Spike initiation by transmembrane current: a white-noise analysis

What causes a neuron to spike?

The computation performed by a neuron can be formulated as a combination of dimensional reduction in stimulus space and the nonlinearity inherent in a spiking output. White noise stimulus and reverse correlation (the spike-triggered average and spike-triggered covariance) are often used in experimental neuroscience to "ask" neurons which dimensions in stimulus space they are sensitive to and to characterize the nonlinearity of the response. In this article, we apply reverse correlation to the simplest model neuron with temporal dynamics-the leaky integrate-and-fire model-and find that for even this simple case, standard techniques do not recover the known neural computation. To overcome this, we develop novel reverse-correlation techniques by selectively analyzing only "isolated" spikes and taking explicit account of the extended silences that precede these isolated spikes. We discuss the implications of our methods to the characterization of neural adaptation. Although these methods are developed in the context of the leaky integrate-and-fire model, our findings are relevant for the analysis of spike trains from real neurons.

Sensitivity of neurons to weak electric fields

Weak electric fields modulate neuronal activity, and knowledge of the interaction threshold is important in the understanding of neuronal synchronization, in neural prosthetic design, and in the public health assessment of environmental extremely low frequency fields. Previous experimental measurements have placed the threshold between 1 and 5 mV/mm, although theory predicts that elongated neurons should have submillivolt per millimeter sensitivity near 100 microV/mm. We here provide the first experimental confirmation that neuronal networks are detectably sensitive to submillivolt per millimeter electrical fields [Gaussian pulses 26 msec full width at half-maximal, 140 microV/mm root mean square (rms), 295 microV/mm peak amplitude], an order of magnitude below previous findings, and further demonstrate that these networks are more sensitive than the average single neuron threshold (185 microV/mm rms, 394 microV/mm peak amplitude) to field modulation.

Sensitivity of neurons to weak electric fields

Weak electric fields modulate neuronal activity, and knowledge of the interaction threshold is important in the understanding of neuronal synchronization, in neural prosthetic design, and in the public health assessment of environmental extremely low frequency fields. Previous experimental measurements have placed the threshold between 1 and 5 mV/mm, although theory predicts that elongated neurons should have submillivolt per millimeter sensitivity near 100 microV/mm. We here provide the first experimental confirmation that neuronal networks are detectably sensitive to submillivolt per millimeter electrical fields [Gaussian pulses 26 msec full width at half-maximal, 140 microV/mm root mean square (rms), 295 microV/mm peak amplitude], an order of magnitude below previous findings, and further demonstrate that these networks are more sensitive than the average single neuron threshold (185 microV/mm rms, 394 microV/mm peak amplitude) to field modulation.

Spike generating dynamics and the conditions for spike-time precision in cortical neurons

Temporal precision of spiking response in cortical neurons has been a subject of intense debate. Using a canonical model of spike generation, we explore the conditions for precise and reliable spike timing in the presence of Gaussian white noise. In agreement with previous results we find that constant stimuli lead to imprecise timing, while aperiodic stimuli yield precise spike timing. Under constant stimulus the neuron is a noise perturbed oscillator, the spike times follow renewal statistics and are imprecise. Under an aperiodic stimulus sequence, the neuron acts as a threshold element; the firing times are precisely determined by the dynamics of the stimulus. We further study the dependence of spike-time precision on the input stimulus frequency and find a non-linear tuning whose width can be related to the locking modes of the neuron. We conclude that viewing the neuron as a non-linear oscillator is the key for understanding spike-time precision.

Amplitude and frequency dependence of spike timing: implications for dynamic regulation

The spike-time reliability of motoneurons in the Aplysia buccal motor ganglion was studied as a function of the frequency content and the relative amplitude of the fluctuations in the neuronal input, calculated as the coefficient of variation (CV). Measurements of spike-time reliability to sinusoidal and aperiodic inputs, as well as simulations of a noisy leaky integrate-and-fire neuron stimulated by spike trains drawn from a periodically modulated process, demonstrate that there are three qualitatively different CV-dependent mechanisms that determine reliability: noise-dominated (CV < 0.05 for Aplysia motoneurons) where spike timing is unreliable regardless of frequency content; resonance-dominated (CV approximately 0.05-0.25) where reliability is reduced by removal of input frequencies equal to motoneuron firing rate; and amplitude-dominated (CV >0.35) where reliability depends on input frequencies greater than motoneuron firing rate. In the resonance-dominated regime, changes in the activity of the presynaptic inhibitory interneuron B4/5 alter motoneuron spike-time reliability. The increases or decreases in reliability occur coincident with small changes in motoneuron spiking rate due to changes in interneuron activity. Injection of a hyperpolarizing current into the motoneuron reproduces the interneuron-induced changes in reliability. The rate-dependent changes in reliability can be understood from the phase-locking properties of regularly spiking motoneurons to periodic inputs. Our observations demonstrate that the ability of a neuron to support a spike-time code can be actively controlled by varying the properties of the neuron and its input.

Ergodicity of spike trains: when does trial averaging make sense?

Neuronal information processing is often studied on the basis of spiking patterns. The relevant statistics such as firing rates calculated with the peri-stimulus time histogram are obtained by averaging spiking patterns over many experimental runs. However, animals should respond to one experimental stimulation in real situations, and what is available to the brain is not the trial statistics but the population statistics. Consequently, physiological ergodicity, namely, the consistency between trial averaging and population averaging, is implicitly assumed in the data analyses, although it does not trivially hold true. In this letter, we investigate how characteristics of noisy neural network models, such as single neuron properties, external stimuli, and synaptic inputs, affect the statistics of firing patterns. In particular, we show that how high membrane potential sensitivity to input fluctuations, inability of neurons to remember past inputs, external stimuli with large variability and temporally separated peaks, and relatively few contributions of synaptic inputs result in spike trains that are reproducible over many trials. The reproducibility of spike trains and synchronous firing are contrasted and related to the ergodicity issue. Several numerical calculations with neural network examples are carried out to support the theoretical results.

The geometry of chaos synchronization

Chaos synchronization in coupled systems is often characterized by a map phi between the states of the components. In noninvertible systems, or in systems without inherent symmetries, the synchronization set--by which we mean graph(phi)--can be extremely complicated. We identify, describe, and give examples of several different complications that can arise, and we link each to inherent properties of the underlying dynamics. In brief, synchronization sets can in general become nondifferentiable, and in the more severe case of noninvertible dynamics, they might even be multivalued. We suggest two different ways to quantify these features, and we discuss possible failures in detecting chaos synchrony using standard continuity-based methods when these features are present.

