Motor ControlNeuroscience

The Central Pattern Generator Experiments – Sten Grillner

A detailed academic exploration of Sten Grillner’s pioneering central pattern generator experiments, detailing spinal networks, cellular mechanics, and locomotion.

memjavad
PUBLISHED
Scientifically Reviewed · Dr. Marwa Abd-Alazim · September 12, 2026
Medically & Scientifically Reviewed Verified: September 12, 2026
Dr. Marwa Abd-Alazim Ph.D.
Professor of Psychology • University of Kerbala
Review Criteria & Clinical Standards

This content undergoes rigorous scientific peer-review and medical editorial standards at Arab Psychology Network to ensure clinical accuracy, validity, and compliance with evidence-based guidelines from leading psychological and healthcare authorities (APA / WHO).

The control of animal movement represents one of the most sophisticated achievements of biological evolution. From the rhythmic undulation of an eel navigating aquatic currents to the steady, quadrupedal pacing of a cheetah or the bipedal striding of a human, locomotion demands an intricate synchronization of motor units across multiple joints and axial segments. For decades, classical neurophysiology was locked in a profound conceptual dispute regarding the fundamental architecture of this motor execution: does coordinated locomotion emerge as a continuous chain of peripheral sensory reflexes, where each mechanical displacement triggers the subsequent muscular contraction, or is it generated from within the central nervous system by intrinsic, autonomous neural networks? The resolution of this debate fundamentally transformed modern neuroscience, inaugurating the contemporary paradigm of motor control.

At the center of this transformation stands the Swedish neurophysiologist Sten Grillner, whose foundational investigations at the Karolinska Institute dismantled classical reflex-dominant dogmas. Grillner demonstrated that the vertebrate central nervous system possesses specialized, self-contained neural assemblies known as Central Pattern Generators (CPGs). These distributed microcircuits, situated within the spinal cord and lower brainstem, are intrinsically capable of generating the precise spatiotemporal patterns of rhythmic motor output required for swimming, walking, flying, and respiration—even in the total absence of rhythmic sensory feedback or descending cortical timing signals. By meticulously isolating and decoding the biological components of these networks, Grillner achieved what few neuroscientists had accomplished: a complete, multi-level mechanistic description of a vertebrate motor behavior spanning from the biophysics of individual ion channels to synaptic topology, computational modeling, and whole-body mechanical dynamics.

The cornerstone of Grillner’s scientific triumph was his strategic choice of the sea lamprey (Petromyzon marinus) as an experimentally tractable vertebrate model system. Through this phylogenetically ancient, unmyelinated cyclostome, Grillner and his colleagues unveiled the universal blueprint of vertebrate motor coordination. The subsequent sections of this comprehensive treatise explore the historical antecedents, experimental paradigms, cellular microcircuits, biophysical mechanisms, supraspinal hierarchies, neuromodulatory dynamics, sensory integration systems, and translational legacies of Sten Grillner’s seminal central pattern generator experiments.

1. Introduction to Sten Grillner and the Conceptual Paradigm of Central Pattern Generators

The formal conceptualization of central pattern generators revolutionized the neurobiology of movement by shifting the locus of motor timing from the periphery to the central nervous system. Rather than viewing the spinal cord as a passive conduit for sensory-motor reflex arcs, modern neurophysiology identifies the spinal neuraxis as an active, computational organ capable of autonomous pattern formation.

1.1 Historical Context and Definition of Central Pattern Generators

The conceptual origins of autonomous rhythmic neural networks trace back to early twentieth-century debates regarding the genesis of motor patterns. For nearly a century, the prevailing dogma of motor physiology was dominated by reflexological concepts, which posited that all coordinated movements represented an orchestrated sequence of discrete reflex arcs. In this schema, the sensory consequences of one movement phase served as the obligate trigger for the subsequent phase. A central pattern generator is formally defined as an ensemble of interconnected neurons capable of producing rhythmic, patterned motor output autonomously—that is, without requiring rhythmic sensory feedback from the periphery or rhythmic timing commands from higher brain centers.

Sten Grillner played a pioneering role in establishing the vertebrate CPG framework during the late 1960s and 1970s at the Karolinska Institute in Stockholm. Synthesizing early, often marginalized observations with rigorous quantitative electrophysiology, Grillner articulated the operational principles that govern vertebrate locomotion. He demonstrated that while sensory feedback dynamically adjusts and calibrates the motor output to match environmental constraints, the fundamental rhythmicity, the intermuscular phase delays, and the left-right antagonistic alternation are generated intrinsically within the spinal gray matter.

The general organizational principles of neural rhythm generation, as formulated through Grillner’s work, dictate that biological locomotion relies on decentralized, modular network architectures. Rather than utilizing a centralized master clock, the spinal cord is organized into a metameric series of interconnected segmental or unit oscillators. Each oscillator controls a specific degree of freedom or mechanical segment, and their mutual synaptic coupling guarantees both stability and flexible phase coordination across varying locomotor frequencies.

1.2 The Search for a Tractable Vertebrate Model System

The quest to decipher the cellular basis of vertebrate locomotion was initially hindered by the immense complexity of mammalian spinal systems. In the adult cat or rodent, the spinal cord contains hundreds of thousands of densely packed neurons, heavily myelinated axon tracts, and a complex vascular architecture that precludes direct optical visualization of living cells. Furthermore, mammalian spinal tissues deteriorate rapidly when deprived of vascular perfusion, making sustained, stable intracellular microelectrode recordings from identified spinal interneurons during active locomotion extraordinarily challenging with mid-twentieth-century technologies.

Recognizing these severe methodological bottlenecks, Sten Grillner made the transformative decision in the late 1970s to adopt the sea lamprey (Petromyzon marinus), a jawless cyclostome representing the oldest surviving lineage of vertebrates, as a reductionist model system. The lamprey separated from the main vertebrate lineage approximately 450 to 500 million years ago, occupying a critical evolutionary juncture. Crucially, the general plan of the vertebrate nervous system—encompassing the forebrain, basal ganglia, brainstem locomotor centers, and spinal circuitry—was already fully established in these primitive creatures, exhibiting profound phylogenetic conservation with modern mammals.

The lamprey offered unparalleled translational and biophysical advantages. Its spinal cord is completely unmyelinated, ribbon-shaped, and structurally transparent, with a total thickness of only 200 to 300 micrometers. This exquisite planar geometry allows oxygen and nutrients to diffuse freely into the core of the tissue from an external physiological saline solution without requiring a functioning vascular system. Consequently, the entire lamprey brainstem and spinal cord can be excised and maintained in vitro for several days at low physiological temperatures (8–12°C), permitting microelectrode impalements of living, identified neurons during the generation of rhythmic motor output.

1.3 Methodological Paradigms: From In Vivo Electrophysiology to In Vitro Preparations

The technological breakthrough that unlocked the fine mechanics of the spinal CPG was the development of the isolated lamprey brainstem-spinal cord preparation. Grillner and his collaborators excised the neuroaxis from the caudal end of the brainstem down to the mid-trunk segments, pinning the flat spinal ribbon in a Sylgard-lined recording chamber superfused with oxygenated, chilled Ringer’s solution. This preparation was completely physically decoupled from muscle tissue, skin, and mechanosensory organs, eliminating all movement artifacts and peripheral sensory loops.

To record motor output under these isolated conditions, the investigators deployed suction electrodes onto the cut ventral roots emerging from the spinal cord. When the preparation was stimulated—either pharmacologically by bath application of neuroactive substances or electrically via descending brainstem tracts—the ventral roots discharged rhythmic, alternating barrages of action potentials that mirrored the muscular activation patterns observed in freely swimming animals. This phenomenon was termed fictive locomotion, serving as an unambiguous, measurable proxy for real locomotor behavior.

