Further Studies on Synaptic Transmission in Insects: II. Relations Between Sensory Information and its Synaptic Integration at the Level of a Single Giant Axon in the Cockroach

One of the major problems in modern neurophysiology concerns the mechanism by which information transmitted from receptors is transformed in the central nervous system into a more or less complex coded information, and in addition, how it is conducted to other parts of the nervous system (interneurones) or to muscles (moto-neurones). The analysis of these integration processes has been carried out most often in relatively simple preparations such as ganglia in the nervous system of invertebrates or sympathetic nervous systems of vertebrates, for reviews, see (Tauc, 1967; Kandel & Kupfermann, 1970). Only relatively little information is available concerning integration of ‘unit’ presynaptic impulses (Maynard, 1966; Kandel & Wachtel, 1968; Kennedy, 1968; Gorman & Mirolli, 1969). The main difficulty most frequently encountered in this kind of study is the lack of a sufficient control of the sensory afferences, especially when the effect of single presynaptic impulses on the post-synaptic activity is studied.

Another inconvenience of many synaptic preparations is that their integrative properties are limited. A few presynaptic spikes are often sufficient to trigger a post-synaptic action potential.

The last abdominal ganglion of the cockroach provides a particularly suitable preparation for integration analysis. In the neuropile of this ganglion a large number of sensory fibres conducting information from very well-defined cereal mechanoreceptors are synaptically connected with a small number of giant axons which play some part in the escape reflex of this insect (Roeder, 1948; Dagan & Parnas, 1970).

A puff of air of sufficient strength applied to the cerci leads to a reflex escape of the animal. According to Pumphrey & Rawdon-Smith (1937) and Roeder (1948) this reflex involves one monosynaptic relay in the sixth abdominal ganglion between the cereal sensory fibres and the giant interneurones. This first abdominal relay has been used for the study of synaptic transmission in insects (see review by Boistel, 1969). It was not until 1966, however, that, using very fine tipped capillary microelectrodes (0.1-0.20µm), we have been able to study some of the electrical characteristics of this transmission, when the cereal nerves are stimulated electrically (Callec & Boistel, 1966 a, b). Recent experiments involving the use of an oil-gap technique to record postsynaptic potentials in a single postsynaptic giant axon (Pichon & Callec, 1970 has allowed for considerable progress to be made in this field. We will describe in this paper the characteristic features of the input-output relations across this synaptic relay. They deal with the postsynaptic events associated with: (1) a single presynaptic impulse; (2) a volley of impulses corresponding to the stimulation of a single cereal mechanoreceptor; (3) the asynchronous (mechanical stimulation) or synchronous (electrical stimulation) activity of a variable number of presynaptic fibres.

These results will be analysed and discussed in terms of integrative properties of the synaptic junctions.

A preliminary account of this work has already appeared (Callec, et al. 1969).

The sensory input to the 6th abdominal ganglion consists in volleys of nerve impulses generated essentially at the level of the cereal mechanoreceptors and conducted by the cereal nerves X and XI.

Although a complete mapping of all the homolateral or contralateral receptors which are connected to a given giant axon has not yet been made, it has been demonstrated that the number of these receptors is rather high (we have for example shown a relation between a dozen receptors and the median ventral giant axon).

These receptors are long (up to 700 µm) and thin (6 µm) sensilla with (most often) one single nerve cell. The axonal processes of these cells constitute the sensory cereal nerves. The firing pattern of these receptors can be classified in two classes corresponding to morphologically different receptors.

The longer ones (700µm) give phasic-tonic patterns which are predominantly tonic ; they will be termed ‘tonic patterns’ in this paper. This first category is extremely sensitive to air movements (the response rises 40 % of maximum when the hair describes only a two degree arc along the preferential plane, Nicklauss, 1965. The others (less than 350 µm) adapt more or less rapidly to the stimulation and give responses which will be termed ‘phasic-tonic patterns’.