Influence of subthreshold nonlinearities on signal-to-noise ratio and timing precision for small signals in neurons: minimal model analysis

Subthreshold voltage- and time-dependent conductances can subserve different roles in signal integration and action potential generation. Here, we use minimal models to demonstrate how a non-inactivating low-threshold outward current (I(KLT)) can enhance the precision of small-signal integration. Our integrate-and-fire models have only a few biophysical parameters, enabling a parametric study of I(KLT) effects. I(KLT) increases the signal-to-noise ratio (SNR) for firing when a subthreshold 'signal' EPSP is delivered in the presence of weak random input. The increased SNR is due to the suppression of spontaneous firings to random input. In accordance, SNR grows as the EPSP amplitude increases. SNR also grows as the unitary synaptic current's time constant increases, leading to more effective suppression of spontaneous activity. Spike-triggered reverse correlation of the injected current indicates that, to reach spike threshold, a cell with I(KLT) requires a briefer time course of injected current. Consistent with this narrowed integration time window, I(KI.T) enhances phase-locking. measured as vector strength, to a weak noisy and periodically modulated stimulus. Thus subthreshold negative feedback mediated by I(KLT) enhances temporal processing. An alternative suppression mechanism is voltage- and time-dependent inactivation of a low-threshold inward current. This feature in an integrate-and-fire model also shows SNR enhancement, in comparison with a case when the inward current is non-inactivating. Small-signal detection can be significantly improved in noisy neuronal systems by subthreshold negative feedback, serving to suppress false positives.

Enhancement of signal-to-noise ratio and phase locking for small inputs by a low-threshold outward current in auditory neurons

Neurons possess multiple voltage-dependent conductances specific for their function. To investigate how low-threshold outward currents improve the detection of small signals in a noisy background, we recorded from gerbil medial superior olivary (MSO) neurons in vitro. MSO neurons responded phasically, with a single spike to a step current injection. When bathed in dendrotoxin (DTX), most cells switched to tonic firing, suggesting that low-threshold potassium currents (I(KLT)) participated in shaping these phasic responses. Neurons were stimulated with a computer-generated steady barrage of random inputs, mimicking weak synaptic conductance transients (the "noise"), together with a larger but still subthreshold postsynaptic conductance, EPSG (the "signal"). DTX reduced the signal-to-noise ratio (SNR), defined as the ratio of probability to fire in response to the EPSG and the probability to fire spontaneously in response to noise. The reduction was mainly attributable to the increase of spontaneous firing in DTX. The spike-triggered reverse correlation indicated that, for spike generation, the neuron with I(KLT) required faster inward current transients. This narrow temporal integration window contributed to superior phase locking of firing to periodic stimuli before application of DTX. A computer model including Hodgkin-Huxley type conductances for spike generation and for I(KLT) (Rathouz and Trussell, 1998) showed similar response statistics. The dynamic low-threshold outward current increased SNR and the temporal precision of integration of weak subthreshold signals in auditory neurons by suppressing false positives.

An analysis of the reliability phenomenon in the FitzHugh-Nagumo model

The reliability of single neurons on realistic stimuli has been experimentally confirmed in a wide variety of animal preparations. We present a theoretical study of the reliability phenomenon in the FitzHugh-Nagumo model on white Gaussian stimulation. The analysis of the model's dynamics is performed in three regimes-the excitable, bistable, and oscillatory ones. We use tools from the random dynamical systems theory, such as the pullbacks and the estimation of the Lyapunov exponents and rotation number. The results show that for most stimulus intensities, trajectories converge to a single stochastic equilibrium point, and the leading Lyapunov exponent is negative. Consequently, in these regimes the discharge times are reliable in the sense that repeated presentation of the same aperiodic input segment evokes similar firing times after some transient time. Surprisingly, for a certain range of stimulus intensities, unreliable firing is observed due to the onset of stochastic chaos, as indicated by the estimated positive leading Lyapunov exponents. For this range of stimulus intensities, stochastic chaos occurs in the bistable regime and also expands in adjacent parts of the excitable and oscillating regimes. The obtained results are valuable in the explanation of experimental observations concerning the reliability of neurons stimulated with broad-band Gaussian inputs. They reveal two distinct neuronal response types. In the regime where the first Lyapunov has negative values, such inputs eventually lead neurons to reliable firing, and this suggests that any observed variance of firing times in reliability experiments is mainly due to internal noise. In the regime with positive Lyapunov exponents, the source of unreliable firing is stochastic chaos, a novel phenomenon in the reliability literature, whose origin and function need further investigation.

Enhancement of signal-to-noise ratio and phase locking for small inputs by a low-threshold outward current in auditory neurons

Neurons possess multiple voltage-dependent conductances specific for their function. To investigate how low-threshold outward currents improve the detection of small signals in a noisy background, we recorded from gerbil medial superior olivary (MSO) neurons in vitro. MSO neurons responded phasically, with a single spike to a step current injection. When bathed in dendrotoxin (DTX), most cells switched to tonic firing, suggesting that low-threshold potassium currents (I(KLT)) participated in shaping these phasic responses. Neurons were stimulated with a computer-generated steady barrage of random inputs, mimicking weak synaptic conductance transients (the "noise"), together with a larger but still subthreshold postsynaptic conductance, EPSG (the "signal"). DTX reduced the signal-to-noise ratio (SNR), defined as the ratio of probability to fire in response to the EPSG and the probability to fire spontaneously in response to noise. The reduction was mainly attributable to the increase of spontaneous firing in DTX. The spike-triggered reverse correlation indicated that, for spike generation, the neuron with I(KLT) required faster inward current transients. This narrow temporal integration window contributed to superior phase locking of firing to periodic stimuli before application of DTX. A computer model including Hodgkin-Huxley type conductances for spike generation and for I(KLT) (Rathouz and Trussell, 1998) showed similar response statistics. The dynamic low-threshold outward current increased SNR and the temporal precision of integration of weak subthreshold signals in auditory neurons by suppressing false positives.

Surfing a spike wave down the ventral stream

Numerous theories of neural processing, often motivated by experimental observations, have explored the computational properties of neural codes based on the absolute or relative timing of spikes in spike trains. Spiking neuron models and theories however, as well as their experimental counterparts, have generally been limited to the simulation or observation of isolated neurons, isolated spike trains, or reduced neural populations. Such theories would therefore seem inappropriate to capture the properties of a neural code relying on temporal spike patterns distributed across large neuronal populations. Here we report a range of computer simulations and theoretical considerations that were designed to explore the possibilities of one such code and its relevance for visual processing. In a unified framework where the relation between stimulus saliency and spike relative timing plays the central role, we describe how the ventral stream of the visual system could process natural input scenes and extract meaningful information, both rapidly and reliably. The first wave of spikes generated in the retina in response to a visual stimulation carries information explicitly in its spatio-temporal structure: the most salient information is represented by the first spikes over the population. This spike wave, propagating through a hierarchy of visual areas, is regenerated at each processing stage, where its temporal structure can be modified by (i). the selectivity of the cortical neurons, (ii). lateral interactions and (iii). top-down attentional influences from higher order cortical areas. The resulting model could account for the remarkable efficiency and rapidity of processing observed in the primate visual system.