Building upon the fictive locomotion paradigm, Grillner’s laboratory integrated advanced cellular electrophysiology, utilizing sharp microelectrodes and subsequently whole-cell patch-clamp recordings to measure the membrane potentials of specific interneurons and motor neurons. By deploying simultaneous paired and triple intracellular recordings, they systematically probed synaptic connections between individual identified cells. They delivered current pulses to a presynaptic interneuron while recording postsynaptic currents or potentials in neighboring targets, quantitatively establishing the sign, latency, amplitude, and pharmacological sensitivity of the underlying monosynaptic and polysynaptic pathways.

2. Historical Antecedents: The Shift from Reflex Chains to Autonomous Spinal Rhythmicity

The modern understanding of central pattern generation emerged from a protracted theoretical and empirical struggle spanning more than a century. To appreciate the impact of Grillner’s work, one must contextualize the intellectual climate that preceded his research, characterized by a persistent tension between peripheralist reflex theories and centralist oscillatory paradigms.

2.1 Sherrington Versus Brown: The Foundations of Motor Control Theory

At the turn of the twentieth century, Sir Charles Sherrington formulated the concept of the reflex as the fundamental integrative unit of the nervous system. Sherrington proposed that locomotion was governed by a proprioceptive reflex chain. In this model, the passive mechanical extension of a limb during the stance phase of walking activated stretch receptors in extensor muscles and articular capsules; the resulting sensory discharge triggered a reflex contraction of flexor muscles, initiating the swing phase. In turn, the mechanical termination of the swing phase triggered the subsequent extensor contraction. Sherrington’s framework was elegant, intuitively appealing, and supported by extensive empirical analyses of spinal reflexes in decerebrate and spinalized mammals.

However, in 1911, Sherrington’s younger colleague, Thomas Graham Brown, challenged this view. Brown performed a series of experiments on low-spinal cats in which he severed all dorsal sensory roots (complete surgical deafferentation) and abolished all muscular contractions using curare. Remarkably, when Brown exposed the transected spinal cord to low levels of chemical stimulation or asphyxia, he recorded rhythmic, alternating contractions in the antagonistic flexor and extensor nerves. Because there were no muscle movements and no sensory feedback signals arriving at the cord, Brown concluded that the neural rhythm was generated centrally.

Brown formulated the legendary half-centre hypothesis, postulating that the spinal locomotor center consists of two symmetrical pools of interneurons: one driving flexor motor neurons and the other driving extensor motor neurons. These two half-centers were proposed to be coupled by reciprocal inhibitory connections. A steady, unpatterned descending excitatory drive would activate both half-centers, but minor stochastic asymmetries would allow one center to fire first, simultaneously suppressing the other through mutual inhibition. Through an intrinsic fatigue mechanism or depression of inhibition, the active half-center would eventually cease firing, permitting the contralateral center to escape from inhibition and initiate its own burst. Despite its profound explanatory power, Brown’s model was overshadowed for half a century by the dominant Sherringtonian reflex paradigm.

2.2 The Deafferentation Experiments and Evidence of Intrinsic Timing

The rediscovery and definitive validation of Brown’s central rhythm hypothesis began in the 1960s and culminated in Grillner’s rigorous experimental synthesis in the 1970s. The conceptual turning point required indisputable proof that sensory feedback was neither necessary nor primary for the generation of the timing, sequencing, and phase relationships of locomotor patterns. Grillner and other pioneers, such as Donald Wilson working on the locust flight system, established a stringent set of operational criteria to test the autonomy of spinal circuits.

Grillner executed comprehensive deafferentation protocols in vertebrates, systematically transecting all dorsal root sensory fibers along the neuroaxis. When these animals were placed on motorized treadmills or stimulated via descending spinal pathways, they maintained rhythmic, alternating locomotor activity with normal interlimb phase sequences. To eliminate the counterargument that undetected visceral or micro-sensory inputs could still provide a peripheral timing signal, preparations were paralyzed using neuromuscular blocking agents such as gallamine or curare. Under these curarized conditions, the recorded motor nerves continued to fire in precise alternating bursts matching normal locomotion.

Through these experiments, Grillner formalized the criteria required to establish the existence of a Central Pattern Generator:

  • The neural network must produce rhythmic motor nerve output in the complete absence of rhythmic sensory afferent feedback.
  • The phase relationships, burst durations, and frequency modulation of this fictive output must qualitatively match the kinematic profiles of intact locomotion.
  • The network must be able to sustain rhythmicity in response to tonic, non-rhythmic descending or pharmacological drives, demonstrating that the temporal patterning is endogenous to the circuit.

These demonstrations decisively shifted the locus of primary pattern generation from the periphery to the central nervous system, establishing that sensory input serves a modulatory rather than an obligate generative role.

2.3 Conceptual Architecture of the Modern Central Pattern Generator

With the physical reality of the CPG confirmed, Grillner and subsequent motor systems neurobiologists addressed the organizational architecture of these circuits. The modern conceptual framework divides the central pattern generator into two functionally distinct, hierarchically interacting network tiers: the Rhythm Generator (RG) and the Pattern Formation (PF) network.

The rhythm generator tier is responsible for the fundamental clock-like timing of the motor behavior. It dictates the overall cycle period, frequency, and initiation of the rhythmic cadence. The rhythm generator neurons project downstream to the pattern formation network, a specialized computational layer that distributes, sculpts, and coordinates this rhythmic drive across distinct motor pools. The pattern formation network determines which specific muscles fire, the exact duration and amplitude of their bursts, and the subtle phase delays required across multiple joints (e.g., knee, ankle, hip) to generate efficient propulsion.

A critical feature of this dual-layer architecture is its flexibility. Rather than executing a rigid, immutable motor tape, the CPG maintains a balance between structural stereotypy and functional adaptability. Sensory feedback from load receptors, muscle spindles, and cutaneous mechanoreceptors acts directly upon both the rhythm-generating and pattern-forming tiers. Feedback can access the rhythm generator to reset the phase of the locomotor clock (e.g., when an unexpected obstacle prematurely terminates a swing phase), or it can selectively modulate the pattern-forming tier to alter the amplitude of an individual muscle contraction without disrupting the overarching timing cadence.

3. The Lamprey Spinal Model: Experimental Architecture and In Vitro Dissection

The transition from theoretical formulation to cellular-level validation required an experimental preparation that brought the microcircuitry of the vertebrate spinal cord into sharp focus. Sten Grillner’s deployment of the lamprey spinal cord yielded unprecedented mechanistic access, bridging the gap between whole-animal kinematics and synaptic physiology.

3.1 Anatomical Advantages of the Lamprey Spinal Cord

The sea lamprey possesses a structural simplicity that makes its spinal cord uniquely suited for cellular neurobiology. Unlike the cylindrical, heavily vascularized, and opaque spinal cord of mammals, the lamprey spinal cord is flattened dorsoventrally, resembling a thin, ribbon-like structure. Its thickness does not exceed 300 micrometers, allowing complete metabolic sustenance via passive diffusion when submerged in an artificial cerebrospinal fluid bath. The tissue maintains physiological intracellular resting potentials, low internal sodium concentrations, and robust synaptic transmission for days in vitro at temperatures between 8°C and 12°C.

A further structural advantage is the complete absence of myelin. In mammalian spinal cords, dense myelin sheaths scatter light, making direct visualization of neuronal cell bodies impossible without destructive tissue slicing. In the intact, isolated lamprey spinal cord, individual living interneurons, sensory edge cells, and motor neurons can be directly visualized through Differential Interference Contrast (DIC) and infrared microscopy. Neurobiologists can visually guide recording microelectrodes to specific neuronal somata residing within their native, unsevered network architecture.

Furthermore, the neuronal density in the lamprey spinal cord is several orders of magnitude lower than that of tetrapods. Rather than confronting millions of heterogeneous neurons, the lamprey spinal cord contains approximately 1,000 to 2,000 neurons per millimeter of spinal length. These cells can be grouped into a small number of morphologically and neurochemically distinct classes. This architectural reductionism enabled Grillner and his team to construct a comprehensive catalog of every major interneuronal subtype participating in locomotor pattern generation.