These responses are characterized by a short burst of a few spikes to each stimulation. The maximum frequency which can be reached during the phasic component of these different responses is of about 360 spikes/s.

The output from the 6th abdominal ganglion is carried by different kinds of interneurones. Some of them, the giant axons, are relatively large in diameter. These giant intemuncials are distributed within each connective in two groups: a ventral consisting of three larger giant axons (up to 40–45 µm in diameter), and a dorsal group consisting of five to seven smaller axons (20–25 µm in diameter) (Roeder, 1948).

It appears from electromicroscopical studies (Farley & Milbum, 1969) that in the 6th ganglion each giant fibre bears numerous ramifications analogous to dendritic tree which most of the afferent cereal fibres seem to make synaptic contacts. These ures are similar to those found in other invertebrates as in the crayfish (Kennedy, Selverston & Remler, 1969), in the crab (Sandeman, 1969) and in the leech (Stuart, 1970; Nicholls & Purves, 1970).

The cell bodies of these giant axons are located at the periphery of the 6th ganglion. According to Farley & Milburn (1969) one cluster of lateroventral somata are related to each fibre. Preliminary intracellular microelectrode recordings have shown that cell bodies in this region cannot give rise to full-sized action potentials, and it therefore seems that they cannot directly affect the electrical activity of the giant axons.

The stimulation of the cereal mechanoreceptors can be achieved in different ways. Most often some of them seem to be ‘spontaneously’ active, thus giving a ground activity which can be recorded on the cereal nerve. This activity results in fact from the stimulation of some of the receptors which are bent by Ringer or dust; it can be at least provisionally suppressed by a careful drying and cleaning of the cerci (Fig. 2).

Each receptor can be stimulated separately using a small glass rod held by manipulators. Several receptors can be activated more or less simultaneously by puffs of air of various strengths and durations, and the synchronous activity of the cereal fibres can be achieved by a direct electrical stimulation of the cereal nerves.

The arrangement diagrammatically shown in Fig. 1 allows simultaneous recordings of the presynaptic and of the postsynaptic activity. We can thus follow with a good accuracy and at the unitary level the different patterns of information arising from a well-defined receptor until its integration at the first synaptic relay.

Presynaptic activity conducted in the cereal nerves is recorded between two baths of saline separated by a vaseline seal, care being taken to make this seal as light as possible to avoid any conduction block in the superficial fibres due to the compression. Unit spikes are of small amplitude (o°2-1 mV) but can usually be distinguished from the ‘noise’. It is possible to get greater potentials by using a microelectrode placed at the base of a given receptor, but this technique does not generally permit a long-duration recording and damages the receptor. Thus, the former method has been preferred except for the records of Fig. 6.

Postsynaptic events have been recorded using the previously described ‘oil-gap’ technique (Pichon & Callee, 1970). This technique provides some advantages: a low impedance preparation (1–5 MΩ) and thus a low noise level, a very good stability whenever the preparation is stimulated mechanically, and the unequivocal identification of the postsynaptic element at the level of which the recordings are made. Most of the experiments described here have been made on the median ventral giant axon, few of them on the lateral ventral giants. The ganglion is continuously irrigated with an oxygenated physiological saline (composition: Na⁺ 214 mM, K⁺ 3·1 mM, Ca²⁺ 1·8 mM, H₂PO₄⁻ 0·2 mM, H-PO₄²⁻ 4·8 mM, in which the synaptic transmission could be maintained unimpaired during several hours. In the last experiments the Ca®+ concentration has been raised to 5·4 mM to provide a better stabilization of the fibre membrane and to enhance slightly the synaptic potentials.