Limits to the experimental detection of nonlinear synchrony

Chaos synchronization is often characterized by the existence of a continuous function between the states of the components. However, in coupled systems without inherent symmetries, the synchronization set can be extremely complicated. We describe and illustrate three typical complications that can arise, and we discuss how existing methods for detecting synchronization will be hampered by the presence of these features.

Synchronization of noise-induced bursts in noncoupled sensory neurons

We report experimental observation of phase synchronization in an array of nonidentical noncoupled noisy neuronal oscillators, due to stimulation with external noise. The synchronization derives from a noise-induced qualitative change in the firing pattern of single neurons, which changes from a quasiperiodic to a bursting mode. We show that at a certain noise intensity the onsets of bursts in different neurons become synchronized, even though the number of spikes inside the bursts may vary for different neurons. We demonstrate this effect both experimentally for the electroreceptor afferents of paddlefish, and numerically for a canonical phase model, and characterize it in terms of stochastic synchronization.

The reliability of the stochastic active rotator

The reliability of firing of excitable-oscillating systems is studied through the response of the active rotator, a neuronal model evolving on the unit circle, to white gaussian noise. A stochastic return map is introduced that captures the behavior of the model. This map has two fixed points: one stable and the other unstable. Iterates of all initial conditions except the unstable point tend to the stable fixed point for almost all input realizations. This means that to a given input realization, there corresponds a unique asymptotic response. In this way, repetitive stimulation with the same segment of noise realization evokes, after possibly a transient time, the same response in the active rotator. In other words, this model responds reliably to such inputs. It is argued that this results from the nonuniform motion of the active rotator around the unit circle and that similar results hold for other neuronal models whose dynamics can be approximated by phase dynamics similar to the active rotator.

Random dynamics of the Hodgkin-Huxley neuron model

Noise can alter the response of neurons, enhancing their ability to detect weak inputs. We analyze how the Hodgkin-Huxley equations, a canonical neuron model, respond to white noise stimulation. We show that this model possesses a stochastic attractor, reduced to a unique stochastic equilibrium point that attracts all trajectories.

Noise-enhanced neuronal reliability

This work shows that noise can enhance the discharge time reliability in Hodgkin-Huxley neuron models stimulated by weak periodic and aperiodic inputs. By expanding the Fokker-Planck equation of an elementary model for excitable systems, the dependence of the optimal noise intensity on input characteristics is discussed.

Action potential threshold of hippocampal pyramidal cells in vivo is increased by recent spiking activity

Understanding the mechanisms that influence the initiation of action potentials in single neurons is an important step in determining the way information is processed by neural networks. Therefore, we have investigated the properties of action potential thresholds for hippocampal neurons using in vivo intracellular recording methods in Sprague-Dawley rats. The use of in vivo recording has the advantage of the presence of naturally occurring spatio-temporal patterns of synaptic activity which lead to action potential initiation. We have found there is a large variability in the threshold voltage (5.7+/-1.7 mV; n=22) of individual action potentials. We have identified two separate factors that contribute to this variation in threshold: (1) fast rates of membrane potential change prior to the action potential are associated with more hyperpolarized thresholds (increased excitability) and (2) the occurrence of other action potentials in the 1 s prior to any given action potential is associated with more depolarized thresholds (decreased excitability). We suggest that prior action potentials cause sodium channel inactivation that recovers with approximately a 1-s time constant and thus depresses action potential threshold during this period.

Input-output functions of mammalian motoneurons

Our intent in this review was to consider the relationship between the biophysical properties of motoneurons and the mechanisms by which they transduce the synaptic inputs they receive into changes in their firing rates. Our emphasis has been on experimental results obtained over the past twenty years, which have shown that motoneurons are just as complex and interesting as other central neurons. This work has shown that motoneurons are endowed with a rich complement of active dendritic conductances, and flexible control of both somatic and dendritic channels by endogenous neuromodulators. Although this new information requires some revision of the simple view of motoneuron input-output properties that was prevalent in the early 1980's (see sections 2.3 and 2.10), the basic aspects of synaptic transduction by motoneurons can still be captured by a relatively simple input-output model (see section 2.3, equations 1-3). It remains valid to describe motoneuron recruitment as a product of the total synaptic current delivered to the soma, the effective input resistance of the motoneuron and the somatic voltage threshold for spike initiation (equations 1 and 2). However, because of the presence of active channels activated in the subthreshold range, both the delivery of synaptic current and the effective input resistance depend upon membrane potential. In addition, activation of metabotropic receptors by achetylcholine, glutamate, noradrenaline, serotonin, substance P and thyrotropin releasing factor (TRH) can alter the properties of various voltage- and calcium-sensitive channels and thereby affect synaptic current delivery and input resistance. Once motoneurons are activated, their steady-state rate of repetitive discharge is linearly related to the amount of injected or synaptic current reaching the soma (equation 3). However, the slope of this relation, the minimum discharge rate and the threshold current for repetitive discharge are all subject to neuromodulatory control. There are still a number of unresolved issues concerning the control of motoneuron discharge by synaptic inputs. Under dynamic conditions, when synaptic input is rapidly changing, time- and activity-dependent changes in the state of ionic channels will alter both synaptic current delivery to the spike-generating conductances and the relation between synaptic current and discharge rate. There is at present no general quantitative expression for motoneuron input-output properties under dynamic conditions. Even under steady-state conditions, the biophysical mechanisms underlying the transfer of synaptic current from the dendrites to the soma are not well understood, due to the paucity of direct recordings from motoneuron dendrites. It seems likely that resolving these important issues will keep motoneuron afficiandoes well occupied during the next twenty years.

Coherence resonance and discharge time reliability in neurons and neuronal models

Neurons are subject to internal and external noise that have been known to modify the way they process incoming signals. Recent studies have suggested that such alterations have functional roles and can also be used in biomedical applications. The present work goes over experimental and theoretical descriptions of the response of neurons to white noise stimulation. It examines various forms of noise related behavior in a standard neuronal model, namely the leaky integrate and fire. This clarifies the conditions under which specific noise induced changes occur in neurons, and consequently can help in determining whether nervous systems operate under similar circumstances.