3.2 Pharmacological Induction of Fictive Swimming

To engage the locomotor circuitry of an isolated lamprey spinal cord in the absence of the brain, Grillner pioneered the use of neuropharmacological activation. In the intact animal, descending glutamatergic axons from the brainstem reticulospinal system release excitatory amino acids onto spinal circuits to initiate swimming. Grillner replicated this descending excitatory drive by bath-applying glutamate receptor agonists directly to the isolated spinal preparation.

Application of excitatory amino acids—specifically N-methyl-D-aspartate (NMDA), alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), and kainate—reliably induces stable, long-lasting fictive swimming. The preparation discharges rhythmic, alternating bursts of action potentials recorded from bilateral ventral roots that can persist uninterrupted for hours. Significantly, NMDA receptor activation produces remarkably stable, slow-to-moderate locomotor rhythms (0.1 to 3 Hz), whereas AMPA and kainate activation drive higher-frequency swimming (up to 8 to 10 Hz), effectively spanning the complete behavioral dynamic range of the living animal.

Grillner’s team conducted split-bath experiments using multi-compartment recording chambers sealed with Vaseline barriers. By selectively applying NMDA to isolated segments of the spinal cord while keeping adjacent segments in normal saline, they proved that even a short section consisting of only two or three spinal segments possesses the autonomous capacity to generate rhythmic, alternating locomotor bursts. This demonstrated that the lamprey spinal cord is not governed by a single centralized pacemaker; instead, it consists of a longitudinal chain of coupled, autonomous segmental oscillators, each capable of generating local rhythmicity.

3.3 Validation of Fictive Locomotion Against Intact Kinematics

A critical component of Grillner’s methodology was validating that the fictive motor output recorded in a Petri dish faithfully reproduced the biomechanics of intact, freely swimming lampreys. To establish this fidelity, Grillner, along with Wallén and other colleagues, conducted rigorous high-speed cinematographic and electromyographic (EMG) analyses of freely swimming animals, subsequently comparing these measurements directly with extracellular ventral root recordings from in vitro preparations.

In the freely swimming lamprey, propulsion is achieved through lateral undulatory waves that travel down the body from head to tail. This undulation requires two fundamental coordination parameters:

  • Bilateral alternation: At any given axial segment, the muscle contractions on the left side alternate with those on the right side, establishing an exact phase difference of 0.5 (180 degrees out of phase).
  • Rostrocaudal phase lag: A constant, fractional phase delay exists along the length of the body, amounting to approximately 1% of the total swimming cycle period per segment. This ensures that exactly one full mechanical sine wave is present along the body during swimming, regardless of swimming speed.

Quantitative comparisons confirmed that the in vitro fictive locomotion preparation reproduced both features with mathematical precision. The ventral roots discharged in strict left-right alternation, and multi-electrode recordings along the rostrocaudal axis demonstrated a stable, descending phase delay of approximately 1% per segment. Furthermore, just as in the living animal, where swimming speed is increased by reducing the total cycle period while keeping burst duration proportionally constant, the fictive motor burst duration scaled linearly with cycle period over a wide range of frequencies. This confirmed that the isolated spinal cord contains the complete neural apparatus required for natural locomotion.

4. Cellular and Synaptic Microcircuitry of the Segmental Oscillator

Having established that isolated spinal segments can generate autonomous rhythmic motor output, Grillner embarked on an exhaustive campaign to map the underlying cellular wiring diagram. Through thousands of targeted microelectrode impalements, his laboratory identified the essential interneuronal populations and their precise synaptic interconnections.

4.1 Identification and Morphology of Key Segmental Neurons

The segmental locomotor network of the lamprey is composed of four principal neuronal classes: Motor Neurons (MNs), Excitatory Interneurons (EINs), Crossed Caudal Inhibitory Interneurons (CCINs), and Lateral Interneurons (LINs). Each class possesses a unique morphological signature, dendritic topology, and neurochemical profile that corresponds to its functional role within the rhythm-generating circuit.

Excitatory Interneurons (EINs) are small-to-medium-sized multipolar neurons with locally projecting, uncrossed (ipsilateral) axons. These axons extend both rostrally and caudally over a few spinal segments, branching profusely within the lateral and ventral motor columns. Neurochemically, EINs are glutamatergic, releasing L-glutamate to excite their postsynaptic targets via both NMDA and AMPA/kainate receptors. They serve as the primary rhythm-generating engine and internal driver of the segmental network.

Crossed Caudal Inhibitory Interneurons (CCINs) are prominent, visually identifiable cells whose axons cross the ventral midline (commissure) and project caudally over multiple segments (ranging from 2 to 20 segments). CCINs are glycinergic; their primary role is to transmit reciprocal inhibition to the contralateral side of the spinal cord, enforcing strict left-right alternation. Lateral Interneurons (LINs), by contrast, are large cells with ipsilaterally descending axons that release glycine, delivering feedback inhibition to ipsilateral interneurons. Finally, Motor Neurons (MNs) reside in the ventral horn, extending elaborate dendrites into the lateral funiculus and projecting their axons out through the ventral roots to innervate the myotomal swimming musculature with acetylcholine.

4.2 The Synaptic Wiring Diagram of the Core Segmental Network

By executing paired intracellular recordings, Grillner and his colleagues reconstructed the synaptic connectivity matrix linking these cell types into a functional segmental oscillator. The circuit operates via a combination of local recurrent excitation and reciprocal commissural inhibition.

On any given side of a spinal segment (e.g., the left side), the network is energized by the EIN population. Crucially, EINs form mutual, recurrent excitatory synapses with one another. When an initial excitatory stimulus depolarizes a subset of EINs, this recurrent positive feedback rapidly recruits and synchronizes the entire local EIN pool. The EINs, in turn, provide direct, monosynaptic glutamatergic excitation to the ipsilateral Motor Neurons, driving the active motor burst, as well as to the ipsilateral CCINs.

The activation of the left CCINs triggers a rapid shutdown of contralateral activity. The axons of the left CCINs cross the midline and establish direct, monosynaptic glycinergic inhibitory synapses onto contralateral EINs, contralateral CCINs, and contralateral MNs. This architecture ensures that while the left side is actively discharging action potentials, the right side is held in a hyperpolarized, quiescent state. The duration of this ipsilateral burst is constrained by intrinsic burst-termination mechanisms (detailed in Section 5). As the left EINs terminate their firing, the contralateral network escapes from CCIN-mediated inhibition, depolarizes via its own recurrent EIN connections, and initiates the opposing half-cycle, activating right-sided CCINs that cross back to silence the left network.

4.3 Functional Verification of Circuit Connectivity

To definitively validate this proposed microcircuit architecture, Grillner’s team conducted dual and triple intracellular recordings. By impaling an EIN and a target Motor Neuron simultaneously, they demonstrated that an action potential in a single EIN generates a robust, short-latency Excitatory Postsynaptic Potential (EPSP) in the motor neuron. Pharmacological dissection revealed that this EPSP is biphasic: it contains a rapid, fast-decaying component mediated by AMPA receptors and a prolonged, voltage-dependent component mediated by NMDA receptors.

Similarly, paired impalements between CCINs and contralateral target neurons confirmed the existence of short-latency, monosynaptic Inhibitory Postsynaptic Potentials (IPSPs). These IPSPs were abolished by bath application of the glycine receptor antagonist strychnine, confirming that glycine is the primary neurotransmitter mediating reciprocal inhibition. The reversal potential of these IPSPs shifted systematically in response to intracellular chloride ion injections, corroborating classic chloride-dependent postsynaptic inhibition.

The most decisive proof of the network’s functional wiring came from pharmacological disinhibition experiments. When strychnine was added to a bath-applied NMDA solution during active fictive swimming, reciprocal inhibition was abolished. Under these conditions, the preparation did not cease to oscillate; instead, the left-right alternating motor burst pattern converted into a synchronous, bilateral bursting rhythm. Both sides of the spinal cord discharged high-amplitude bursts simultaneously at the same frequency. This demonstrated that each half of a spinal segment possesses its own autonomous rhythm generator capable of generating oscillatory bursts, and that the commissural CCIN pathway functions primarily to phase-lock these two hemicords into 180-degree reciprocal alternation.

