The two small white (W) cells in the buccal ganglia of Tritonia hombergi can initiate and modulate cyclic activity in the pattern generating neurones which drive feeding activity in the buccal mass.
They also make extensive monosynaptic connections with the buccal moto-neurones, generating EPSPs on protractor (P) cells, IPSPs on retractor (R) cells, and EIPSPs on the small radula flattener (F) cells. Two F motoneurones receive a chemically mediated, facilitating EPSP from the W cells.
Inactive W cells receive weak excitatory feedback from the pattern generating network interneurones (FPG) in phase I of the feeding cycle and also from some F cells. Prolonged depolarization of one W cell recruits the other. When both are active they adopt a patterned burst mode with a common inhibitory input in phase I.
The neural mechanisms which control the rhythmic cycle of activity shown by the buccal mass and structures associated with the mouth during the capture, ingestion and swallowing of food have been studied in a variety of pulmonate and opistho-branch molluscs (Kater, 1974; Siegler, Mpitsos & Davis, 1974; Woollacott, 1974; Bulloch & Dorsett, 1979a,b; Rose & Benjamin, 1979).
In Tritonia hombergi the buccal mass undergoes three phases of activity during the feeding cycle. The first phase, protraction, is followed by two phases of retraction - radula retraction and flattening – each phase being driven by groups of identified motoneurones in the buccal ganglia called the P, R and F cells (Bulloch & Dorsett, 1979a). Essentially similar sequences have been reported in the closely related species T. diomedea (Willows, 1978) and in the pulmonate Lymnaea (Goldschmeding & DeVlieger, 1975; Rose & Benjamin, 1979), the individual differences between these genera probably reflecting adaptations for dealing with the different physical nature of the food.
In Tritonia hombergi the patterned output of the buccal motoneurones is primarily determined by common inputs from three multi-action interneurones whose activity can be recognized by their characteristic and synchronized synaptic input to the motoneurone groups (Bulloch & Dorsett, 1979b). In Lymnaea similar pattern generating neurones have been identified by direct recording (Rose & Benjamin, 1981). Other sources have been found to influence the buccal motor network either by direct synaptic input or by their action on the pattern generating interneurones (FPGs). These include both mechanoreceptors and chemoreceptors in the anterior gut (Goldschmeding & Jager, 1973; Kater & Rowell, 1973; Audesirk, 1979; Audesirk & Audesirk, 1980; Dorsett & Sigger, 1981). In T. diomedea a presumed buccal motoneurone may also initiate and modulate activity in the pattern generating network (Willows, 1980).
In the search for the pattern generating interneurones in T. hombergi, a number of interneurones were found whose properties could not be fully characterized as part of the pattern generating network yet whose action reliably initiated feeding activity from the motoneurones (Bulloch & Dorsett, 1979b). Among these were two small white cells which are often visible on the dorsal surface of the buccal ganglia. These interact with the motoneurones either by direct synaptic connections, indirectly through the pattern generating interneurones, or both. White cells with some comparable properties have also been reported from the buccal ganglia of T. diomedea (Willows, 1980; Masinovsky, Lloyd & Willows, 1985) and Pleuro-branchaea (Gillette & Davis, 1977; Gillette, Gillette & Davis, 1980, 1982). Other interneurones such as the paired cerebral giant cells also influence the feeding output pattern in T. hombergi and other molluscs (Pleurobranchaea, Gillette & Davis, 1977; Lymnaea, McCrohan & Benjamin, 1980) but are not always active during spontaneous bouts of feeding activity. In Aplysia these cells have been implicated in the arousal associated with feeding behaviour (Kupfermann & Weiss, 1982).