The transient precision of integrate and fire neurons: effect of background activity and noise

We study the response of an integrate and fire neuron to a randomly timed step stimulus. We calculate the latency to the first spike after stimulus onset and its jitter. Background activity, seen in most neurons, reduces latency but causes substantial jitter in the response, indicating a tradeoff between timing precision and latency. The effect of intrinsic noise and synaptic noise on this tradeoff is studied. For synaptic noise we find that, unexpectedly, jitter does not increase for larger synaptic amplitudes, instead, jitter is practically independent of synaptic amplitude. Constant intrinsic noise interacts counterintuitively with latency and jitter, and depending on the stimulus strength, noise shifts the tradeoff in either direction.

Noise in neurons is message dependent

Neuronal responses are conspicuously variable. We focus on one particular aspect of that variability: the precision of action potential timing. We show that for common models of noisy spike generation, elementary considerations imply that such variability is a function of the input, and can be made arbitrarily large or small by a suitable choice of inputs. Our considerations are expected to extend to virtually any mechanism of spike generation, and we illustrate them with data from the visual pathway. Thus, a simplification usually made in the application of information theory to neural processing is violated: noise is not independent of the message. However, we also show the existence of error-correcting topologies, which can achieve better timing reliability than their components.

Channel noise in neurons

The probabilistic gating of voltage-dependent ion channels is a source of electrical 'channel noise' in neurons. This noise has long been implicated in limiting the reliability (repeatability) of neuronal responses to repeated presentations of identical stimuli. More recently, it has been shown to increase the range of spiking behaviors exhibited in some neural populations. Channel numbers are tied to metabolic efficiency and the stability of resting potential, and channel noise might be exploited by future cochlear implants in order to improve the temporal representation of sound.

Noise in integrate-and-fire neurons: from stochastic input to escape rates

We analyze the effect of noise in integrate-and-fire neurons driven by time-dependent input and compare the diffusion approximation for the membrane potential to escape noise. It is shown that for time-dependent subthreshold input, diffusive noise can be replaced by escape noise with a hazard function that has a gaussian dependence on the distance between the (noise-free) membrane voltage and threshold. The approximation is improved if we add to the hazard function a probability current proportional to the derivative of the voltage. Stochastic resonance in response to periodic input occurs in both noise models and exhibits similar characteristics.

Functional identification of the input-output transforms of mammalian motoneurones

We studied the responses of rat hypoglossal and cat lumbar motoneurones to a variety of excitatory and inhibitory injected current transients during repetitive discharge. The amplitudes and time courses of the transients were comparable to those of the synaptic currents underlying postsynaptic potentials (PSPs) recorded in these cells. Poisson trains of these current transients were combined with an additional independent, high frequency random waveform to approximate band-limited white noise. The composite, white noise waveform was then superimposed on long duration suprathreshold current steps. We used the responses of the motoneurones to the white noise stimulus to derive zero-, first- and second-order Wiener kernels, which provide a quantitative description of the relation between injected current and discharge probability. The convolution integral computed for an injected current waveform and the first-order Wiener kernel provides the best linear prediction of the associated peristimulus time histogram (PSTH). This linear model provided good matches to most of the PSTHs compiled between the times of occurrence of individual current transients and motoneurone discharges. However, for the largest amplitude current transients, a significant improvement in the PSTH match was often achieved by expanding the model to include the convolution of the second-order Wiener kernel with the input. The overall transformation of current inputs into firing rate could be approximated by a second-order Wiener Model, i.e., a cascade of a dynamic, linear filter followed by a static non-linearity. At a given mean firing rate, the non-linear component of the motoneurone's response could be described by the square of the linear component multiplied by a constant coefficient. The amplitude of the response of the linear component increased with the average firing rate, whereas the value of the multiplicative coefficient in the nonlinear component decreased. As a result, the overall transform could be predicted from the mean firing rate and the linear impulse response, yielding a relatively simple, general description of the motoneurone's input-output function.

Noisy spiking neurons and networks: useful approximations for firing probabilities and global behavior

Electrophysiological properties of spiking neurons receiving complex stimuli perturbed by noise are investigated. A semi-analytical estimate of firing probabilities and subthreshold behavior of the stochastic system can be made in terms of the solution of a purely deterministic system. The method comes from an approximation for the distribution function and moments of the underlying non linear multidimensional diffusion process. This so called moment method works for general conductance-based systems and an application is presented for the Hodgkin-Huxley neuronal model. Statistical properties obtained from the moment method are compared with direct numerical integration of the stochastic system. The firing probability due to external noise is derived as a closed formula. Results are given for different forms of the deterministic component of the stimulus. A generalization to neural networks of conductance-based systems with internal currents perturbed by noise can be obtained using the same approach. In the case of fully connected networks, a mean field population equation is derived which may be compared to Kuramoto's master equation for weakly coupled neural oscillators.

Efficient discrimination of temporal patterns by motion-sensitive neurons in primate visual cortex

Although motion-sensitive neurons in macaque middle temporal (MT) area are conventionally characterized using stimuli whose velocity remains constant for 1-3 s, many ecologically relevant stimuli change on a shorter time scale (30-300 ms). We compared neuronal responses to conventional (constant-velocity) and time-varying stimuli in alert primates. The responses to both stimulus ensembles were well described as rate-modulated Poisson processes but with very high precision (approximately 3 ms) modulation functions underlying the time-varying responses. Information-theoretic analysis revealed that the responses encoded only approximately 1 bit/s about constant-velocity stimuli but up to 29 bits/s about the time-varying stimuli. Analysis of local field potentials revealed that part of the residual response variability arose from "noise" sources extrinsic to the neuron. Our results demonstrate that extrastriate neurons in alert primates can encode the fine temporal structure of visual stimuli.

Impact of synaptic unreliability on the information transmitted by spiking neurons

The spike generating mechanism of cortical neurons is highly reliable, able to produce spikes with a precision of a few milliseconds or less. The excitatory synapses driving these neurons are by contrast much less reliable, subject both to release failures and quantal fluctuations. This suggests that synapses represent the primary bottleneck limiting the faithful transmission of information through cortical circuitry. How does the capacity of a neuron to convey information depend on the properties of its synaptic drive? We address this question rigorously in an information theoretic framework. We consider a model in which a population of independent unreliable synapses provides the drive to an integrate-and-fire neuron. Within this model, the mutual information between the synaptic drive and the resulting output spike train can be computed exactly from distributions that depend only on a single variable, the interspike interval. The reduction of the calculation to dependence on only a single variable greatly reduces the amount of data required to obtain reliable information estimates. We consider two factors that govern the rate of information transfer: the synaptic reliability and the number of synapses connecting each presynaptic axon to its postsynaptic target (i.e., the connection redundancy, which constitutes a special form of input synchrony). The information rate is a smooth function of both mechanisms; no sharp transition is observed from an "unreliable" to a "reliable" mode. Increased connection redundancy can compensate for synaptic unreliability, but only under the assumption that the fine temporal structure of individual spikes carries information. If only the number of spikes in some relatively long-time window carries information (a "mean rate" code), an increase in the fidelity of synaptic transmission results in a seemingly paradoxical decrease in the information available in the spike train. This suggests that the fine temporal structure of spike trains can be used to maintain reliable transmission with unreliable synapses.