5. Intrinsic Biophysical Properties: Ion Channels and Oscillatory Burst Dynamics

A seminal contribution of Sten Grillner was demonstrating that central pattern generation cannot be explained by synaptic connectivity alone. Instead, network function emerges from the non-linear interaction between specific synaptic wiring topologies and the intrinsic biophysical properties of the constituent neurons, governed by voltage-gated and ligand-gated ion channels.

5.1 NMDA Receptor-Mediated Membrane Potential Oscillations

One of Grillner’s most significant discoveries was that the NMDA receptor is not merely an excitatory synaptic conduit, but an active biophysical oscillator when operating within the vertebrate spinal cord. When the isolated lamprey spinal cord is treated with tetrodotoxin (TTX)—a potent neurotoxin that blocks voltage-gated sodium channels and abolishes all action potentials and synaptic transmission—the addition of NMDA continues to induce rhythmic, slow oscillations of the membrane potential in isolated spinal neurons.

These TTX-resistant pacemaker-like oscillations depend upon the unique biophysical kinetics of the NMDA receptor channel, specifically its voltage-dependent block by extracellular magnesium ions ($Mg^{2+}$). At hyperpolarized resting membrane potentials (around -65 to -70 mV), extracellular $Mg^{2+}$ ions enter the outer pore of the NMDA channel and become lodged, electrostatically obstructing the flow of other cations. Consequently, the channel conductance is low.

However, when the neuron experiences a slight depolarizing shift, the positive intracellular potential repels the divalent magnesium ion back into the extracellular space. This $Mg^{2+}$ unblock creates a region of negative-slope conductance in the current-voltage relationship of the neuronal membrane. As $Mg^{2+}$ leaves the channel pore, sodium ($Na^+$) and calcium ($Ca^{2+}$) ions flow inward down their electrochemical gradients, generating a depolarizing current that drives further unblocking. This positive feedback loop produces a regenerative, all-or-none upward phase transition, shifting the membrane potential into a sustained plateau depolarization state during which active burst discharge occurs.

5.2 Calcium-Dependent Potassium Conductances and Burst Termination

Because the NMDA-mediated depolarization is self-reinforcing, the network requires a powerful, intrinsic burst-termination mechanism to bring the membrane potential back down to the hyperpolarized state, ending the active motor burst. Grillner and his colleagues revealed that this termination mechanism is driven by intracellular calcium dynamics coupled to calcium-activated potassium conductances.

During the depolarized plateau phase of the burst, calcium ions continuously flood into the soma and dendrites through two distinct pathways: directly through the open pore of the unblocked NMDA receptor channels, and through voltage-gated L-type calcium channels activated by the sustained depolarization. Over the course of several hundred milliseconds, the concentration of free $Ca^{2+}$ in the sub-membrane cytoplasmic space rises progressively.

This localized accumulation of intracellular $Ca^{2+}$ binds to and activates small-conductance calcium-activated potassium channels (commonly referred to as $K_{Ca}$ or SK channels). The opening of these channels causes an efflux of potassium ions ($K^+$) down their concentration gradient, generating an outward hyperpolarizing current. As this outward $K^+$ current steadily increases, it eventually surpasses the inward NMDA- and $Ca^{2+}$-mediated depolarizing currents. The membrane potential begins to drop, allowing $Mg^{2+}$ ions to re-enter and block the NMDA channel pores. This precipitates an abrupt, regenerative deactivation phase, returning the neuron to a hyperpolarized resting state.

Grillner’s team experimentally confirmed this mechanism by applying apamin, a selective pharmacological blocker of SK channels. When apamin was applied to the spinal preparation during fictive swimming, the burst-termination mechanism was severely impaired: the duration of individual locomotor bursts expanded dramatically, the burst frequency slowed, and the regularity of the swimming rhythm deteriorated. Intracellular calcium buffering using BAPTA microinjections produced identical effects, cementing the role of $Ca^{2+}$-dependent $K^+$ currents as the master regulators of burst termination.

5.3 Low-Threshold and Hyperpolarization-Activated Currents

The hyperpolarizing phase transition that concludes a burst initiates a cascade of subthreshold conductances that prime the neuron for its subsequent burst, establishing a self-sustaining oscillatory loop. Two specific intrinsic currents are central to this phase transition: the low-voltage-activated T-type calcium current ($I_T$) and the hyperpolarization-activated cation current ($I_h$).

When the membrane hyperpolarizes—driven by $K_{Ca}$ conductances and reinforced by contralateral CCIN-mediated glycinergic IPSPs—voltage-gated T-type calcium channels, which were inactivated during the preceding plateau depolarization, recover from inactivation (de-inactivation). Simultaneously, the hyperpolarized membrane potential activates the non-selective cation current $I_h$. The slow, inward flow of cations through $I_h$ channels prevents excessive hyperpolarization and produces a gentle, upward depolarizing ramp (the pacemaker potential).

As the inhibitory synaptic inputs from the contralateral side decline and the membrane depolarizes toward threshold, the de-inactivated T-type calcium channels activate, driving a transient, low-threshold calcium spike. This physiological phenomenon, termed post-inhibitory rebound, accelerates the depolarizing trajectory of the neuron, carrying the membrane potential past the threshold required to expel $Mg^{2+}$ from the NMDA channels. Thus, the neuron transitions seamlessly back into the depolarized burst phase. The synthesis of these intrinsic ion channel kinetics with the synaptic wiring of the network creates an oscillator that is stable yet responsive to external modulation.

6. Intersegmental Coordination and Propagating Metachronal Waves

Locomotion in the lamprey requires more than the rhythmic bilateral alternation of individual spinal segments; it demands the seamless propagation of a mechanical bending wave down the length of the body. Sten Grillner addressed how a linear chain of approximately one hundred individual segmental oscillators coordinates its activity to generate this propagating wave.

6.1 The Phase-Lag Phenomenon in Undulatory Propulsion

To propel itself through water, an undulatory swimmer must pass a transverse mechanical wave along its body in a rostrocaudal direction at a speed slightly greater than the forward velocity of the animal. This mechanical wave is driven by a underlying neural wave of muscular activation. When Grillner and his team measured the phase relationship between distant ventral roots along the spinal cord, they discovered the constant fractional phase-lag phenomenon.

The total electrical delay from the first spinal segment to the final caudal segment corresponds to approximately 100% of a cycle period, meaning that exactly one full wavelength ($lambda$) is represented along the neuroaxis during steady swimming. Remarkably, as the animal changes its swimming speed—and thus changes its cycle period from several seconds down to 100 milliseconds—the phase lag per segment remains constant at approximately 1% of the cycle period ($\Phi \approx 0.01$).

Because the fractional phase lag is scale-invariant with respect to frequency, the spatial wavelength of the undulatory wave remains constant across all swimming speeds. From a hydrodynamic perspective, this scale invariance is essential: if the wavelength shifted dramatically with speed, the lamprey would experience turbulence and loss of propulsive thrust. Grillner demonstrated that this constant phase relationship is generated by intrinsic spinal mechanisms, as it persists intact in completely isolated spinal cord preparations devoid of all sensory feedback and brainstem inputs.

6.2 Longitudinal Axonal Projections and Intersegmental Coupling

The physical substrate for this intersegmental coordination lies in the longitudinal axonal projections of the CPG interneurons themselves. Morphological tract-tracing experiments revealed that both Excitatory Interneurons (EINs) and Crossed Inhibitory Interneurons (CCINs) send collaterals that extend along the spinal cord over varying segmental distances.

Importantly, these intersegmental projections exhibit a structural asymmetry. Ascending axons typically project over shorter distances (2 to 4 segments) with dense synaptic arborizations, whereas descending axons extend over longer distances (4 to 10 segments, and up to 20 segments for some CCINs). This structural polarity introduces an asymmetry in the synaptic coupling strength between adjacent segmental oscillators. Grillner formulated the hypothesis that this descending bias in synaptic coupling ensures that the more rostral oscillators exert a stronger entraining influence on their caudal neighbors than vice versa, encouraging the rostrocaudal propagation of the motor wave.

