matter. The intermediolateral cell column occupies the lateral horn of the thoracic and upper lumbar segments, one lamina from the ventral horn motor neurons of the same level. A compact group of central sympathetic fibers surrounds that horn and terminates in the intermediolateral and intermediomedial cell columns Nathan 1996. What the thinning of that column costs a body is set out in autonomic regulation. One level of the cord governs movement and organ control together, and one lesion takes both.

Read together, these signs place a lesion by column and by segment before any scanner is involved. The spinal cord is a map whose tracts and segments locate a problem in space before a single image is taken.

08Tone

How this system expresses tone

Cut the neuraxis above the lumbar cord and the segment below still computes. A train at 25 to 60 Hz over the second lumbar segment drove alternating stance and swing in legs their owner could not move. Tone shows up here as what the columns permit, what the afferents deliver, and what one segment settles between them.

Set point: a segment sits at a resting gain, and after complete spinal cord injury the resting tonic stretch reflex gain averaged below 0.5 microvolts per degree. Gain: strip the descending brakes and the same reflexes run unchecked below the level as hyperreflexia and clonus. Coupling: one Ia afferent excites its own motor pool and inhibits the antagonist through a single interposed interneuron, so one stretch is answered on both sides of the joint at once. Time course: a reflex closes in 31 ms, a manipulated motoneuron pool resets in 60 seconds, and the conus finishes its ascent over three months. Load: the ventral two thirds of the cord depend on one anterior artery whose lower feeder cannot be imaged in nearly a third of patients. Prediction: Clarke's column posts proprioception to the cerebellum ahead of the movement it belongs to, so the correction is drafted before the limb has finished arriving.

09Across the library

How this page relates to the rest of the library

10Frequently asked

Questions about this topic

11The sources

References

Sources: primary literature, linked inline.