A hemicord locomotor network of excitatory interneurons: a simulation study.

Locomotor burst generation is simulated using a full-scale network model of the unilateral excitatory interneuronal population. Earlier small-scale models predicted that a population of excitatory neurons would be sufficient to produce burst activity, and this has recently been experimentally confirmed. Here we simulate the hemicord activity induced under various experimental conditions, including pharmacological activation by NMDA and AMPA as well as electrical stimulation. The model network comprises a realistic number of cells and synaptic connectivity patterns. Using similar distributions of cellular and synaptic parameters, as have been estimated experimentally, a large variation in dynamic characteristics like firing rates, burst, and cycle durations were seen in single cells. On the network level an overall rhythm was generated because the synaptic interactions cause partial synchronization within the population. This network rhythm not only emerged despite the distributed cellular parameters but relied on this variability, in particular, in reproducing variations of the activity during the cycle and showing recruitment in interneuronal populations. A slow rhythm (0.4-2 Hz) can be induced by tonic activation of NMDA-sensitive channels, which are voltage dependent and generate depolarizing plateaus. The rhythm emerges through a synchronization of bursts of the individual neurons. A fast rhythm (4-12 Hz), induced by AMPA, relies on spike synchronization within the population, and each burst is composed of single spikes produced by different neurons. The dynamic range of the fast rhythm is limited by the ability of the network to synchronize oscillations and depends on the strength of synaptic connections and the duration of the slow after hyperpolarization. The model network also produces prolonged bouts of rhythmic activity in response to brief electrical activations, as seen experimentally. The mutual excitation can sustain long-lasting activity for a realistic set of synaptic parameters. The bout duration depends on the strength of excitatory synaptic connections, the level of persistent depolarization, and the influx of Ca(2+) ions and activation of Ca(2+)-dependent K(+) current.

A working memory model based on fast Hebbian learning.

Recent models of the oculomotor delayed response task have been based on the assumption that working memory is stored as a persistent activity state (a 'bump' state). The delay activity is maintained by a finely tuned synaptic weight matrix producing a line attractor. Here we present an alternative hypothesis, that fast Hebbian synaptic plasticity is the mechanism underlying working memory. A computational model demonstrates a working memory function that is more resistant to distractors and network inhomogeneity compared to previous models, and that is also capable of storing multiple memories.

A Bayesian attractor network with incremental learning.

A realtime online learning system with capacity limits needs to gradually forget old information in order to avoid catastrophic forgetting. This can be achieved by allowing new information to overwrite old, as in a so-called palimpsest memory. This paper describes an incremental learning rule based on the Bayesian confidence propagation neural network that has palimpsest properties when employed in an attractor neural network. The network does not suffer from catastrophic forgetting, has a capacity dependent on the learning time constant and exhibits faster convergence for newer patterns.

Modeling of substance P and 5-HT induced synaptic plasticity in the lamprey spinal CPG: consequences for network pattern generation.

Consequences of synaptic plasticity in the lamprey spinal CPG are analyzed by means of simulations. This is motivated by the effects substance P (a tachykinin) and serotonin (5-hydroxytryptamin; 5-HT) have on synaptic transmission in the locomotor network. Activity-dependent synaptic depression and potentiation have recently been shown experimentally using paired intracellular recordings. Although normally activity-dependent plasticity presumably does not contribute to the patterning of network activity, this changes in the presence of the neuromodulators substance P and 5-HT, which evoke significant plasticity. Substance P can induce a faster and larger depression of inhibitory connections but potentiation of excitatory inputs, whereas 5-HT induces facilitation of both inhibitory and excitatory inputs. Changes in the amplitude of the first postsynaptic potential are also seen. These changes could thus be a potential mechanism underlying the modulatory role these substances have on the rhythmic network activity. The aim of the present study has been to implement the activity dependent synaptic depression and facilitation induced by substance P and 5-HT into two alternative models of the lamprey spinal locomotor network, one relying on reciprocal inhibition for bursting and one in which each hemicord is capable of oscillations. The consequences of the plasticity of inhibitory and excitatory connections are then explored on the network level. In the intact spinal cord, tachykinins and 5-HT, which can be endogenously released, increase and decrease the frequency of the alternating left-right burst pattern, respectively. The frequency decreasing effect of 5-HT has previously been explained based on its conductance decreasing effect on K(Ca) underlying the postspike afterhyperpolarization (AHP). The present simulations show that short-term synaptic plasticity may have strong effects on frequency regulation in the lamprey spinal CPG. In the network model relying on reciprocal inhibition, the observed effects substance P and 5-HT have on network behavior (i.e., a frequency increase and decrease respectively) can to a substantial part be explained by their effects on the total extent and time dynamics of synaptic depression and facilitation. The cellular effects of these substances will in the 5-HT case further contribute to its network effect.