MATERIALS AND METHODS
Details of the head–buccal mass preparation used in these experiments were as in a previous paper (Bulloch & Dorsett, 1979a). The buccal mass and peri-oral region were dissected out and supported on a wax-covered platform inserted down the oesophagus. The buccal ganglia were stabilized by pins pushed through the oesophageal wall into the wax. Apart from cutting the peripheral pleural and pedal nerves, the circumoesophageal ganglionic complex was not disturbed. The white neurones are normally located in the area F3–4 on the topographical map given previously (Bulloch & Dorsett, 1979a). Details of cell morphology were obtained by dye-iontophoresis using microelectrodes filled to the tips with a 4% solution of Lucifer Yellow CH layered with a 1 % solution of lithium chloride. The initial resistance of electrodes used was in the range of 40–50 MΩ. Dye was injected using currents of 10 nA until the cell and axon could be distinguished when viewed through a dark blue filter. Some preparations were left overnight at 5 °C to allow the dye to diffuse. For drawing and photography the ganglia were mounted in 120% sea water and viewed by u.v. light under a Leitz Dialux 20 microscope fitted with a B12 excitation filter and a K515 barrier filter. The hypertonic saline helped to flatten the ganglia by partially dehydrating the living cells and considerably improved the resolution of fine processes arising from the axons. Identification of the white cells was confirmed physiologically in each case by simultaneous recording with a second electrode from the neurones LB8 or RB8 which receive powerful excitatory input from the white cells.
For most experiments intracellular recordings were made from three buccal neurones simultaneously in a combination chosen from the two white cells and a pair of identified motoneurones. Sequences of activity from selected combinations were Stored on separate tracks of a Racal Store 14DS tape recorder. Paired interactions between W cells and motoneurones were routinely checked for electrotonic coupling by the application of hyperpolarizing and depolarizing currents to the soma of each of the pair. Other interactions were considered to be monosynaptic when an EPSP followed the presynaptic impulse 1:1 at frequencies up to 10 s−1, with short but constant latencies of 6–10 ms for ipsilateral combinations and in the region of 12–20 ms for contralateral ganglionic pairings (Fig. 3). These criteria have been considered as strong evidence for monosynaptic connections (Kandel, 1976). Further tests such as the introduction of high Mg2+-low Ca2+ in the bathing solution were not possible in the through-flow sea water bath used in these experiments.
White cell morphology
The white cells are between 20 and 30 μm in diameter. The colouration is variable, ranging from an obvious white to an indeterminate grey tone. The cell lies below the surface and can be difficult to penetrate as the electrode pressure may displace it to one side. Eight white neurones were filled with Lucifer Yellow CH and showed extensive diffusion into the axons. The white cells are bipolar neurones, the axons arising separately but close together on the medial surface of the soma (Fig. 1A,B). The two axons are consistently seen to be of unequal diameter, the larger following the central axon tracts to enter the ipsilateral gastro-oesophageal nerve in which it continues without branching until just below the gastro-oesophageal ganglion. Here it divides, the larger branch entering a nerve which passes to the oesophageal wall and the finer branch continuing through the ganglion to enter the most medial nerve on its distal border (Fig. 1A). The structures innervated by these nerves are not known. Other small branches arise from the main axon just before it enters the gastro-oesophageal nerve, and terminate in an arborization in the neuropile where axons enter and leave the ganglion through the trunk formed by the fused bases of the buccal nerves 2, 3 and the cerebrobuccal connective (CBC).
The smaller diameter axon passes towards the buccal commissure; in three preparations the axon underwent a characteristic flexion in the neuropile some 100 μm from the soma. A number of fine processes arise from this segment of the axon, generally directed posteriorly and towards the areas of the neuropile occupied by the dense dendritic arborization of the motoneurones, (Fig. 2A–C), but extensive branching in this region was seen in only two preparations. On entering the contralateral ganglion the axon passes towards the lateral nerve trunk where it terminates in a small arborization comparable to that of the other side. In all preparations three major branches were seen arising from this axon as it crossed the contralateral neuropile, directed towards and overlapping the areas occupied by the dense dendritic arborizations of the motoneurones grouped along the posterior border of the ganglion (Fig. 2; and see Bulloch, 1977). These processes show both secondary and tertiary branching but the fine arborizations characteristic of the postsynaptic motoneurones were never observed. Two preparations of W cells examined after allowing some 12 h for dye diffusion to occur revealed faint outlines of one or two cell bodies in the opposite ganglion which were apparently dye-coupled to the white cell (Fig. 1A). No corresponding cells were seen on the same side. Dyecoupling has frequently been associated with electrotonic coupling between cells, but numerous recordings from all the accessible surface neurones in the area failed to reveal any that were electrically coupled to the contralateral white neurone.