Analytical and simulation results for stochastic Fitzhugh-Nagumo neurons and neural networks

An analytical approach is presented for determining the response of a neuron or of the activity in a network of connected neurons, represented by systems of nonlinear ordinary stochastic differential equations--the Fitzhugh-Nagumo system with Gaussian white noise current. For a single neuron, five equations hold for the first- and second-order central moments of the voltage and recovery variables. From this system we obtain, under certain assumptions, five differential equations for the means, variances, and covariance of the two components. One may use these quantities to estimate the probability that a neuron is emitting an action potential at any given time. The differential equations are solved by numerical methods. We also perform simulations on the stochastic Fitzugh-Nagumo system and compare the results with those obtained from the differential equations for both sustained and intermittent deterministic current inputs with superimposed noise. For intermittent currents, which mimic synaptic input, the agreement between the analytical and simulation results for the moments is excellent. For sustained input, the analytical approximations perform well for small noise as there is excellent agreement for the moments. In addition, the probability that a neuron is spiking as obtained from the empirical distribution of the potential in the simulations gives a result almost identical to that obtained using the analytical approach. However, when there is sustained large-amplitude noise, the analytical method is only accurate for short time intervals. Using the simulation method, we study the distribution of the interspike interval directly from simulated sample paths. We confirm that noise extends the range of input currents over which (nonperiodic) spike trains may exist and investigate the dependence of such firing on the magnitude of the mean input current and the noise amplitude. For networks we find the differential equations for the means, variances, and covariances of the voltage and recovery variables and show how solving them leads to an expression for the probability that a given neuron, or given set of neurons, is firing at time t. Using such expressions one may implement dynamical rules for changing synaptic strengths directly without sampling. The present analytical method applies equally well to temporally nonhomogeneous input currents and is expected to be useful for computational studies of information processing in various nervous system centers.

Functional identification of the input-output transforms of motoneurones in the rat and cat

1. We studied the responses of rat hypoglossal and cat lumbar motoneurones to a variety of excitatory and inhibitory injected current transients during repetitive discharge. The amplitudes and time courses of the transients were comparable to those of the synaptic currents underlying unitary and small compound postsynaptic potentials (PSPs) recorded in these cells. Poisson trains of ten of these excitatory and ten inhibitory current transients were combined with an additional independent, high-frequency random waveform to approximate band limited white noise. The white noise waveform was then superimposed on long duration (39 s) suprathreshold current steps. 2. We measured the effects of each of the current transients on motoneurone discharge by compiling peristimulus time histograms (PSTHs) between the times of occurrence of individual current transients and motoneurone discharges. We estimated the changes in membrane potential associated with each current transient by approximating the passive response of the motoneurone with a simple resistance-capacitance circuit. The relations between the features of these simulated PSPs and those of the PSTHs were similar to those reported previously for real PSPs: the short-latency PSTH peak (or trough) was generally longer than the initial phase of the PSP derivative, but shorter than the time course of the PSP itself. Linear models of the PSP to PSTH transform based on the PSP time course, the time derivative of the PSP, or a linear combination of the two parameters could not reproduce the full range of PSTH profiles observed. 3. We also used the responses of the motoneurones to the white noise stimulus to derive zero-, first- and second-order Wiener kernels, which provide a quantitative description of the relation between injected current and discharge probability. The convolution integral computed for an injected current waveform and the first-order Wiener kernel should provide the best linear prediction of the associated PSTH. This linear model provided good matches to the PSTHs associated with a wide range of current transients. However, for the largest amplitude current transients, a significant improvement in the PSTH match was often achieved by expanding the model to include the convolution of the second-order Wiener kernel with the input. 4. The overall transformation of current inputs into firing rate could be approximated by a second-order Wiener model, i.e. a cascade of a dynamic, linear filter followed by a static non-linearity. At a given mean firing rate, the non-linear component of the response of the motoneurone could be described by the square of the linear component multiplied by a constant coefficient. The amplitude of the response of the linear component increased with the average firing rate, whereas the value of the multiplicative coefficient in the non-linear component decreased. As a result, the overall transform could be predicted from the mean firing rate and the linear impulse response, yielding a relatively simple, general description of the motoneurone input-output function.

Computer simulation of noise resulting from random synaptic activities

Neuronal behavior is dependent on random inputs from a multitude of synaptic contacts on the soma and dendritic tree. Therefore, simulations of different types of noise are often required in the experimental and theoretical investigation of the properties of neurons and neuronal assemblies. The direct simulation of these noise sources by simple difference equations may therefore be quite useful and a general approach is presented in this paper. Initially, a first order model and its time-discretization are analyzed in detail, followed by a generalization to more complex models. The firing patterns of neurons are dependent on the random behaviors of their membrane potentials at the trigger zone. These depend on the propagation of the randomly occurring postsynaptic potentials from specific places on the dendritic tree or soma to the trigger zone. Different models may represent a variety of circumstances in which random membrane potentials arise at the trigger zone. Simulations of different types of noise are often required in the experimental and theoretical investigation of the properties of neurons and neuronal assemblies. The direct simulation of these noise sources by simple difference equations may therefore be quite useful and a general approach is presented in this paper. This paper presents a detailed analysis of the very useful first order model and its time discretization. The criterion used is that the autocovariance sequence of the discrete time model be a sample of the original autocovariance function. Several cases are presented which are of practical interest, including the case of constant output variance independent of the model's time constant. General models are time-discretized by the impulse response invariance method. Two applications are presented, one is related to the modeling of the synaptic currents by the alpha function instead of the delta function and the second deals with analog synaptic noise generation by D/A conversion of computer generated noise sequences.