Glucose-induced [Ca2+]i abnormalities in human pancreatic islets: important role of overstimulation.

Chronic hyperglycemia desensitizes beta-cells to glucose. To further define the mechanisms behind desensitization and the role of overstimulation, we tested human pancreatic islets for the effects of long-term elevated glucose levels on cytoplasmic free Ca2+ concentration ([Ca2+]i) and its relationship to overstimulation. Islets were cultured for 48 h with 5.5 or 27 mmol/l glucose. Culture with 27 mmol/l glucose obliterated postculture insulin responses to 27 mmol/l glucose. This desensitization was specific for glucose versus arginine. Desensitization was accompanied by three major [Ca2+]i abnormalities: 1) elevated basal [Ca2+]i, 2) loss of a glucose-induced rise in [Ca2+]i, and 3) perturbations of oscillatory activity with a decrease in glucose-induced slow oscillations (0.2-0.5 min(-1)). Coculture with 0.3 mmol/l diazoxide was performed to probe the role of overstimulation. Neither glucose nor diazoxide affected islet glucose utilization or oxidation. Coculture with diazoxide and 27 mmol/l glucose significantly (P < 0.05) restored postculture insulin responses to glucose and lowered basal [Ca2+]i and normalized glucose-induced oscillatory activity. However, diazoxide completely failed to revive an increase in [Ca2+]i during postculture glucose stimulation. In conclusion, desensitization of glucose-induced insulin secretion in human pancreatic islets is induced in parallel with major glucose-specific [Ca2+]i abnormalities. Overstimulation is an important but not exclusive factor behind [Ca2+]i abnormalities.

Neural mechanisms potentially contributing to the intersegmental phase lag in lamprey.II. Hemisegmental oscillations produced by mutually coupled excitatory neurons.

Most previous models of the spinal central pattern generator (CPG) underlying locomotion in the lamprey have relied on reciprocal inhibition between the left and right side for oscillations to be produced. Here, we have explored the consequences of using self-oscillatory hemisegments. Within a single hemisegment, the oscillations are produced by a network of recurrently coupled excitatory neurons (E neurons) that by themselves are not oscillatory but when coupled together through N-methyl-d-aspartate (NMDA) and alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionicacid (AMPA)/kainate transmission can produce oscillations. The bursting mechanism relies on intracellular accumulation of calcium that activates Ca(2+)-dependent K(+). The intracellular calcium is modeled by two different intracellular calcium pools, one of which represents the calcium entry following the action potential, Ca(AP) pool, and the other represents the calcium inflow through the NMDA channels, Ca(NMDA) pool. The Ca(2+)-dependent K(+) activated by these two calcium pools are referred to as K(CaAP) and K(CaNMDA), respectively, and their relative conductances are modulated and increase with the background activation of the network. When changing the background stimulation, the bursting activity in this network can be made to cover a frequency range of 0.5-5.5 Hz with reasonable burst proportions if the adaptation is modulated with the activity. When a chain of such hemisegments are coupled together, a phase lag along the chain can be produced. The local oscillations as well as the phase lag is dependent on the axonal conduction delay as well as the types of excitatory coupling that are assumed, i.e. AMPA/kainate and/or NMDA. When the caudal excitatory projections are extended further than the rostral ones, and assumed to be of approximately equal strength, this kind of network is capable of reproducing several experimental observations such as those occurring during strychnine blockade of the left-right reciprocal inhibition. Addition of reciprocally coupled inhibitory neurons in such a network gives rise to antiphasic activity between the left and right side, but not necessarily to any change of the frequency if the burst proportion of the hemisegmental bursts is well below 50%. Prolongation of the C neuron projection in the rostrocaudal direction restricts the phase lag produced by only the excitatory hemisegmental network by locking together the interburst intervals at different levels of the spinal cord.

Neural mechanisms potentially contributing to the intersegmental phase lag in lamprey.I. Segmental oscillations dependent on reciprocal inhibition.