Morphology of neurones B8, B81
Two F-type neurones in each buccal ganglion respond monosynaptically to impulses in the W cells with large, facilitating EPSPs (Fig. 3A–F). One pair of these neurones corresponds topographically to the identified cells B8, while the other pair B81, located in area D-6 (Bulloch & Dorsett, 1979a, fig. 3), were not previously specified. LB8 and RB8 are monopolar neurones around 100 pm in diameter (Fig. 2A,B). The single axon penetrates the neuropile and turns laterally to enter buccal nerve 2 (BN2) on the same side. Between two and four fine neurites 2–300 μm long arise directly from the soma, passing centrally into the neuropile where they undergo terminal branching to form a three-dimensional array of fine varicose processes whose exact configuration is difficult to resolve (Fig. 2A,B). A number of similar processes arise from the initial 2–300 μm segment of the axon to occupy adjacent areas of the neuropile. It is assumed that these structures represent the principal synaptic areas of the neurones, as little or no branching occurs in more distal regions of the axon within the ganglion. The morphology of LB81 and RB81 generally corresponds to that of the B8 neurones, except the axon divides in the lateral nerve trunk to enter both BN2 and BN3. In this pair the finely divided varicose branches tend to arise from a single thicker stem process adjacent to the main axon (Fig. 2C), and the fluorescent cloud they present in Lucifer Yellow preparations occupies a more central area of the neuropile.
W cell interactions with protractor motoneurones (P cells)
It has been reported that the left white cell (LBW) generates a facilitating EPSP on the ipsilateral P motoneurone LB2 (Bulloch & Dorsett, 1979b). Subsequent recordings from this and other members of the P cell group have confirmed that the P cells 2, 6 and 13 receive 1:1 EPSPs with a short and constant latency from both white cells (Fig. 3H). In LB2 and LB6 the amplitude of the EPSP from the ipsilateral white cell is 0·5–1 mV which summates at frequencies above 3 s−1 but does not markedly facilitate (Fig. 4A,B). LB13 receives a very small EPSP from LBW but at impulse frequencies above 3 s−1 a second intemeurone input appears which closely follows the firing pattern in the W cell, but has a latency of 700 ms (Fig. 4C,D). This is too long for a monosynaptic connection. A prolonged depolarization of the W cell (Fig. 4E) can lead to patterned bursting in P cells which continues after the W cell activity has terminated, resulting from activation of the synaptic inputs associated with the feeding pattern generating interneurones 1 and 1A described previously (Bulloch & Dorsett, 1979a).
Interactions with retractor motoneurones (R cells)
The action of the W cells on members of the R group motoneurones is more complex than that on the P cells. In quiescent preparations individual impulses in the W cells result in monosynaptic IPSPs in B7 and B9 (Figs 3I–L, 5A,B), but in B4 and B5 at spike frequencies above 3 s−1 these summate to produce a 3–5 mV hyperpolarization of the cell membrane and often involve the recruitment of a second inhibitory interneurone (Fig. 5C,E). In each case the short latency inhibition is followed by a depolarization of variable amplitude with a latency of around 2 s, probably due to the activation of interneurone 2 (see below). The R cell LB3 receives a monosynaptic EIPSP from the W cells, repetitive stimulation causing summation of the inhibitory component and a progressive hyperpolarization of 2–3 mV with an enhancement of the EPSP amplitude (Fig. 5D). Increasing levels of hyperpolarizing current applied to LB3 lead to reversal of the slow inhibitory wave (Fig. 5Fi–iii).
The inhibition in R group neurones that follows a short burst of impulses in the W cells is frequently followed by a wave of depolarization that can lead to spiking (Fig. 5D,E). This is probably due to the activation of interneurone 2 in the FPG network, which drives R group neurones during phase II of the feeding cycle. Prolonged firing of a W cell leads to patterned bursting in the R group neurones in which the three types of synaptic input from the FPG network can be resolved (Fig. 5G). Termination of W cell activity is followed by a further two cycles in the pattern generating network which also has synaptic feedback to the W cell.