Estimation of certain parameters of a stationary hybrid process involving a time series and a point process

A method is presented for estimating the cross-spectral density of a hybrid process involving a time series and a point process. The method is based on the generalized cross-periodogram statistic, which is smoothed by splitting the whole record of the data into a number of disjoint subrecords. Estimates of the coherence function and the cross-covariance function can also be obtained by using the estimate of the cross-spectral density. The distribution of the cross-covariance function between a time series and a point process is shown to be asymptotically normal. The theoretical results are used in the study of a complex physiological system. It is shown that the presence of a gamma motor neuron (gamma stimulation) modifies the effect of the length changes on the complex system at low frequencies (the length changes, and the response of the system become uncorrelated in the range 3-30 Hz) while the effect remains unchanged at higher frequencies. As a comparison it is shown that the presence of the length changes weakens the effect of the gamma stimulation on the complex system.

Third-order reverse correlation analysis of muscle spindle primary afferent fiber responses to random muscle stretch

The response of primary muscle spindle afferent fibers to muscle stretch is nonlinear. Now spindle responses (trains of action potentials) to band-limited Gaussian white noise length perturbations of the gastrocnemius muscles (input signal) are described in cats. The input noise upper cutoff frequency was clearly above the frequency range of physiological length changes in cat hindleg muscles. The input-output relation was analyzed by means of peri-spike averages (PSAs), which could be shown to correspond to the kernels of Wiener's white noise approach to systems identification. The present approach (the reverse correlation analysis) was applied up to the third order. An experiment consisted of two recordings: one (the source recording) to determine PSAs and the other (the test recording) to provide an input signal for predicting responses. The predictions of different orders were compared with the actual neuronal response (the observation) of the test recording. Four different approximation procedures were developed to adapt prediction and observation and to determine weighting factors for the predictions of different orders. The approximations also yielded the value of the power density P of the input noise signal: at a variety of stimulus parameters, P from approximations had the same magnitude as P determined directly from the input signal amplitude spectrum. The prediction of a sequence of action potentials improved the higher the order of components. 37 of 42 action potentials of a test recording (the observation) could be confidently predicted from PSAs or kernels. Compared with the size of the linear first-order prediction curve, the relative sizes of the second and third-order prediction curves were: 1.0:0.47:0.26.

Reliability of spike timing in neocortical neurons

It is not known whether the variability of neural activity in the cerebral cortex carries information or reflects noisy underlying mechanisms. In an examination of the reliability of spike generation using recordings from neurons in rat neocortical slices, the precision of spike timing was found to depend on stimulus transients. Constant stimuli led to imprecise spike trains, whereas stimuli with fluctuations resembling synaptic activity produced spike trains with timing reproducible to less than 1 millisecond. These data suggest a low intrinsic noise level in spike generation, which could allow cortical neurons to accurately transform synaptic input into spike sequences, supporting a possible role for spike timing in the processing of cortical information by the neocortex.

Spike discharges of skeletomotor neurons during random noise modulated transmembrane current stimulation and muscle stretch

Spike discharges of skeletomotor neurons innervating triceps surae muscles elicited by white noise modulated transmembrane current stimulation and muscle stretch were studied in decerebrated cats. The white noise modulated current intensity ranged from 4.3 to 63.2 nA peak-to-peak, while muscle stretches ranged from 100 microns to 4.26 mm peak-to-peak. The neuronal responses were studied by averaging the muscle length records centered at the skeletomotor action potentials (peri-spike average, PSA) and by Wiener analysis. Skeletomotor spikes appeared after a sharp peak in PSA of the injected current, preceded by a longer-lasting smaller wavelet of either depolarizing or hyperpolarizing direction. The PSA amplitude was not related to the injected current amplitude nor showed any differences related to the motor unit type. The PSA amplitudes were virtually independent of the stretching amplitude sigma, after an initial increase with stretching amplitudes in the range of 15-40 microns (S.D.), or 100-270 microns peak-to-peak. Analyses of cross-spectra indicated a small or absent increase in gain with frequency in response to injected current, but about 20 dB/decade in the range 10-100 Hz in response to muscle stretch. The peaks of both Wiener kernels in response to current injection appear to decrease with the amplitude of injected current, but this decrease was not statistically significant. The narrow first-order kernels suggest that the transfer function between the current input and spike discharge is lowpass with a wide passband, i.e. there is very little change in dynamics. The values of the second-order kernels appear to be nonzero only along the main diagonal.(ABSTRACT TRUNCATED AT 250 WORDS)

Membrane potential changes of skeletomotor neurons in response to random stretches of the triceps surae muscles in decerebrate cats

The properties of membrane potential changes of skeletomotor neurons (S, FR, and FF) innervating triceps surae muscles during pseudorandom stretching of these muscles were studied in decerebrate cats. Peak amplitudes of pseudorandom muscle stretches ranged from 119 microns to 4.15 mm peak-to-peak. Sequences of ten identical stretching periods were applied for averaging. Shapes of membrane potential changes and probability density distribution of amplitudes of the input and output signals and power spectra suggest that the skeleto-motor neuron membrane has nonlinear properties. First- and second-order Wiener kernels were determined by applying the cross-correlation (Lee-Schetzen) method. The results suggest that the transfer function between muscle stretches and subthreshold membrane potentials is a Wiener-type cascade. This cascade is consistent with a linear, second-order, underdamped transfer function followed by a simple quadratic nonlinearity [linear (L) system followed by nonlinear (N) system, or LN cascade]. Including the nonlinear component calculated from the second-order Wiener kernel improved the model significantly over its linear counterpart, especially in S-type motoneurons. Qualitatively similar results were obtained with all types of motoneurons studied.

Reverse correlation analysis of the stretch response of primary muscle spindle afferent fibers

The nonlinear responses of deefferented primary muscle spindle afferent fibers to muscle stretching consisted of a train of action potentials which was analyzed when random changes in muscle length (band-limited gaussian white noise) were applied in cats. The upper cutoff frequency of the applied noise (the source stimulus) was varied between 1.6 and 570 Hz; the amplitude of the random input was varied between 0.002 and 1.2 mm. In a previous report the reverse correlation of 1st and 2nd order was studied for its ability to analyze data of a continuous input signal and pulsatile events in the output. Computations of the Wiener kernels h1 and h2 or their equivalents, the perispike averages of the 1st and 2nd order, were computed from the random stretch responses of muscle-spindle afferents. Then the 1st- and the 2nd-order predictions and the summation of both to random muscle stretch was estimated. A general finding was that the 1st-order component was approximately 10 times that of the 2nd-order component, when both were combined in approximation procedures to give the closest prediction of observed responses to random test stimuli. The approximation was poor when the source stimulus was less than 0.03 mm and improved when it was greater. With the increase in the upper cutoff frequency of the random source input, the approximation worsened continuously. Predictions to ramp-and-hold stimuli were computed, as well as responses to random stimulation. Limiting the upper cutoff frequency did not diminish the value of the techniques applied.