Factors contributing to the production of a phase lag along chains of oscillatory networks consisting of Hodgkin-Huxley type neurons are analyzed by means of simulations. Simplified network configurations are explored consisting of the basic building blocks of the spinal central pattern generator (CPG) generating swimming in the lamprey. It consists of reciprocally coupled crossed inhibitory C interneurons and ipsilateral excitatory E interneurons that activate C neurons and other E neurons. Oscillatory activity in the model network can, in the simplest case, be produced by a pair of reciprocally coupled C interneurons oscillating through an escape mechanism. Different levels of tonic excitation drive the network over a wide burst frequency range. In this type of network, powerful frequency-regulating factors are the effective inhibition produced by the active side, in combination with the tendency of the inactive side to escape from the inhibition. These two mechanisms can be affected by several factors, e.g. spike frequency adaptation (calcium-dependent K(+) channels), N-methyl-D-aspartate membrane properties as well as presence of low-voltage activated calcium channels. A rostrocaudal phase lag can be produced either by extending the contralateral inhibitory projections or the ipsilateral excitatory projections relatively more in the caudal than the rostral direction, since both an increased inhibition and a phasic excitation slow down the receiving network. The phase lag becomes decreased if the length of the intersegmental projections is increased or if the projections are extended symmetrically in both the rostral and the caudal directions. The simulations indicate that the conditions in the ends of an oscillator chain may significantly affect sign, magnitude and constancy of the phase lag. Also, with short and relatively weak intersegmental connections, the network remains robust against perturbations as well as intrinsic frequency differences along the chain. The phase lag (percentage of cycle duration) increases, however, with burst frequency also when the coupling strength is comparatively weak. The results are discussed and compared with previous "phase pulling" models as well as relaxation oscillators.

A model of cortical associative memory based on a horizontal network of connected columns.

An attractor network model of cortical associative memory functions has been constructed and simulated. By replacing the single cell as the functional unit by multiple cells in cortical columns connected by long-range fibers, the model is improved in terms of correspondence with cortical connectivity. The connectivity is improved, since the original dense and symmetric connectivity of a standard recurrent network becomes sparse and asymmetric at the cell-to-cell level. Our simulations show that this kind of network, with model neurons of the Hodgkin-Huxley type arranged in columns, can operate as an associative memory in much the same way as previous models having simpler connectivity. The network shows attractor-like behaviour and performs the standard assembly operations despite differences in the dynamics introduced by the more detailed cell model and network structure. Furthermore, the model has become sufficiently detailed to allow evaluation against electrophysiological and anatomical observations. For instance, cell activities comply with experimental findings and reaction times are within biological and psychological ranges. By introducing a scaling model we demonstrate that a network approaching experimentally reported neuron numbers and synaptic distributions also could work like the model studied here.

Activity-dependent modulation of adaptation produces a constant burst proportion in a model of the lamprey spinal locomotor generator.

The neuronal network underlying lamprey swimming has stimulated extensive modelling on different levels of abstraction. The lamprey swims with a burst frequency ranging from 0.3 to 8-10 Hz with a rostrocaudal lag between bursts in each segment along the spinal cord. The swimming motor pattern is characterized by a burst proportion that is independent of burst frequency and lasts around 30%-40% of the cycle duration. This also applies in preparations in which the reciprocal inhibition in the spinal cord between the left and right side is blocked. A network of coupled excitatory neurons producing hemisegmental oscillations may form the basis of the lamprey central pattern generator (CPG). Here we explored how such networks, in principle, could produce a large frequency range with a constant burst proportion. The computer simulations of the lamprey CPG use simplified, graded output units that could represent populations of neurons and that exhibit adaptation. We investigated the effect of an active modulation of the degree of adaptation of the CPG units to accomplish a constant burst proportion over the whole frequency range when, in addition, each hemisegment is assumed to be self-oscillatory. The degree of adaptation is increased with the degree of stimulation of the network. This will make the bursts terminate earlier at higher burst rates, allowing for a constant burst proportion. Without modulated adaptation the network operates in a limited range of swimming frequencies due to a progressive increase of burst duration with increasing background stimulation. By introducing a modulation of the adaptation, a broad burst frequency range can be produced. The reciprocal inhibition is thus not the primary burst terminating factor, as in many CPG models, and it is mainly responsible for producing alternation between the left and right sides. The results are compared with the Morris-Lecar oscillator model with parameters set to produce a type A and type B oscillator, in which the burst durations stay constant or increase, respectively, when the background stimulation is increased. Here as well, burst duration can be controlled by modulation of the slow variable in a similar way as above. When oscillatory hemisegmental networks are coupled together in a chain a phase lag is produced. The production of a phase lag in chains of such oscillators is compared with chains of Morris-Lecar relaxation oscillators. Models relating to the intact versus isolated spinal cord preparation are discussed, as well as the role of descending inhibition.