Interactions with the F group motoneurones
There is now some evidence that the F group motoneurones that fire during phase III of the buccal cycle can be divided into two physiological types. The majority of this group, including neurones B10–19 respond to impulses in the white cells with a small monosynaptic EIPSP having an amplitude of 1 mV (Fig. 3J,K). The inhibitory component summates at frequencies above 2 s−1 to hyperpolarize the cell (Fig. 6A–C). As in the R group, a burst of impulses in the W cells often leads to delayed excitation of the F cells through the FPG network; the synaptic inputs from interneurones 1, 1A and the characteristic long inhibitory wave from interneurone 2 following in their proper sequence after a latency of several seconds (Fig. 6C). Prolonged depolarization of the W cells leads to the development of cycles of patterned bursting in the F group characteristic of the feeding sequence (Fig. 6D,E). At frequencies around 1 s−1 one sometimes observes progressive activation of the FPG intemeurone inputs, with the intensity of the interneurone 2 discharge increasing until its full amplitude is expressed, causing the F cell to fire by post-inhibitory rebound and overcoming the inhibitory influence of the W cell input. The effects of the FPG network on the W cell can also be observed as an inhibition in phase I and a vigorous burst extending from the onset of phase II into phase III (Fig. 6D,E). Normally no evidence is obtained of any feedback between the motoneurones and the W cells. On one occasion a long latency excitation feedback was found between LW and LB 15. Bursts of impulses in the F cell evoked by depolarization, postinhibitory rebound or following interneurone 2 inhibition resulted in 1–2 impulses in the W cell after a delay of 2–4 s (Fig. 6F).
Interactions with F cells R8, R81
The F cells may be separated into two groups on the basis of their interactions with the W neurones. Two symmetrical pairs of F cells L-RB8, L-RB81 respond uniquely to single W cell impulses with large facilitating EPSPs having amplitudes of 1–2 mV (Fig. 7A,C). L8 and R8 are normally so responsive to W cell spikes that it is difficult to obtain isolated EPSPs even when the cells are strongly hyperpolarized; this property is particularly noticeable in the contralateral connections. The neurones L81 and R81 respond to W cell impulses with an EPSP which rapidly facilitates so that 1–2 impulses in either W cell are sufficient to bring both L81 and R81 to spike threshold. Under moderate hyperpolarization, and at a spike frequency of 3 s−1 in the contralateral W cell, the EPSP in R81 facilitates from 1 mV to a peak of 6 mV after seven impulses (Fig. 7A–B). The latency for the W cell EPSP in both pairs of F cells is approximately 7 s for the ipsilateral connections and between 15 and 17 ms for the contralateral pairing. The synaptic inputs from the W cells are 2–3 times larger than those from other unidentified sources. Unlike other pairs of F cells reported previously (Bulloch & Dorsett, 1979b) no evidence was obtained of electrotonic coupling between these two cell pairs in any combination. Despite the dye-coupling that was observed between the W cells and a pair of neurones in the opposite buccal ganglion in positions approximately corresponding to neurones B8 and B81, no electrotonic junctions were found between these neurones and the opposite W cell. An extensive search for coupling to other F cells also proved negative, as did attempts to initiate feeding cycles by depolarizing neurones 8 and 81.
Both B8 and B81 tend to fire spike doublets or triplets resulting from depolarizing afterpotentials (DAPs) in both pairs of cells (Thompson & Smith, 1976; Bulloch & Willows, 1981). In B8 the peak amplitude of the DAP occurs 100ms after the spike and decays exponentially over 2-300 ms (Fig. 7E). A train of impulses in the W cell at 3–4 s−1 can lead to a high frequency burst in B8 with spike ratios of about 4:1 (Fig. 7F).
Prolonged depolarization of a W cell for periods between 15 and 160 ms has two principal actions on the B8 and B81 neurones. The initial effect, due to the direct synaptic pathway, is to generate a high frequency discharge in both cells. After a period of 5–15 s activity, an inhibition occurs in the motoneurone coincident with a slight increase in the firing rate of the W cell (Fig. 7G). The motoneurone then fires a second burst which continues until a further inhibition occurs, this time common to both neurones. The W cell recovers from this inhibition within 1–2 s to resume firing at an increased rate, but the B8 neurone experiences a second large inhibitory wave which has the characteristics of the intemeurone 2 input to these neurones. As the inhibitory influence of interneurone 2 declines, B8 fires its characteristic phase III burst resulting from postinhibitory rebound and the excitatory drive from the W cell (Fig. 7G,H). Interneurone 2 inhibition in the F group following W cell activation often shows a progressive build up in intensity during the first few cycles of FPG activity (Fig. 6D) and occasionally B8 continues to fire in phase II as a result of the strong excitation from the W cells.