To evaluate the intrinsic oscillatory properties of different spinal regions, Grillner’s group conducted split-bath experiments where distinct sections of the spinal cord were exposed to different concentrations of NMDA. These studies demonstrated that the intrinsic natural frequency of rostral segments is slightly higher than that of caudal segments when exposed to identical levels of unpatterned chemical drive. This intrinsic rostrocaudal excitability gradient, combined with asymmetrical intersegmental coupling, locks the network into a descending phase sequence.

6.3 Mathematical and Biophysical Models of Intersegmental Coordination

To rigorously understand how a chain of coupled non-linear oscillators produces stable phase lags, Sten Grillner collaborated with mathematicians and theoretical neurobiologists, notably Nancy Kopell and G. Bard Ermentrout. Together, they applied coupled non-linear oscillator theory to the lamprey spinal cord, formalizing the biophysical constraints that govern metachronal wave propagation.

The Kopell-Ermentrout-Grillner models demonstrated that a chain of oscillators with equal intrinsic frequencies and symmetric bidirectional coupling will settle into a synchronous state (zero phase lag) or an anti-phase state, neither of which supports undulatory swimming. To generate a stable non-zero phase lag, the system requires either:

  • An intrinsic frequency gradient along the chain (e.g., rostral oscillators possessing a higher uncoupled frequency than caudal oscillators), or
  • An asymmetry in the coupling functions (e.g., descending synaptic projections transmitting a different phase-resetting torque than ascending projections).

The mathematical models generated an unexpected, testable prediction: if the excitability gradient along the spinal cord were experimentally reversed—by bathing caudal segments in a high concentration of NMDA while bathing rostral segments in a lower concentration—the direction of wave propagation should flip. Grillner’s laboratory performed this experiment and confirmed the mathematical prediction: the isolated spinal cord switched to backward fictive swimming, propagating a metachronal wave from tail to head. This confirmed that the direction and wavelength of the locomotor wave are dynamic, emergent properties of intersegmental coupling parameters rather than hard-wired unidirectional pathways.

7. Supraspinal Control: Descending Brainstem Pathways and Locomotor Initiation

While the spinal cord contains the complete microcircuitry required to form the locomotor pattern, it does not act autonomously in the living animal. Under physiological conditions, locomotion is initiated, steered, and terminated by descending commands originating in the brainstem. Sten Grillner’s investigations unraveled the conserved brainstem pathways that transform high-level motor intent into spinal rhythmicity.

7.1 The Mesencephalic Locomotor Region (MLR) and Diencephalic Locomotor Region (DLR)

In the mid-twentieth century, Russian physiologists Shik, Severin, and Orlovsky identified a localized region in the cat midbrain—the Mesencephalic Locomotor Region (MLR)—whose low-intensity electrical stimulation elicited walking on a treadmill. Grillner demonstrated that this command center is an evolutionarily ancient structure present in the lamprey brainstem, alongside an adjacent control center, the Diencephalic Locomotor Region (DLR).

When Grillner and his colleagues delivered focal electrical or chemical microinjections of glutamate into the lamprey MLR, the animal initiated coordinated swimming. The most remarkable feature of MLR stimulation is its graded control: delivering a low-intensity, tonic electrical stimulus to the MLR produces slow swimming; as the stimulus intensity is increased, the swimming frequency increases proportionally, transitioning through the animal’s entire dynamic speed range. The MLR does not specify the timing of individual muscle contractions; instead, it issues an unpatterned, tonic scalar command signal that encodes the desired vigor or velocity of locomotion.

Subsequent comparative studies proved that the MLR is present in all vertebrate classes, including cyclostomes, teleost fish, amphibians, reptiles, birds, and mammals. Neurochemically, the MLR consists of a collection of glutamatergic and cholinergic neurons situated in the pedunculopontine and cuneiform nuclei. It serves as the common vertebrate engine for initiating locomotor output, translating motivational decisions into motor drive.

7.2 The Reticulospinal System as the Primary Executive Conduit

The command signals generated in the MLR and DLR do not project directly to the spinal cord. Instead, they synapse within the hindbrain reticular formation, which houses the Reticulospinal (RS) system. In the lamprey, the reticulospinal system represents the primary descending executive pathway, equivalent in many functional respects to the mammalian corticospinal and rubrospinal tracts combined.

The lamprey reticulospinal system comprises four major bilateral nuclear groups: the mesencephalic, anterior, middle, and posterior reticular nuclei. These groups contain visually prominent, stereotypically positioned giant command neurons known as Müller cells. These reticulospinal neurons extend thick, unmyelinated axons down the entire length of the spinal cord, coursing through the ventral and lateral fiber tracts.

Grillner’s laboratory proved that reticulospinal axons release L-glutamate directly onto the spinal CPG elements, forming monosynaptic dual-component (AMPA and NMDA) excitatory connections onto both Excitatory Interneurons (EINs) and Motor Neurons. Reticulospinal recruitment follows an orderly size principle: during slow swimming, only small, low-threshold reticulospinal neurons discharge; as the required swimming speed rises, progressively larger reticulospinal cells, including the giant Müller cells, are recruited into the active ensemble. While bilateral, symmetrical activation of reticulospinal neurons drives straight forward swimming, unilateral stimulation of the reticulospinal system induces asymmetric spinal burst amplitudes, triggering lateral steering maneuvers and turning movements.

7.3 Sensory-Motor Transformations in Supraspinal Structures

The reticulospinal system does not serve as a passive relay station; it integrates multi-sensory inputs to adjust descending motor commands in real time. Sten Grillner and his collaborators elucidated the complex sensory-motor transformations that occur within these brainstem nuclei, focusing on visual, vestibular, and tectal pathways.

During swimming, the lamprey must maintain its dorsal-side-up orientation in the water column. The vestibular system, via the octavolateralis nuclei, transmits detailed gravitational and angular acceleration signals directly to the reticulospinal neurons. If the animal rolls to the left, the left vestibular apparatus detects the tilt and monosynaptically excites the ipsilateral reticulospinal cells while disynaptically inhibiting the contralateral reticulospinal cells. This asymmetric reticulospinal drive alters the bilateral amplitude of the spinal CPG output, rotating the animal back to horizontal equilibrium.

Similarly, the optic tectum (the homolog of the mammalian superior colliculus) projects heavily to the reticulospinal system. Visual detection of an approaching predator or a targeted prey item evokes a topographical burst of activity within the tectal map. Tectal efferents directly excite specific sub-ensembles of reticulospinal neurons, triggering either an explosive, large-amplitude escape turn (C-start response) or targeted orienting behavior. Grillner showed that the reticulospinal system serves as an integrative computational hub where visual, vestibular, and intrinsic motor programs converge before descending to the spinal cord.

8. The Basal Ganglia-Brainstem Axis: Upstream Command and Locomotor Selection

For locomotion to be adaptive, an animal must select appropriate behaviors while suppressing competing, disadvantageous motor outputs. In the late 1990s and 2000s, Grillner’s research expanded rostrally into the forebrain, leading to the profound discovery that the core microcircuit architecture of the vertebrate basal ganglia was already fully operational in the lamprey, serving as the master upstream gatekeeper of brainstem locomotor centers.

8.1 Evolutionary Blueprint of the Basal Ganglia in Primitive Vertebrates

Prior to Grillner’s comparative investigations, many neuroscientists considered the basal ganglia to be a specialized forebrain innovation of advanced amniotes (reptiles, birds, and mammals). Grillner challenged this view by executing a comprehensive immunohistochemical, tract-tracing, and electrophysiological dissection of the lamprey telencephalon, demonstrating that every fundamental component of the mammalian basal ganglia possesses a direct structural and functional homolog in the cyclostome brain.

The lamprey forebrain contains a distinct striatum that receives topographically organized glutamatergic inputs from both the pallium (the evolutionary precursor to the cerebral cortex) and the thalamus. The striatum is populated by GABAergic medium spiny neurons (MSNs) that express dopamine receptors. Furthermore, Grillner identified a localized dopaminergic nucleus in the lamprey mesencephalon that corresponds precisely to the mammalian Substantia Nigra pars compacta (SNc), sending dense dopaminergic projections to the striatum.