A Bayesian neural network method for adverse drug reaction signal generation.

OBJECTIVE: The database of adverse drug reactions (ADRs) held by the Uppsala Monitoring Centre on behalf of the 47 countries of the World Health Organization (WHO) Collaborating Programme for International Drug Monitoring contains nearly two million reports. It is the largest database of this sort in the world, and about 35,000 new reports are added quarterly. The task of trying to find new drug-ADR signals has been carried out by an expert panel, but with such a large volume of material the task is daunting. We have developed a flexible, automated procedure to find new signals with known probability difference from the background data. METHOD: Data mining, using various computational approaches, has been applied in a variety of disciplines. A Bayesian confidence propagation neural network (BCPNN) has been developed which can manage large data sets, is robust in handling incomplete data, and may be used with complex variables. Using information theory, such a tool is ideal for finding drug-ADR combinations with other variables, which are highly associated compared to the generality of the stored data, or a section of the stored data. The method is transparent for easy checking and flexible for different kinds of search. RESULTS: Using the BCPNN, some time scan examples are given which show the power of the technique to find signals early (captopril-coughing) and to avoid false positives where a common drug and ADRs occur in the database (digoxin-acne; digoxin-rash). A routine application of the BCPNN to a quarterly update is also tested, showing that 1004 suspected drug-ADR combinations reached the 97.5% confidence level of difference from the generality. Of these, 307 were potentially serious ADRs, and of these 53 related to new drugs. Twelve of the latter were not recorded in the CD editions of The physician's Desk Reference or Martindale's Extra Pharmacopoea and did not appear in Reactions Weekly online. CONCLUSION: The results indicate that the BCPNN can be used in the detection of significant signals from the data set of the WHO Programme on International Drug Monitoring. The BCPNN will be an extremely useful adjunct to the expert assessment of very large numbers of spontaneously reported ADRs.

Intrinsic function of a neuronal network - a vertebrate central pattern generator.

The cellular bases of vertebrate locomotor behaviour is reviewed using the lamprey as a model system. Forebrain and brainstem cell populations initiate locomotor activity via reticulospinal fibers activating a spinal network comprised of glutamatergic and glycinergic interneurons. The role of different subtypes of Ca2+ channels, Ca2+ dependent K+ channels and voltage dependent NMDA channels at the neuronal and network level is in focus as well as the effects of different metabotropic, aminergic and peptidergic modulators that target these ion channels. This is one of the few vertebrate networks that is understood at a cellular level.

Modulation of burst frequency by calcium-dependent potassium channels in the lamprey locomotor system: dependence of the activity level.

It is crucial to determine the effects on the network level of a modulation of intrinsic membrane properties. The role calcium-dependent potassium channels, KCa, in the lamprey locomotor system has been investigated extensively. Earlier experimental studies have shown that apamin, which affects one type of KCa, increases the cycle duration of the locomotor network, due to effects on the burst termination. The effects of apamin were here larger when the network had a low level of activity (burst frequency 0.5 to 1 Hz) as compared to a higher rate (> 2 Hz). By using a previously developed simulation model based on the lamprey locomotor network, we show that the model could account for the frequency dependence of the apamin modulation, if only the KCa conductance activated by Ca2+ entering during the action potential was altered and not the KCa conductance activated by Ca2+ entering through NMDA channels. The present simulation model of the spinal network in the lamprey can thus account for earlier experimental results with apamin on the network and cellular level that previously appeared enigmatic.

Modeling of the spinal neuronal circuitry underlying locomotion in a lower vertebrate.