The preparation consists of the cerci, the cereal nerves and the abdominal ganglion. The single giant axon is dissected over 1-2 mm between the 5th and the 6th ganglia as close as possible from the 6th ganglion using the previously described technique (Pichon & Boistel, 1966 ; Pichon & Callec, 1970). It is then transferred to the recording chamber shown in Fig. 1. The cerci are then carefully cleaned and dried until all ‘spontaneous’ activity in the cereal nerves has completely disappeared. The level of polarization of the postsynaptic fibre membrane is continuously monitored by a pen recorder. The mechanical stimulation and the lack of oxygenation during the dissection lead to a moreor less important depolarization of thepostsynaptic membrane. After the irrigation is started the preparation is left unstimulated during a variable period of time to allow the normal polarization to be reached again. The duration of the repolarization is variable (5–15 min.) depending essentially upon the time between be beginning of the dissection and the beginning of the irrigation. It is only after the postsynaptic membrane has reached a stable level that one begins to look for receptors related to the dissected axon, this being done using a fine glass rod mounted on a manipulator.

When the preparation is placed in the recording chamber, one can usually record an important electrical activity as shown in Fig. 2 A. This activity which consists essentially in so-called ‘spontaneous’ excitatory postsynaptic potentials (EPSPs) can be, as already mentioned, related to the mechanical stimulation of the cereal receptors. Usually it is suppressed by a careful drying of the cerci. In these conditions one can also record, in most cases, small hyperpolarizing waves which will be referred to as inhibitory postsynaptic potentials (IPSPs).

(a) EPSPs. The mean amplitude of these spontaneous EPSPs varies between 0·15 and 0·9 mV with a mean value of 0·5 mV (in Ringer 1·8 mM-Ca^a+). A histogram of the amplitudes of these potentials does not show any characteristic peak as shown on Fig. 3 A. The time to peak can vary between 1·7 and 3·4013, but presents a sharp peak for 2·0 ms (Fig. 3 B). The time necessary for these EPSPs to decrease by $23$ of their peak value ranges between 3·9 to 5·3 ms.

(b) IPSPs. The IPSPs (Fig. 4), can be observed only in special conditions, when the excitatory activity is reduced or suppressed (Fig. 2) and when the polarization the postsynaptic element remains below a certain level. Experiments have shown therefore, that the polarization of the postsynaptic element can slightly increase after 2 or 3 h of experimentation. As the reversal of these IPSPs can be obtained for hyperpolarizations of about + 5 mV from the normal’ resting level, one can understand why these potentials have not been observed in our previous experiments, in which the time for dissection and mounting was longer than an hour and a half.

The mean amplitude of these IPSPs varies generally between 40 and 320 µV (in Ringer 1·8 mM Ca²⁺). The histogram of the amplitudes of these IPSPs shows two peaks in Fig. 5. Their shape is essentially similar to that of the EPSPs, but they are generally slower (time to peak ranging from 3·3 to 5·0 ms ; total duration ranging from 10 to 15 ms). Their frequency remains low (4–20 s). These potentials still occur in completely de-afferented preparations showing that they must be related to an in-ganglionic presynaptic activity. One does not know whether this activity originates in axonal regions or in cell bodies. We have reported earlier (Callee & Boistel, 1966a) that some cell bodies located in the mediodorsal part of the ganglion may fire spontaneously, and it does not seem impossible that this spike activity might be the origin of at least some of the ‘spontaneous’ IPSPs. These intraganglionary IPSPs cannot be considered as sensory information despite their evident role in modulating the excitatory postsynaptic responses. They will be studied in more detail in another paper.

Simultaneous recording of presynaptic and postsynaptic activity have shown that, in most cases, each spike at the receptor triggers an EPSP (Fig. 6).

The amplitude of the EPSPs corresponding to various receptors might be different (Fig. 7). Furthermore, the amplitude of the EPSPs corresponding to a single receptor might vary by a factor of three, whereas the time to peak remains relatively constant (about 2 ms) (Fig. 8).

The delay between the presynaptic spike recorded where it enters the ganglion and the rising phase of the corresponding EPSP is of about 0·85 ms (mean for five proparafions). This would correspond after correction for the conduction time in the presynaptic element (about 350 µm at 2 ms, that is 0·17 m/s) to a mean synaptic delay of 0·68 ms (Fig. 7).