The RF-cinematogram. A cross-correlation technique for mapping several visual receptive fields at once

We present a spike-triggered averaging method capable of mapping the visual receptive fields of several neurons simultaneously. The stimulation is general and the mapping proceeds automatically without the need to match the stimulation to the cells' preference for position, orientation, direction, etc. The maps are spatiotemporal; receptive field (RF) structures are quantitatively determined in three dimensions: the two dimensions of visuotopic space, and time. The method presented is one of a family of "reverse correlation" or "spike-triggered averaging" techniques (DeBoer and Kuyper 1968) capable of revealing linear aspects of stimulus-response coupling. The formal relationship of these methods to stimulus-response cross-correlation is shown. The analysis is extended to provide some second-order axis-of-motion information ("direction marks"). The stimulus is a constantly illuminated, randomly jumping bright or dark spot, not an elongated bar. Spot diameters between one-third to 1 x RF width are effective. The method ascertains for each recorded action potential or "spike" the prior visual field position of the spot. The average or most probable spot positions define the receptive field spatially. Repeating the process for a succession of times prior to observed spikes defines the field temporally, presented here as a succession of spatial maps. We term this portrayal a receptive field cinematogram, RFc or ciné. The RFc reveals and economically portrays the spread of excitability and suppression across the receptive field, culminating in the generation of a spike. RFcs for LGN neurons and for simple cells recorded in cat cortical areas 17 and 18 are presented and interpreted in terms of classic ON/OFF regions. The availability of temporal information permits the separation of an excitatory exit response, generated when a moving bright spot leaves an OFF region, from an excitatory entrance response occurring when a bright spot enters an ON region, because these responses occur at different times (exit responses earlier). Spike emission remains coupled to (cross-correlated with) stimulus events over time periods as long as 96 ms, implying that some stimulus drive or afferent visual input is delayed by as much as 96 ms more than other input. This is a striking instance of temporal dispersion in the visual system. In some cells, said to be "spatiotemporally inseparable", the delay (latency) varies systematically across the visual field; i.e., the place for optimal stimulation varies with the time prior to spike emission. In these cells, the RFc shows receptive field structures which move across the visual field over trajectories equal to approximately twice the total conventional RF width.(ABSTRACT TRUNCATED AT 400 WORDS)

Signal transduction and nonlinearities revealed by white noise inputs in the fast adapting crayfish stretch receptor

Input-output relations were investigated in the fast adapting stretch receptor organ (RM2) of the crayfish by matching gaussian white noise (GWN) length inputs, with the resulting spike output. The analysis revealed the expected sensitivity to lengthening velocity, a behavior termed phasic. It also disclosed a sensitivity to sustained elongation, a performance termed tonic and previously not recognized in the RM2. Spectral analysis indicated the properties of a low-pass filter, confirming the tonic sensitivity. A variety of individual length trajectories could lead to a spike. The average trajectory consisted in a biphasic shortening-lengthening wave. The range of possible trajectories and their averages changed with stimulus prestretch and GWN amplitude, indicating that system properties depended on the input characteristics; i.e., a nonlinear operation. Length waveforms in the GWN were isolated by computing methods and the corresponding responses were calculated. Symmetric stimuli led to responses that reflected magnitudes and velocities asymmetrically. Nonlinear interactions between responses in the past and present were negligible. In conclusion, depending on the input, the RM2 modifies its operation to enhance the detectability of the functionally relevant signal in each natural situation.

Dynamic relationship between the slow potential and spikes in cockroach ocellar neurons

The relationship between the slow potential and spikes of second-order ocellar neurons of the cockroach, Periplaneta americana, was studied. The stimulus was a sinusoidally modulated light with various mean illuminances. A solitary spike was generated at the depolarizing phase of the modulation response. Analysis of the relationship between the amplitude/frequency of voltage modulation and the rate of spike generation showed that (a) the spike initiation process was bandpass at approximately 0.5-5 Hz, (b) the process contained a dynamic linearity and a static nonlinearity, and (c) the spike threshold at optimal frequencies (0.5-5 Hz) remained unchanged over a mean illuminance range of 3.6 log units, whereas (d) the spike threshold at frequencies of less than 0.5 Hz was lower at a dimmer mean illuminance. The voltage noise in the response was larger and the mean membrane potential level was more positive at a dimmer mean illuminance. Steady or noise current injection during sinusoidal light stimulation showed that (a) the decrease in the spike threshold at a dimmer mean illuminance was due to the increase in the noise variance: the noise had facilitatory effects on the spike initiation; and (b) the change in the mean potential level had little effect on the spike threshold. We conclude that fundamental signal modifications occur during the spike initiation in the cockroach ocellar neuron, a finding that differs from the spike initiation process in other visual systems, including Limulus eye and vertebrate retina, in which it is presumed that little signal modification occurs at the analog-to-digital conversion process.

The maximum likelihood approach to the identification of neuronal firing systems

The concern of this work is the identification of the (nonlinear) system of a neuron firing under the influence of a continuous input in one case, and firing under the influence of two other neurons in a second case. In the first case, suppose that the data consist of sample values Xt, Yt, t = 0, +/- 1, +/- 2,... with Yt = 1 if the neuron fires in the time interval t to t + 1 and Yt = 0 otherwise, and with Xt denoting the (sampled) noise value at time t. Suppose that Ht denotes the history of the process to time t. Then, in this case the model fit has the form Prob[Yt = 1/Ht] = phi(Ut-theta) where (formula; see text) where gamma t denotes the time elapsed since the neuron last fired and phi denotes the normal cumulative. This model corresponds to quadratic summation of the stimulus followed by a random threshold device. In the second case, a network of three neurons is studied and it is supposed that (formula; see text) with Xt and Zt zero-one series corresponding to the firing times of the two other neurons. The models are fit by the method of maximum likelihood to Aplysia californica data collected in the laboratory of Professor J.P. Segundo. The paper also contains some general comments of the advantages of the maximum likelihood method for the identification of nonlinear systems.