The neural circuitry generating lamprey undulatory swimming is among the most accessible and best known of the vertebrate neuronal locomotor systems. It therefore serves as an experimental model for such systems. Modeling and computer simulation of this system was initiated at a point when a significant part of the network had been identified, although much detail was still lacking. The model has been further developed over 10 years in close interaction with experiments. The local burst generating circuitry is formed by ipsilateral excitatory neurons and crossed reciprocal inhibitory neurons. Early models also incorporated an off-switch lateral interneuron (L), the connectivity of which suggested it could contribute to burst termination at moderate to high bursting frequencies. Later examination of this model suggested, however, that the L interneuron was not of primary importance for burst termination, and this was later verified experimentally. Further, early models explained the effects of 5-HT on bursting frequency, spike frequency, and burst duration as being due to its modulatory action on the spike frequency adaptation of lamprey premotor interneurons. In current network models, accumulated adaptation is in addition the main burst terminating factor. Drive-related modulation of adaptation is explored as a mechanism for control of burst duration. This produces an adequate burst frequency range and a constant burst proportion within each cycle. It further allows for hemisegmental bursting, which has been observed experimentally. The local burst generator forms the basis of a network model of the distributed pattern generator that extends along the spinal cord. Phase constancy and flexibility of intersegmental coordination has been studied in such a simulated network. Current modeling work focuses on neuromodulator circuitry and action, network responses to input transients, how to model the intact versus an isolated piece of spinal cord, as well as on improving an earlier neuromechanical model of lamprey swimming.

Low-voltage-activated calcium channels in the lamprey locomotor network: simulation and experiment.

To evaluate the role of low-voltage-activated (LVA) calcium channels in the lamprey spinal locomotor network, a previous computer simulation model has been extended to include LVA calcium channels. It is also of interest to explore the consequences of a LVA conductance for the electrical behavior of the single neuron. The LVA calcium channel was modeled with voltage-dependent activation and inactivation using the m3h form, following a Hodgkin-Huxley paradigm. Experimental data from lamprey neurons was used to provide parameter values of the single cell model. The presence of a LVA calcium conductance in the model could account for the occurrence of a rebound depolarization in the simulation model. The influence of holding potential on the occurrence of a rebound as well the latency at which it is elicited was investigated and compared with previous experiments. The probability of a rebound increased at a more depolarized holding potential and the latency was also reduced under these conditions. Furthermore, the effect of changing the holding potential and the reversal potential of the calcium dependent potassium conductance were tested to determine under which conditions several rebound spikes could be elicited after a single inhibitory pulse in the simulation model. A reduction of the slow afterhyperpolarization (sAHP) after the action potential reduced the tendency for a train of rebound spikes. The experimental effects of gamma-aminobutyric acid-B (GABA(B)) receptor activation were simulated by reducing the maximal LVA calcium conductance. A reduced tendency for rebound firing and a slower rising phase with sinusoidal current stimulation was observed, in accordance with earlier experiments. The effect of reducing the slow afterhyperpolarization and reducing the LVA calcium current was tested experimentally in the lamprey spinal cord, during N-methyl-D-aspartate (NMDA)-induced fictive locomotion. The reduction of burst frequency was more pronounced with GABA(B) agonists than with apamin (inhibitor of K(Ca) current) when using high NMDA concentration (high burst frequency). The burst frequency increased after the addition of a LVA calcium current to the simulated segmental network, due to a faster recovery during the inhibitory phase as the activity switches between the sides. This result is consistent with earlier experimental findings because GABA(B) receptor agonists reduce the locomotor frequency. These results taken together suggest that the LVA calcium channels contribute to a larger degree with respect to the burst frequency regulation than the sAHP mechanism at higher burst frequencies. The range in which a regular burst pattern can be simulated is extended in the lower range by the addition of LVA calcium channels, which leads to more stable activity at low locomotor frequencies. We conclude that the present model can account for rebound firing and trains of rebound spikes in lamprey neurons. The effects of GABA(B) receptor activation on the network level is consistent with a reduction of the calcium current through LVA calcium channels even though GABA(B) receptor activation will affect the sAHP indirectly and also presynaptic inhibition.

A higher order Bayesian neural network with spiking units.

We treat a Bayesian confidence propagation neural network, primarily in a classifier context. The one-layer version of the network implements a naive Bayesian classifier, which requires the input attributes to be independent. This limitation is overcome by a higher order network. The higher order Bayesian neural network is evaluated on a real world task of diagnosing a telephone exchange computer. By introducing stochastic spiking units, and soft interval coding, it is also possible to handle uncertain as well as continuous valued inputs.

Neural networks that co-ordinate locomotion and body orientation in lamprey.