The preceding recordings have usually been obtained with receptors firing at a rather low frequency so that each spike could be considered as isolated. This is far from being general, and one knows that the frequency of the spikes initiated in one cereal mechanoreceptor can reach 360/s. It was thus important in a second step to analyse the evolution of the postsynaptic events when the unitary presynaptic input consisted in bursts of spikes.

These postsynaptic responses, associated with the different patterns of presynaptic activity, differ somewhat according to the simulated receptor.

The time course of the postsynaptic response approximately follows the changes in the frequency of the afferent volley. The level of the induced postsynaptic depolarization increases with increasing frequency. Nevertheless, the amplitude of the unitary EPSPs decreases with increasing depolarization; this is a limiting factor so that the maximum depolarization which could be obtained following natural stimulation of these receptors (corresponding to a presynaptic frequency of 300/3 in this example is not sufficient to trigger a postsynaptic spike. Parameters, such as potentiation or depression, have little or no effect; for a given level of polarization the amplitude of unitary EPSPs remains nearly constant. When the frequency undergoes a rapid and brief variation, we observe a subsequent increase in the depolarization.

As far as we know, purely tonic receptors (that is, receptors whose frequency of firing remains constant throughout the stimulation) are only sparsely distributed. Almost all the receptors we have used for this study were in fact phasic-tonic but showed a very slow adaptation to the stimulus, so that for short times the frequency of their firing could be considered as constant during the tonic component.

The time course of the postsynaptic response following this kind of afferent volley begins by a rapid depolarization followed by a progressive repolarization whenever the frequency remains constant (this is particularly obvious in Fig. 10 B, where the sequence has no initial phasic component).

The amplitude of the unitary EPSPs decreases during the first phase and then remains smaller and nearly constant, during the whole of the tonic phase.

If the stimulation is temporarily stopped at the end of the depolarizing phase (Fig. 10D), we can observe a slight and slow hyperpolarization. It is also present in the other records (Fig. 10A–C), but in part masked by the background activity. These events can be related to some kind of fatigue or depression. This phenomenon is particularly striking at the beginning of each sequence and also during the tonic component of Fig. 10B, where an increase in the frequency is no longer followed by a larger depolarization.

The mechanical stimulation of the receptors of this category gives rise to short volleys of one to a dozen of spikes. The frequency within one volley is relatively high (between 180 and 300/3). As for the other receptors, this frequency is high enough to induce summation phenomena as can be seen in Fig. 11. This summation is often accompanied by some potentiation or depression, depending upon the interval between the presynaptic spikes (Fig. 12). When this interval ranges between 3 and 8 ms, one can observe an increase of the second EPSP which could be over 20 % ; for intervals reaching 10–12 ms the phenomenon is inverted and the second EPSP is smaller than the first (10% decrease after about 32 ms). This depression lasts up to 50 ms.

Repeated stimulation of this same receptor results in a reduction of the size of the unitary EPSPs as illustrated in Fig. 13. This can safely be referred to EPSP-habitua-tion at the cellular level as defined by Bruner & Tauc (1966).

The restoration of the full amplitude is complete after a rest of 1-2 min.

We have seen in the preceding sections how a single postsynaptic giant axon is affected by afferent activity coming from different mechanoreceptors stimulated isolately and how complex the response can be. In this section we shall study the effect of asynchronous (mechanical) or synchronous (electrical) stimulation of several presynaptic fibres (which correspond to the integrative properties of the giant axons).

The effect of stimulation of the cerci by puffs of air is illustrated in Fig. 14. For a slight stimulation (A) there is only a small infraliminal depolarization corresponding to the summation of EPSPs related to a very small number of receptors. In B a rather strong puff of air gives rise to a complex response ; there is an early rapid summation which nearly reaches the threshold followed by a second step which is sufficient to trigger a propagated action potential. With a longer and stronger puff the depolarization is strong enough to induce several action potentials (6 in C).