Dynamics of the ganglion cell response in the catfish and frog retinas

Responses were evoked from ganglion cells in catfish and frog retinas by a Gaussian modulation of the mean luminance. An algorithm was devised to decompose intracellularly recorded responses into the slow and spike components and to extract the time of occurrence of a spike discharge. The dynamics of both signals were analyzed in terms of a series of first-through third-order kernels obtained by cross-correlating the slow (analog) or spike (discrete or point process) signals against the white-noise input. We found that, in the catfish, (a) the slow signals were composed mostly of postsynaptic potentials, (b) their linear components reflected the dynamics found in bipolar cells or in the linear response component of type-N (sustained) amacrine cells, and (c) their nonlinear components were similar to those found in either type-N or type-C (transient) amacrine cells. A comparison of the dynamics of slow and spike signals showed that the characteristic linear and nonlinear dynamics of slow signals were encoded into a spike train, which could be recovered through the cross-correlation between the white-noise input and the spike (point process signals. In addition, well-defined spike correlates could predict the observed slow potentials. In the spike discharges from frog ganglion cells, the linear (or first-order) kernels were all inhibitory, whereas the second-order kernels had characteristics of on-off transient excitation. The transient and sustained amacrine cells similar to those found in catfish retina were the sources of the nonlinear excitation. We conclude that bipolar cells and possibly the linear part of the type-N cell response are the source of linear, either excitatory or inhibitory, components of the ganglion cell responses, whereas amacrine cells are the source of the cells' static nonlinearity.

Neuronal sources of theta rhythm in the entorhinal cortex of the rat. II. Phase relations between unit discharges and theta field potentials

The discharge patterns and layer distribution of entorhinal cortex (EC) units were investigated in paralysed and locally anesthetized rats injected with physostigmine in order to induce theta (theta) rhythm. Entorhinal unit activity and field potentials were recorded simultaneously with the same micropipette. Hippocampal CA1 theta rhythm was used as reference. Statistical analysis included auto- and cross-correlations and interval histograms. Results showed: a. the existence of rhythmic and non-rhythmic cells, both tending to fire in a constant phase relationship with theta rhythm; b. in all EC subdivisions, most rhythmic cells were located in superficial cell layers (II-III); c. on the average, rhythmic cells from the medial EC fired synchronously; d. non-rhythmic cells tended also to fire synchronously but with an opposite phase relationship with respect to rhythmic neurons. Although a complex organization in the rhythmicity of EC units is revealed, it is concluded that the neuronal sources of theta activity in the EC are located in superficial cell layers, and it is strongly suggested that the EC output through the perforant path may rhythmically modulate the discharge pattern of hippocampal pyramidal and dentate granule cells.

Cross-correlation analysis of septohippocampal neurons during theta-rhythm

The activity from 55 septohippocampal neuron pairs was examined in rats anesthetized with urethane. In addition to the statistical characterization of the firing patterns of the recorded units, the functional interactions between pairs of neurons and between neurons and hippocampal theta (theta) waves were investigated with cross-correlation techniques. Pairs were classified according to the rhythmic or non-rhythmic discharge pattern of their neurons. (a) theta-Pairs were those in which both the medial septal (MS) and hippocampal (HPC) units were rhythmic (type 1 units). (b) Pairs with a rhythmic and a theta-related non-rhythmic unit (type 2 unit) were called mixed pairs. (c) Pairs composed of type 2 units were called type 2 pairs. theta-Pairs showed periodic cross-correlations and frequently fired with a phase difference which could change in different pairs. Mixed pairs also showed periodic cross-correlations although one of the units was non-rhythmic. Type 2 pairs showed non-periodic positive cross-correlations. Our data provide new information regarding the temporal relationship between MS and HPC rhythmic activities supporting the role of the MS in providing the afferent timing for the generation of theta-rhythm in the HPC.

Poisson process stimulation of an excitable membrane cable model

The convergence of multiple inputs within a single-neuronal substrate is a common design feature of both peripheral and central nervous systems. Typically, the result of such convergence impinges upon an intracellularly contiguous axon, where it is encoded into a train of action potentials. The simplest representation of the result of convergence of multiple inputs is a Poisson process; a general representation of axonal excitability is the Hodgkin-Huxley/cable theory formalism. The present work addressed multiple input convergence upon an axon by applying Poisson process stimulation to the Hodgkin-Huxley axonal cable. The results showed that both absolute and relative refractory periods yielded in the axonal output a random but non-Poisson process. While smaller amplitude stimuli elicited a type of short-interval conditioning, larger amplitude stimuli elicited impulse trains approaching Poisson criteria except for the effects of refractoriness. These results were obtained for stimulus trains consisting of pulses of constant amplitude and constant or variable durations. By contrast, with or without stimulus pulse shape variability, the post-impulse conditional probability for impulse initiation in the steady-state was a Poisson-like process. For stimulus variability consisting of randomly smaller amplitudes or randomly longer durations, mean impulse frequency was attenuated or potentiated, respectively. Limitations and implications of these computations are discussed.

The non-uniform character of expiratory synaptic activity in expiratory bulbospinal neurones of the cat

Intracellular recordings were made from caudal medullary expiratory neurones in pentobarbitone-anaesthetized, vagotomized and artificially ventilated cats. The sample consisted of thirty-three bulbospinal neurones and seven neurones which were not antidromically excited from either the spinal cord (C2-C3) or vagus nerve. Their rhythmic activity consisted of an alternating inspiratory hyperpolarization due to Cl(-)-dependent inhibitory post-synaptic potentials (i.p.s.p.s) (Mitchell & Herbert, 1974) and an expiratory depolarization. The precise shape of the expiratory depolarizing wave varied within a given neurone depending on the over-all pattern of respiration. This variation extended from a smoothly developing depolarization, continuous throughout its course, through an intermediate state in which depolarization proceeded in two stages with a definite transition between them, to a final state in which the early part of expiration was occupied by a distinct hyperpolarizing component to the membrane potential trajectory. Under conditions of a brisk phrenic nerve discharge, these variations in the shape of the membrane potential profile were related to the time course and intensity of post-inspiratory discharge in the nerve. However, other factors (depth of anaesthesia and stimulation of laryngeal receptors) could influence the time course of the membrane potential profile of expiratory neurones independently of post-inspiratory phrenic discharge. In five of fifteen neurones which were tested, early expiration was occupied by a rapidly developing, decrementing wave of Cl(-)-dependent i.p.s.p.s (post-inspiratory i.p.s.p.s). These i.p.s.p.s were present only under conditions of a strong phrenic rhythm (large amplitude, fairly rapid phrenic discharge). They became weaker and ultimately disappeared when the level of anaesthesia was deepened and the phrenic rhythm became slower. Under these conditions, the post-inspiratory wave of i.p.s.p.s could be restored by stimulation of the superior laryngeal nerve. Adequate stimulation of presumed 'irritant' laryngeal receptors elicited post-inspiratory i.p.s.p.s in seven of ten neurones tested which initially showed either no post-inspiratory i.p.s.p.s or possibly just a weak pattern. In ten of fifteen neurones tested, the responses to current injection revealed clear differences in membrane potential behaviour in early and late expiration, which became intensified following stimulation of the superior laryngeal nerve.(ABSTRACT TRUNCATED AT 400 WORDS)