Downstream of the striatum, Grillner identified the functional homologs of the Globus Pallidus interna (GPi) and the Substantia Nigra pars reticulata (SNr). These output nuclei consist of GABAergic projection neurons that exert powerful, continuous, tonic inhibitory control over downstream motor targets in the MLR and reticulospinal system. Thus, the fundamental wiring blueprint of the vertebrate basal ganglia—comprising striatum, globus pallidus, substantia nigra, and their interconnecting direct and indirect pathways—has been conserved for over 500 million years.

8.2 Action Selection and the Disinhibition Paradigm of Movement Initiation

Grillner established that the initiation of locomotion by the basal ganglia operates through a conserved physiological mechanism known as disinhibition. Under basal, resting conditions, the GABAergic neurons of the globus pallidus homolog maintain high rates of spontaneous, tonic firing (20 to 50 Hz). Because these neurons project directly to the Mesencephalic Locomotor Region and the reticulospinal nuclei, they hold the brainstem locomotor networks in a state of continuous synaptic suppression, preventing unwanted motor discharge.

When the animal decides to initiate swimming—driven by sensory cues or internal motivational states—specific subsets of striatal projection neurons are activated by excitatory inputs from the pallium and thalamus. These striatal neurons project directly to the globus pallidus, releasing GABA and substance P. The activation of this direct pathway transiently silences the spontaneously active pallidal neurons.

By silencing these inhibitory pallidal cells, the tonic brake on the MLR is momentarily lifted. Released from inhibition, the MLR neurons depolarize, enter an active discharge state, and excite the downstream reticulospinal neurons, which in turn activate the spinal CPGs to initiate swimming. Grillner’s group demonstrated this disinhibitory control directly: microinjecting GABA receptor antagonists into the lamprey MLR immediately triggered spontaneous, coordinated swimming by chemically breaking the pallidal inhibitory clamp, mirroring the pathophysiology and therapeutic mechanisms observed in mammalian basal ganglia circuits.

8.3 Sensory and Motivational Integration in Movement Gating

The basal ganglia-brainstem axis functions as an integrative switchboard, filtering and gating behavioral selection based on sensory context and motivational salience. Grillner’s laboratory traced how olfactory and limbic information gains direct access to this action-selection machinery.

In the lamprey, olfaction represents a primary sensory modality for locating food, detecting migratory pheromones, and sensing alarm cues. Grillner showed that the olfactory bulb projects not only to the pallium but sends direct, monosynaptic connections into the striatum and the habenular complex. Strong olfactory stimulation triggers immediate, selective disinhibition of the MLR through this pathway, providing a neural circuit substrate for odor-evoked locomotor tracking.

Furthermore, dopamine plays a critical role in setting the threshold for locomotor initiation within the lamprey striatum. Just as in mammals, lamprey striatal neurons express two distinct subclasses of dopamine receptors: $D_1$ receptors (which enhance excitability and promote movement via the direct pathway) and $D_2$ receptors (which suppress excitability in the indirect pathway). Microinjections of dopamine or $D_1$ receptor agonists into the lamprey striatum lower the threshold of stimulation required to elicit locomotion, whereas $D_2$ antagonists inhibit motor initiation. This demonstrates that the neuromodulatory architecture underlying motivation, reward, and motor readiness was fully established at the dawn of vertebrate evolution.

9. Neuromodulation: Dynamic Tuning of Rhythm, Phase, and Network Output

The hard-wired synaptic connections of the central pattern generator define the boundaries of what the network can do, but the dynamic tuning of its output is governed by neuromodulators. Sten Grillner’s laboratory performed seminal work revealing how monoamines, neuropeptides, and purines reconfigure CPG microcircuit dynamics, enabling a single anatomical circuit to produce a diverse repertoire of behavioral outputs.

9.1 Serotonergic Modulation of Locomotor Burst Parameters

The most extensively characterized neuromodulatory system in the lamprey spinal cord is the endogenous serotonergic (5-HT) system. Unlike mammals, where spinal serotonin is supplied predominantly by descending projections from the medullary raphe nuclei, the lamprey spinal cord contains an intrinsic, dense plexus of serotonergic neurons and varicosities located immediately beneath the central canal, termed the periependymal 5-HT plexus.

Grillner and his colleagues discovered that bath application of serotonin to an isolated spinal cord undergoing fictive locomotion produces marked alterations in the motor burst profile:

  • It significantly reduces the overall locomotor frequency, producing slower, more deliberate swimming.
  • It prolongs the duration of the individual motor bursts relative to the cycle period.
  • It markedly sharpens the burst boundaries, enhancing the amplitude and regularity of the motor discharge while preventing erratic, stray action potentials.

Through cellular and biophysical analyses, Grillner proved that serotonin exerts these effects by directly down-regulating the small-conductance calcium-activated potassium channels (SK channels) responsible for burst termination. Serotonin activates 5-$HT_{1A}$-like receptors on spinal motor neurons and interneurons, initiating an intracellular signaling cascade that reduces the amplitude of the slow calcium-dependent afterhyperpolarization (sAHP). By suppressing the sAHP, serotonin prevents premature hyperpolarization, allowing neurons to fire more action potentials during the active phase of the burst. This demonstrates how a neuromodulator can fine-tune macroscopic motor performance by modifying the biophysical gating of a single ion channel species.

9.2 Peptidergic and Purinergic Modulatory Systems

In addition to classical monoamines, the lamprey spinal network is rich in neuropeptides and purines that are co-stored and co-released with conventional neurotransmitters. Grillner’s laboratory conducted extensive investigations into the modulatory roles of tachykinins (Substance P), somatostatin, and adenosine.

Substance P is co-localized with serotonin within the fibers of the periependymal plexus. When released during high-frequency neuronal discharge, Substance P exerts an effect complementary to that of serotonin: it potentiates the responses of NMDA receptor channels through an intracellular protein kinase C (PKC) pathway. This enhances the plateau depolarization of CPG neurons, increasing burst frequency and boosting the total power of the motor output. The co-release of serotonin and tachykinin allows the spinal cord to balance bursting stability against frequency modulation during high-velocity swimming.

Conversely, purinergic modulation provides a negative-feedback, homeostatic brake on network activity. During sustained swimming, active spinal neurons release adenosine triphosphate (ATP), which is rapidly hydrolyzed in the extracellular space by ecto-nucleotidases into adenosine. Grillner demonstrated that adenosine binds to presynaptic $A_1$ receptors, selectively reducing high-voltage-activated calcium currents ($N$– and $P/Q$-type) and depressing glutamate release at CPG synapses. This negative feedback loop prevents excitotoxic over-activation and provides a metabolic throttle, lowering locomotor frequency during prolonged motor engagement.

9.3 Dopaminergic and GABAergic Presynaptic Modulation

Neuromodulation within the spinal CPG operates not only through postsynaptic alterations of ion channels but also via targeted presynaptic inhibition at axon terminals. Sten Grillner’s team revealed that dopamine and GABA form specialized presynaptic regulatory checkpoints that control the gain of both descending commands and sensory inputs.

Descending reticulospinal axons, as they enter the spinal gray matter, express presynaptic dopamine $D_2$ receptors. Local release of dopamine within the spinal cord acts on these $D_2$ receptors to decrease the amount of glutamate released from reticulospinal terminals onto spinal EINs. This mechanism allows the spinal network to modulate its sensitivity to descending commands dynamically, dampening descending drive during specific behavioral phases.

Similarly, primary sensory afferents entering the dorsal spinal cord are subject to continuous presynaptic inhibition mediated by both $GABA_A$ and $GABA_B$ receptors. GABAergic interneurons form axo-axonic synapses directly onto sensory terminals. When activated, these inputs depolarize the sensory terminal slightly (primary afferent depolarization), inactivating voltage-gated sodium channels and shunting incoming action potentials. This presynaptic gate ensures that during powerful, self-generated swimming movements, the CPG can selectively silence irrelevant or disruptive sensory noise while preserving sensitivity to critical external mechanical perturbations.