When these puffs of relatively constant intensity are repeated at regular intervals (2/s) one can see (Fig. 15 A) a progressive but rapid decrease in the size of the postsynaptic response. This reduction can be related to the already mentioned habituation phenomenon to natural stimulation. The response always returns to its full amplitude after a rest of only a few minutes (Fig. 15 B). It has never been possible to alter this phenomenon, for instance by the interpolation of an extra stimulus applied to the homolateral or contralateral cereal nerve (Pumphrey & Rawdon Smith, 1937).

Sometimes, as for the stimulation of a single tonic receptor (cf. Fig. 10), the response which consists essentially in a depolarization is followed by a long-duration hyperpolarization (Fig. 16).

The postsynaptic response to an electrical stimulation of the cereal nerves is complex, and different arguments which will be discussed later have led us to divide our scription into monosynaptic and polysynaptic phenomena. Furthermore, the frequency of stimulation has been revealed to be critical and we will describe the effects of this parameter in a separate paragraph.

When the electrical stimulation of the homolateral cereal nerve is increased, one can see a very smooth increase in the amplitude of the EPSP (Fig. 17). When the postsynaptic depolarization is larger than 3–10 mV (the threshold is different in different preparations) it gives rise to a full-size (100–120 mV) action potential.

The EPSPs are often preceded by a small potential which increases with the intensity of the stimulation. It is likely to correspond to an electrotonic recording of the cereal (presynaptic) potential. This seems to be the only electrotonic coupling between the afferent cereal nerves and the giant interneurone.

The synaptic delay can be measured with relatively good accuracy between this electrotonic presynaptic potential and the EPSP. In this example it ranges between 83 and 0·91 ms. Whenever in a few other experiments it has been possible to note greater delays (up to 1 ·73 ms) and some variations, we think that the delay is short enough and nearly constant to argue for a monosynaptic relay.

Stimulation of the contralateral cereal nerve also gives rise to an EPSP, which means that the cereal fibres are also connected contralaterally; but it is not possible to trigger a postsynaptic actionjpotential. However, the increase of the EPSP with the progressive increase in presynaptic stimulation was smooth enough to show that the number of synaptic contacts with the contralateral giant fibre is rather important.

In good conditions (for instance after a long period of rest), the electrical stimulation of the homolateral cereal nerve gives rise to a complex response which in Fig. 18 C consists essentially in three depolarizing waves or EPSPs.

The first is the already described EPSP which is thought to be monosynaptic. The second, whose delay is about 20 ms (Fig. 18 A), increases with increasing stimulation and can give rise (but very rarely) to a conducted postsynaptic spike (Fig. 19C2). The third EPSP, well separated from the others (60–100 ms, Fig. 18 C), has a slightly higher threshold. Its amplitude is related rather to the frequency than to the intensity of stimulation. The study of these two last EPSPs shows that their delay, amplitude and time course are essentially variable. This variability, together with the high value of delay, argue for their multineuronic origin.

The delay of the second EPSP could be explained by slower conduction speed along me cereal fibres. It seems, however, that this alternative explanation is unlikely because of the relative homogeneity of the diameters of these fibres. In addition to these three kinds of EPSPs other intermediate or later depolarizations sometimes occur, but with much less regularity.

To study the effects of repetitive stimulation with volleys of given frequencies three main precautions are taken: the stimulation is subthreshold to avoid spikes, the ganglion is isolated from ‘spontaneous’ sensory inputs by cutting the cerci and the rest interval between two volleys is maintained for 3–5 min suppress as far as possible interference with depression or potentiation.

We have already mentioned that repetitive simulation results in modification of the postsynaptic events. In Fig. 20 one can see the evolution of the first (monosynaptic) EPSP to repetitive stimulation at different frequencies. For frequencies less than about 25/s there is first a more or less fast decline of the size of the EPSP followed by a plateau. The higher the frequency, the faster is this decline and the lower is the plateau. This steady state is reached after about ten stimulations. For higher frequencies (more than 25/s) the amplitude of the first EPSPs falls rapidly, then subsequently increases before falling again very slowly to the steady state. This ‘rebound’ phenomenon will be discussed later.