The networks of the brainstem and spinal cord that co-ordinate locomotion and body orientation in lamprey are described. The cycle-to-cycle pattern generation of these networks is produced by interacting glutamatergic and glycinergic neurones, with NMDA receptor-channels playing an important role at lower rates of locomotion. The fine tuning of the networks produced by 5-HT, dopamine and GABA systems involves a modulation of Ca2+-dependent K+ channels, high- and low-threshold voltage-activated Ca2+ channels and presynaptic inhibitory mechanisms. Mathematical modelling has been used to explore the capacity of these biological networks. The vestibular control of the body orientation during swimming is exerted via reticulospinal neurones located in different reticular nuclei. These neurones become activated maximally at different angles of tilt.

Neuronal network models of motor generation and control.

Computer simulation of neuronal networks is rapidly becoming accepted as a powerful tool in neuroscience. We illustrate the trends in this field by looking at motor generation and control, with examples from recent modeling studies of different systems, including the spinal swimming rhythm generator of the lamprey.

Computer models of the brain--how far can they take us?

The recent developments in computer capacity and algorithms, together with a tremendous growth of data in neuroscience have dramatically improved the possibilities of modeling and simulating certain brain structures and activities with a considerable degree of realism. Although there is still a long way to go, some claim that we will one day be able to create artificial "brains" with similar capacity to the human brain, perhaps even surpassing it. Here we focus on these perspectives, discussing the potentials and limitations of today's computer models, and how far they might be able to take us.

A flexible and fault tolerant query-reply system based on a Bayesian neural network.

A query-reply system based on a Bayesian neural network is described. Strategies for generating questions which make the system both efficient and highly fault tolerant are presented. This involves having one phase of question generation intended to quickly reach a hypothesis followed by a phase where verification of the hypothesis is attempted. In addition, both phases have strategies for detecting and removing inconsistencies in the replies from the user. Also described is an explanatory mechanism which gives information related to why a certain hypotheses is reached or question asked. Specific examples of the systems behavior as well as the results of a statistical evaluation are presented.

Computer simulations of NMDA and non-NMDA receptor-mediated synaptic drive: sensory and supraspinal modulation of neurons and small networks.

1. The segmental locomotor network in lamprey can generate the rhythmic burst pattern underlying locomotion when it is driven via synaptic glutamate receptors. Lower rates of activity can be evoked by activation of N-methyl-D-aspartate (NMDA) receptors, whereas a rapid activity can only be induced by non-NMDA receptors [kainate/alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)]. The reticulospinal and sensory inputs are known to act via both NMDA and non-NMDA receptors, but it is unclear how these inputs can provide an appropriate control of the locomotor rate. We have examined the effects of different types of excitatory synaptic input to neurons of the locomotor network with the use of a computer-simulated electrical neuron model, with Na+, K+, Ca(2+)-dependent K+ channels, and with inherent oscillatory properties linked to the NMDA conductance. Synapses were modeled as a modulated ionic conductance in the membrane of the postsynaptic cell comprising a voltage-dependent NMDA component (Na+, K+, Ca2+ conductances) of long duration, and/or a non-NMDA component (Na+, K+ conductance) of short duration. 2. By using two neurons to drive a postsynaptic cell with non-NMDA-type synapses, a continuous range of firing frequencies could be evoked in the postsynaptic cell, by altering the firing rate of the presynaptic cells. If a single presynaptic neuron was used, there was a tendency toward spike synchronization between the pre- and postsynaptic cells. 3. When a postsynaptic neuron was driven via NMDA synapses, an oscillatory burst activity could be evoked. The rate of the oscillations was, however, little affected by the presynaptic firing rate. When a drive neuron with mixed (NMDA and non-NMDA) synapses was used, the rate of the oscillations could be changed within a limited frequency range by altering the presynaptic firing rate. By adding another larger drive neuron, having a larger rheobase current and mixed synapses with smaller relative NMDA components, the frequency range of the postsynaptic oscillations could be markedly increased. The frequency range depended on the parameters selected for each of the two types of mixed synapses. 4. A small rhythm-generating neuronal network, comprising six cells connected as the principal interneurons of the lamprey spinal locomotor network, was used to test the role of a tonic NMDA and non-NMDA receptor activation to drive the network and produce bursting.(ABSTRACT TRUNCATED AT 400 WORDS)