10. Sensory Feedback Integration: Edge Cells, Proprioception, and Phase Resetting

A fundamental misconception regarding central pattern generators is that because they can operate autonomously in a dish, sensory feedback is irrelevant in the intact animal. Sten Grillner’s work disproved this assumption, demonstrating that the real-world execution of locomotion depends upon a continuous, bi-directional dialogue between the central pattern generator and peripheral proprioceptive feedback.

10.1 Identification and Physiology of Intraspinal Mechanoreceptors

In most vertebrates, proprioceptive feedback is supplied by specialized peripheral receptors, such as muscle spindles, Golgi tendon organs, and joint capsules. In the lamprey, Grillner and his colleagues discovered a novel class of primary mechanoreceptors embedded directly within the tissue of the spinal cord itself: the edge cells.

Edge cells are specialized mechanoreceptive neurons positioned along the extreme lateral margins of the flat spinal cord, immediately adjacent to the meninx. They extend long, branched dendritic arborizations that interlace with the structural connective tissue fibers at the cord boundary. When the lamprey’s body bends laterally during swimming, the spinal cord undergoes mechanical deformation: the convex side of the bend is subjected to longitudinal stretch, while the concave side experiences compression.

Grillner’s team showed that edge cells are specialized stretch receptors that fire action potentials in response to this lateral mechanical displacement. Edge cells can be divided into two functional classes:

  • Excitatory edge cells: release L-glutamate onto ipsilateral spinal CPG neurons and motor neurons, providing local reinforcement of ongoing motor activity.
  • Inhibitory edge cells: release glycine, projecting across the midline or to ipsilateral antagonistic centers to terminate ongoing contractions and facilitate transition to the opposite phase.

Because these mechanoreceptors reside directly within the central nervous system, they possess exceptionally short conduction latencies, providing instantaneous feedback regarding the physical curvature of the body.

10.2 Mechanisms of Entrainment and Phase Resetting

To demonstrate how mechanical sensory feedback couples to the central neural clock, Grillner designed mechanical entrainment experiments. The isolated lamprey spinal cord was mounted with its rostral end fixed and its caudal end attached to a mechanical motor that applied rhythmic lateral bending movements to the tissue, mimicking the passive undulations the body experiences when moving through water.

Remarkably, when the spinal cord was mechanically oscillated at frequencies close to its endogenous fictive swimming frequency, the internal neural rhythm shifted to match the mechanical movement precisely—a process termed entrainment. The ventral roots discharged at the exact frequency of the mechanical movement, locking into a fixed phase relationship with the mechanical displacement. If the mechanical driver was accelerated or decelerated, the central neural pattern generator adjusted its timing, following the mechanical rhythm over a wide frequency band.

Furthermore, Grillner performed transient phase-resetting experiments. If a sudden, brief mechanical stretch was applied to one side of the spinal cord during the ipsilateral burst phase, the ongoing burst was prolonged. Conversely, if the same mechanical stretch was delivered during the contralateral burst phase, the active burst was aborted prematurely, and the network was immediately reset to the ipsilateral phase. This established that proprioceptive feedback does not merely drive reflex contractions; it acts directly on the central clock machinery, adjusting the phase and cycle period of the rhythm generator to match environmental mechanics.

10.3 Dynamic Compensation for External Perturbations

The functional utility of this sensory-CPG coupling was demonstrated through experiments examining how animals navigate unpredictable fluid environments. In nature, a swimming lamprey encounters water currents, varying viscosities, and physical obstacles that introduce discrepancies between the motor command issued by the nervous system and the actual physical displacement achieved by the body.

Grillner demonstrated that if a swimming lamprey is physically restrained mid-body or forced to swim through a high-viscosity methylcellulose solution, the edge cell feedback automatically compensates for the mechanical load. The delayed mechanical bending delays the activation of the inhibitory edge cells, prolonging the duration of the ongoing muscle burst. As a result, the animal increases its muscular force output until the resistance is overcome and the mechanical wave can propagate.

Moreover, Grillner showed that the nervous system selectively gates proprioceptive inputs during different locomotor states. During resting states or slow swimming, sensory feedback exerts an entraining influence; during escape swimming, descending reticulospinal drive activates presynaptic inhibitory mechanisms that attenuate sensory feedback, allowing the central feedforward command to run unhindered. This work definitively closed the historical debate between Sherrington and Brown: the motor system is neither a pure reflex chain nor a closed central loop; it is a hybrid feedforward-feedback control architecture.

11. Computational Neuroscience, Bio-Robotics, and Quantitative Circuit Modeling

A distinctive hallmark of Sten Grillner’s career was his early recognition that biological systems of this complexity cannot be fully understood through reductionist biological dissection alone. To verify that the identified cellular conductances and synaptic connections were quantitatively sufficient to account for the macroscopic behavior of the swimming animal, Grillner initiated ambitious computational modeling and bio-robotic programs.

11.1 Biophysical and Kinetic Modeling of the Lamprey Circuit

Beginning in the late 1980s, Grillner collaborated with computational neuroscientists, such as Örjan Ekeberg and Jeanette Hellgren Kotaleski, to develop multi-level biophysical models of the lamprey spinal network. Rather than relying on abstract, connectionist neural networks, these simulations were constructed from the bottom up, utilizing Hodgkin-Huxley style mathematical differential equations to model the specific ion channels identified in their physiological experiments.

Each simulated neuron included multi-compartmental dendritic trees, explicit somatic and axonal compartments, and detailed mathematical kinetics for:

  • Voltage-gated sodium channels ($Na_V$),
  • Delayed rectifier potassium channels ($K_V$),
  • Low-voltage-activated T-type calcium channels ($Ca_T$),
  • High-voltage-activated L-type calcium channels ($Ca_L$),
  • Small-conductance calcium-activated potassium channels ($K_{Ca}$ / SK), and
  • Ligand-gated NMDA and AMPA/kainate channels, including the mathematical formulation of voltage-dependent $Mg^{2+}$ block kinetics.

These computational models succeeded in reproducing the behavior of the real biological network. When a non-specific, tonic excitatory current was injected into the simulated network, it spontaneously generated stable, alternating burst activity across bilateral neuron pools. The burst frequency scaled with the magnitude of the tonic drive, the phase lag along the simulated intersegmental chain stabilized at 1% per segment, and the apamin-sensitive burst termination kinetics matched empirical recordings. These simulations confirmed that Grillner’s biological catalog was mathematically sufficient to generate vertebrate locomotion.

11.2 Neuro-Mechanical Closed-Loop Simulations

Grillner’s theoretical work expanded beyond isolated neural models to embrace neuro-mechanical modeling. He recognized that natural selection does not operate on neural firing patterns in isolation; it operates on the mechanical movements of the physical body interacting with a physical environment. Collaborating with Ekeberg, Grillner developed a closed-loop virtual lamprey simulation.

In this virtual model, the biophysical neural network was coupled directly to a simulated multi-segmented biomechanical body submerged in a simulated fluid environment. The simulated motor neuron action potentials were converted via an excitation-contraction muscle model into dynamic muscular tension. This muscular tension acted upon elastic body segments, producing lateral bending movements governed by Navier-Stokes hydrodynamic equations that calculated the resulting forward propulsion and lateral drag forces.

Crucially, the physical displacements calculated by the hydrodynamic model were fed back into simulated edge cells, closing the loop between body, brain, and water. This comprehensive simulation revealed that smooth, energy-efficient swimming requires a precise matching between the internal mechanical resonance of the body tissues and the firing frequency of the neural CPG. The model demonstrated that the constant 1% fractional phase lag maximizes hydroelastic efficiency, confirming that the central pattern generator has evolved to exploit the physical properties of the aquatic medium.