The second and the third (polysynaptic) EPSPs are far more labile than the first. The second disappears after about 3 s at a rate of stimulation of 5/s whereas the first (monosynaptic) EPSP remains nearly unimpaired (Fig. 18B). In other experiments one can observe some variations in its delay during the repetitive stimulation. The third is still more labile and completely disappears after about five stimulations at a frequency of 5/s (Fig. 18D2). It remains nearly constant at a lower frequency (I/10S in Fig. 18 D1), whereas it shows great variations in its delay. If the preparation is unstimulated for a period of about 3–5 min, these polysynaptic EPSPs are completely restored.

Other observations may be mentioned here (Fig. 19). If one increases further the stimulation of the afferent cereal nerve, the slope of the monosynaptic EPSP becomes steeper. Thus the threshold is more rapidly reached and the delay of the postsynaptic spike decreases. Sometimes we observe a second action potential of smaller amplitude following the first one after about 1-7 ms (about the minimum refractory period at room temperature). Strong electrical stimulation can induce an action potential appearing with no delay, probably corresponding to an electrotonic stimulation of the postsynaptic element and called direct action potential (Fig. 19 C 3). The second spike on the monosynaptic EPSP and this kind of ‘direct’ action potential have already been described by Pichon & Boistel (1965) using capillary microelectrodes introduced in the connective between the 3rd and 4th ganglia.

Our recordings from a single giant axon of the postsynaptic phenomena associated with the activity of the cereal fibres in different experimental conditions give a particularly good insight in the integrative properties of the synaptic junctions between cereal fibres and giant axon. The possible connexions between a given postsynaptic giant axon and the main different inputs which seem to be involved can be synthesized as in Fig. 21. This diagram has been elaborated using our electrophysiological findings and the electron microscopical data of Farley & Milbum (1969). It will be used throughout the discussion in which we shall successively analyse the unitary postsynaptic activity and the mode of triggering one postsynaptic action potential, then the factors influencing synaptic transmission and finally the behavioural significance of the synaptic integration in this preparation.

‘ Spontaneous’ EPSPs are in fact related to receptor activity. The careful analysis of the recordings has shown that each EPSP is preceded by a presynaptic spike. These unitary potentials are analogous to those described in the motoneurone of the cat (Kuno, 1964a; Burke, 1967), in the abdominal ganglion of Aplysia (Tauc, 1965), in the 6th ganglion of the crayfish (D. Kennedy, personal communication) and in insect neuropile (Callee & Boistel, 1966a), but one special feature of this preparation is that the afferences are controlled at the receptor level thus giving valuable indications of the precise relationship between a presynaptic spike and one EPSP.

An EPSP in relation with a given receptor varies in amplitude whereas its time to peak remains practically constant. This can be explained by the fact that a single presynaptic fibre might make several synaptic contacts with the dendritic tree of a given postsynaptic axon, some of them being sometimes ineffective. An alternative explanation might involve synaptic contacts more or less remote from the recording site, but this would involve a variation in the time to peak which seems in contradicho with experimental evidence. A statistical analysis on a larger number of recordings has to be carried out for a further understanding of this point.

We have never observed so-called miniature potentials. It is possible that they are too small to be separated from the noise. In some occasions a few EPSPs (0·5–0·9 mV) remain after the ganglion has been completely isolated from all sensory input. It is, however, more likely that they are given by intraganglion interneurones for they also disappear when tetrodotoxin is applied to the whole ganglion (J. J. Callec, unpublished observations).