11.3 Bio-Inspired Robotics: Embodying the Grillner Architecture

The ultimate physical test of Grillner’s central pattern generator paradigm was realized through bio-inspired robotics. Working alongside roboticists, most notably Auke Ijspeert at the École Polytechnique Fédérale de Lausanne (EPFL), Grillner’s circuit architectures were embedded into physical, amphibious robots.

These bio-robots, such as Lampetra and Salamandra Robotica, featured modular mechanical segments actuated by electric motors, with each segment controlled by an onboard microcontroller programmed with the mathematical equations of Grillner’s segmental CPG. By sending a single scalar electrical signal corresponding to descending reticulospinal drive, the engineers could drive the robot forward. Increasing the signal intensity accelerated the mechanical swimming frequency; modulating the drive asymmetrically between the left and right sides caused the robot to execute smooth turns.

Remarkably, the Salamandra Robotica platform demonstrated how an undulatory swimming CPG could be extended, through evolutionary duplication and specialization of motor modules, to control multi-jointed limbs for walking. When low tonic drive was applied, the robotic salamander exhibited slow, interlimb-coordinated walking; when the drive crossed a threshold, the limb controllers saturated, and the axial CPG transitioned into a high-frequency undulatory swimming wave. These physical embodiments confirmed that the decentralized, modular CPG architecture discovered by Grillner offers unmatched robustness, mechanical fault-tolerance, and energy efficiency for autonomous mobile systems.

12. Evolutionary Conservation and Translational Relevance: From Lamprey to Mammalian Locomotion

The ultimate significance of Sten Grillner’s investigations extends far beyond the biology of the sea lamprey. Grillner demonstrated that the organizational principles deciphered in this jawless cyclostome represent the ancestral, conserved blueprint for motor control across all vertebrates, providing a conceptual and clinical bridge directly to human neurophysiology.

12.1 Comparative Homologies Across Vertebrate Locomotor Networks

During the 2000s and 2010s, modern molecular genetics and developmental neurobiology validated Grillner’s evolutionary hypotheses. Researchers using genetically modified mice—pioneered by laboratories such as those of Martyn Goulding and Ole Kiehn—identified specific classes of spinal interneurons defined by the expression of developmental transcription factors (V0, V1, V2a, V2b, and V3 interneurons).

These mammalian interneuronal lineages match the functional, physiological, and morphological phenotypes identified by Grillner in the lamprey:

  • The V0 interneurons are commissural inhibitory and excitatory interneurons that cross the midline to coordinate left-right alternation, serving as the mammalian homolog of Grillner’s lamprey CCINs. When V0 interneurons are genetically ablated in mice, the animals lose their alternating walk and switch to a synchronous, hopping gait.
  • The V2a interneurons are ipsilaterally projecting glutamatergic cells that provide the primary excitatory drive to the motor pools, functioning as the mammalian counterparts to the lamprey EINs.
  • The V1 interneurons (including Renshaw cells and Ia inhibitory interneurons) provide fast, local inhibitory feedback, mirroring the functional profile of lamprey LINs.

This correspondence demonstrates that evolution did not reinvent the spinal motor engine when animals transitioned from swimming in ancient oceans to walking on land; it repurposed, expanded, and elaborated the ancestral CPG circuitry discovered by Grillner.

12.2 Evidence for Central Pattern Generators in Primates and Humans

For decades, many clinical neurologists maintained that while primitive vertebrates and lower mammals possess autonomous spinal CPGs, human motor control had become entirely encephalized, relying exclusively on direct descending cortical drive from the corticospinal tract. Sten Grillner challenged this view, arguing that while humans exhibit expanded cortical control for dexterous fine-motor manipulation of the hands, our basic rhythmic locomotor movements continue to be driven by spinal central pattern generators.

Compelling evidence for human spinal CPGs can be observed in neonatal physiology. Human newborns, when supported upright over a moving surface, execute spontaneous, coordinated, alternating stepping movements—the stepping reflex. Because cortical myelination is largely incomplete at this stage of human development, and because this rhythmic stepping can be elicited even in anencephalic infants who lack cerebral hemispheres, these rhythmic motor patterns must be generated by intrinsic, pre-wired spinal networks.

Furthermore, Grillner’s conceptual framework provided the theoretical foundation for contemporary rehabilitation protocols in individuals with complete spinal cord injuries. When a paralyzed patient is suspended in an unweighting harness over a motorized treadmill, the combination of tonic sensory load, treadmill movement, and descending neuromodulatory or electrical stimulation can awaken the dormant lumbar spinal cord. Under these conditions, the isolated human lumbar cord can generate rhythmic, coordinated, reciprocal EMG bursting patterns in leg muscles without any descending input from the brain.

12.3 Clinical Applications and Future Directions in Neuromodulation

Sten Grillner’s mechanistic dissection of central pattern generators has inspired modern therapeutic paradigms aimed at restoring functional movement following spinal cord injury (SCI) and neurodegenerative disease. Translating the principles of CPG pharmacology and electrophysiology has moved motor neuro-rehabilitation from passive physical therapy to targeted neuromodulatory restoration.

A prominent realization of this work is the development of targeted epidural electrical stimulation (EES) of the spinal cord, pioneered clinically by researchers such as Grégoire Courtine and Susan Harkema. Rather than delivering indiscriminate electrical currents, modern EES arrays deliver spatiotemporally patterned electrical pulses to specific dorsal root entry zones of the lumbar and sacral spinal cord. These stimulation protocols are timed to coincide with the precise phase requirements of the underlying CPG, restoring the permissive excitability required for remaining, damaged supraspinal commands to recruit the spinal rhythm-generating networks.

Simultaneously, pharmacological strategies directly derived from Grillner’s modulatory studies—specifically the systemic or intrathecal administration of 5-HT, dopamine, and alpha-2 adrenergic receptor agonists—are used to prime spinal circuits, lowering the threshold for locomotor initiation. Looking to the future, the integration of brain-computer interfaces (BCIs) with biomimetic spinal stimulators offers the promise of fully closed-loop neuroprosthetics: cortical intention decoded from the motor cortex is routed around a damaged spinal lesion to drive bio-inspired stimulation patterns delivered directly to the human locomotor central pattern generator. The entire architecture of these restorative neurotechnologies rests on the foundational principles deciphered by Sten Grillner in a humble aquatic vertebrate.

Conclusion

The scientific trajectory of Sten Grillner represents a masterclass in the power of comparative, reductionist neurobiology. By challenging classical reflex dogmas and dedicating decades to the rigorous dissection of the sea lamprey nervous system, Grillner resolved fundamental questions regarding how the vertebrate brain and spinal cord initiate, pattern, modulate, and stabilize rhythmic movement. His work dismantled false dichotomies, demonstrating that motor execution emerges from the seamless integration of intrinsic biophysical channel kinetics, modular synaptic circuit topologies, ascending and descending supraspinal command loops, and closed-loop sensory feedback.

Today, the central pattern generator stands not as a theoretical hypothesis, but as an empirical reality whose organizational architecture is conserved from the simplest jawless fishes to primates. Sten Grillner’s legacy extends across neurobiology, computational neuroscience, autonomous robotics, and clinical neuro-rehabilitation. As neuroscientists continue to map the human motor connectome and develop neuroprosthetic technologies to reverse paralysis, the foundational principles deciphered across his Central Pattern Generator experiments will remain the guiding light for understanding the neural symphony of movement.

References

★

Rate This Content

0.0 / 5 • 0 votes

Cite This Article

memjavad (2026, September 12). The Central Pattern Generator Experiments – Sten Grillner. PSYCHOLOGICAL DATABASE. https://en.arabpsychology.com/experiments/central-pattern-generator-experiments-sten-grillner/
memjavad. “The Central Pattern Generator Experiments – Sten Grillner.” PSYCHOLOGICAL DATABASE, 12 September 2026, https://en.arabpsychology.com/experiments/central-pattern-generator-experiments-sten-grillner/.
memjavad. “The Central Pattern Generator Experiments – Sten Grillner.” PSYCHOLOGICAL DATABASE. September 12, 2026. https://en.arabpsychology.com/experiments/central-pattern-generator-experiments-sten-grillner/.