Unitary IPSPs subsist sometimes after the afferences have been cut. It is thus likely that they have their origin in spontaneous firing of intraganglion interneurones. They have already been recorded in Aplysia (Tauc & Gerschenfeld, 1962) and in motoneuro-nes of the frog (Katz & Miledi, 1963).^* They have been previously observed in some occasions using intraganglion microelectrodes (J. J. Callec, unpublished observations). In fact we have only few data on inhibitory potentials in insect ganglia; they have been recorded in the neuropile (Callec & Boistel, 1966b; Rowe, 1969) or in undetermined cell bodies (J. J. Callec, unpublished observations; Kerkut, Pitman & Walker, 1969). The size and number of the unitary IPSPs in the ‘isolated fibre’ preparation depend upon the quality and the rapidity of the dissection. They are sensitive to oxygen concentration. Furthermore, they can be inverted by a small hyperpolarization which takes place spontaneously after about 2 h.

It seems in our experiments that the depolarization induced by the maximal activity of one receptor stimulated mechanically is not able to reach the threshold. Thus, to trigger an impulse it is necessary to stimulate several receptors simultaneously and strongly.

The electrical stimulation of the cereal nerves shows that there is a great number of homolateral and contralateral sensory fibres connected with a given giant axon, for the EPSP grows progressively with an increase of the stimulation intensity. But in these conditions we can obtain only one action potential, rarely two. The stimulation by a puff of air applied to the cerci gives the longest depolarization which results from the summation of more or less synchronized unitary EPSPs. Nevertheless, it also gives rise to a small number of spikes, generally one or two (maximum six in our experiments). Thus for this kind of integrative synapse it is necessary to summate the action of many sensory impulses to trigger a propagating action potential ; it remains possible, however, that the resting level in vitro is not exactly the same as tn vivo and that, for instance, the already-mentioned slight hyperpolarization which occurs in vitro increases the threshold.

As in the crayfish (see Preston & Kennedy, 1960), when the action potential appears the subjacent EPSP does not disappear. This shows that the site of initiation of the EPSP and of the spike are likely to be differently located (see Fig. 21). All the excitatory and inhibitory afferences (Fig. 21 E, I) are supposed to have their connexions the dendritic tree while the site of initiation of the propagated action potential is thought to be at the beginning of the giant axon near the convergence site of the main dendritic arborizations. This is related to a higher threshold^* or even a complete inexcitability of the thinner ramifications. This geometric arrangement is very well adapted to summate over a wide range a large number of presynaptic influences without giving a spike. These spikes would be blocked at the dilated point of convergence which, in that way, would act like the soma in a vertebrate motoneurone. Similar potentials have also been found in the crayfish (Takeda & Kennedy, 1965), where the anatomical features of the dendritic tree are comparable to those in the cockroach.

In our experiments we have seen that the transmission can be modulated by two principal phenomena: facilitation and depression. These phenomena which have a great importance in the integrative properties of this synapse have been studied at the unitary level by a stimulation of the different kinds of mechanoreceptors. This stimulation was carried out in such a way that it simulated as close as possible the natural patterns of firing.

Facilitation (or potentiation) has been found in many preparations: at the neuromuscular junction (Hubbard, 1963; Braun, Schmidt & Zimmerman, 1966; Mallart & Martin, 1967; Katz & Miledi, 1968; Rahamimoff, 1968; Maeno & Edwards, 1969) or in the central nervous system (Curtis & Eccles, 1960; Kuno, 1964b; Epstein & Tauc, 1970). In the cockroach potentiation seems to be only correlated with phasic receptors, so that their firing, whenever it is brief, is very effective and the threshold of the giant axon is easily reached (see later).

On the other hand the fatigue (or depression) decreases the potentialities of transmission of a coded message or a series of messages. This phenomenon is very frequent. It is often called depression when it follows a previous facilitation (Curtis & Eccles, 1960; Kuno, 1964b) or habituation when one can see a progressive but rapid decrease in the size of the behavioural response. This habituation has been studied by two principal methods ; electrophysiological methods, often at the cellular level such as in Aplysia (Burner & Tauc, 1966; Pinsker et al. 1970), in the crayfish (Krasne, 1969), in the squid (Horn & Wright, 1970), or behavioural methods on the whole animal, as in the cockroach (Zilber-Gachelin, 1966) and in an acridid (Fraser-Rowell & McKay, 1969). Our experiments show that, in the cockroach, this fatigue phenomenon occurs very frequently. It appears during repetitive afferent volleys induced by the mechanical stimulation of receptors. It can be seen as well after the stimulation of a single phasic receptor (Fig. 13) as after the stimulation of a great number of them by puffs of air (Fig. 15). In this last case it has been called EPSP habituation in the text. Repetitive electrical stimulation of afferent fibres (Fig. 20) shows also the same phenomenon complicated by a ‘rebound’ for high frequencies. The analysis of the fatigue phenomenon in these three kinds of experiments has given nearly exponential curves of somewhat comparable size. It can be advanced that in the last two cases we observe summation of the fatigue of individual fibres and that they are all supjxjrted by the same fundamental mechanism. It is also visible after a tonic pattern (Fig. 10) where the postsynaptic depolarization decreases in amplitude even if the frequency remains nearly constant. After a phasic pattern the potentiation wave has been shown to be followed by a depression wave. The curve of Fig. 12 is comparable to the one obtained in the motoneurone of the cat (Curtis & Eccles, 1960). The fatigue is still more striking in polysynaptic pathways (Fig. 18B–D) where EPSPs are very labile and disappear rapidly even at low frequencies.

We think that the fatigue may be due essentially to a depletion of the available transmitter as in the cat motoneurone, but some other factors may also be present, namely, desensitivation of the postsynaptic membrane, modification * at some stage of the secretory process’, relative refractory period in the presynaptic terminals (Bennett, 1968). The last factor has some importance when a receptor gives high-frequency volleys (up to 360 c/s). We therefore know that during refractory period the size of the presynaptic spikes decreases, thus decreasing the quantity of transmitter release for each presynaptic spike (Takeuchi & Takeuchi, 1962). A quantitative study of these phenomena is now being carried out using an artificial (electrical) and well-controlled stimulation of a receptor for a wider range of frequencies. The first hypothesis is reinforced by the observed ‘rebound’ which has already been described by Elmqvist & Quastel (1965) on the human intercostal muscle. These authors have explained this fact by the onset of a transmitter mobilization following the more or less rapid * depletion of a store of transmitter immediately available for release’. This hypothesis has been also mentioned in a recent review (Hubbard, 1970).

Two other limiting factors of less importance have been described in this paper. The first is related to the hyperpolarization component following a burst of great activity coming from a tonic receptor (Fig. 10) or following the mechanical stimulation of the whole cercus (Fig. 16). The second is in correlation with the summation of unitary EPSPs which, during high frequency, drives the resting potential towards the equilibrium potential, thus decreasing their amplitude.

We have seen in this paper that a giant axon in the sixth abdominal ganglion of the cockroach receives many inputs, thus giving rise to a very intense activity consisting in small unitary EPSPs or IPSPs, most of them related to presynaptic spikes in correlation with cereal stimulation. Nevertheless only few postsynaptic spikes are generated. The question then arises: what is the meaning of all this sensory information for a living animal?

The tonic receptors are continuously stimulated by dust, light puffs of air, slight contact with support or other animals, so they provide information about the environment. Their activity is certainly very important and would give a great synaptic depolarization of the giant axon if this was not, in fact, minimized by fatigue or habituation phenomena. It seems that these receptors give essentially a basic tonus and are not able to trigger a propagated action potential.

The phasic receptors, on the other hand, respond to a sudden strong puff of air or greater contacts and induce postsynaptic potentials which may be reinforced by poten-ttion. In these conditions a spike may be generated on the great depolarization which rresponds to the summation of the unitary activities including also the initial component of the tonic response. Then, all the limiting factors (fatigue, habituation, hyperpolarization, …) intervene to shorten the burst of postsynaptic propagated action potentials. The synapses have to be at rest for some time before giving rise again to other spikes. These spikes are rapidly conducted to higher centres (thoracic or cerebral ganglia) where they seem to develop and control the escape response.