When stimulated individually both the B8 and B81 neurones consistently produce movements associated with the contraction of muscle group 4 (inner lip retraction) and muscle groups 5 and 6 (radula retraction and flattening) which occur during phase III of the feeding cycle. Their axons leave the ganglia in nerves which serve these muscles but direct innervation has not been established. As they do not drive any other motoneurones so far recorded, it is provisionally assumed that they are motoneurones serving these muscles.
Interactions between W cells
Simultaneous recordings from both W cells provide no evidence of any monosynaptic pathway between these cells, but they receive common inputs from a number of unidentified sources (Fig. 9C).
Prolonged depolarization of one W cell will activate its partner after a latency of 2–10 s (Fig. 8A–C). The spiking in both cells is periodically interrupted by a simultaneous inhibitory input lasting about 5 s, although this is shorter in the cell being depolarized. The reciprocal excitatory nature of the pathway between the W cells is also emphasized by the observation that the patterned bursting continues for one or two cycles after the depolarizing current in one neurone ceases.
The synchronized burst patterning seen in the W cells depends upon spike activity in both partners (Fig. 10). If one neurone of an active pair is hyperpolarized to prevent spiking or reduce it to frequencies below 2s−1, patterned bursting in the partner is replaced by slow iterative spiking. Upon release from hyperpolarization, burst patterning is resumed by both cells. This mode of activity in the W cells is accompanied by feeding movements of the buccal mass which are more vigorous than those seen during spontaneous buccal cycles when the W cells are silent.
Inputs to the white cells
Evidence has been presented showing that prolonged depolarization of the W cells causes indirect excitation of the buccal motoneurones by activating the FPG network (Figs 5F, 6D,E, 7G,H). Simultaneous recording from a motoneurone and a W cell provides information relative to the timing of two synaptic inputs to the W cells that seem to be derived from the FPG network. In preparations where both W cells are active, there is a periodic inhibition of both neurones in phase I, followed by an acceleration of the firing rate coincident with the onset of phase II. Neither of these events can be elicited by stimulating single motoneurones from any group. W cells do show slight postinhibitory rebound but this does not produce spike frequencies equivalent to those seen during phase II firing.
Concurrent recording from a W cell and a motoneurone during spontaneous FPG cycles where the W cell is inactive reveals that the W cells undergo a 1–2 mV depolarization in phase I at the time of the increase in the firing rate of interneurone 1 and the activation of interneurone 1A (Fig. 9A,B). At the onset of interneurone 2 activity the W cell membrane repolarizes and may temporarily become 1 mV hyperpolarized with respect to its earlier resting level. These fluctuations in membrane potential are common to the W cells and it is improbable that they represent movement artefacts as no disturbance to the recording on the other trace is observed, but no unitary PSPs can be resolved which indicate direct inputs from the multi-action interneurones of the FPG network (Bulloch & Dorsett, 1979b).
The white cells in the buccal ganglia of T. hombergi both initiate and enhance the protraction-retraction cycles in the isolated buccal mass preparation. They do this by exciting the premotor interneurone network which determines the patterned motor output during feeding activity.
In this action they resemble other white buccal neurones present in Pleurobranchaea and T. diomedea (Gillette & Gillette, 1983; Willows, 1980). Although analogous in their function, considerable variation is found in the morphological and physiological characteristics of the white cells in the three species. In T. diomedea the Bll and B12 neurones are unipolar with an axon in the ipsilateral gastro-oesophageal nerve, only the B12 cell having a contralateral branch which makes the synapse with the B5 motoneurone. The white cells of Pleurobranchaea are also unipolar with a dendritic field close to the soma, but the axon crosses to the gastro-oesophageal nerve on the opposite side. In T. hombergi the neurone is bipolar with a well developed system of branches in both ganglia.
The synaptic properties of the white neurones in the three species are equally divergent. In Pleurobranchaea the two cells are strongly coupled electrotonically, yet do not synapse with the motoneurones. In neither species of Tritonia are the cells coupled, but in T. hombergi they make extensive chemical synaptic contacts with the motoneurones, whereas only electrical coupling has been detected with unspecified motoneurones at this level in T. diomedea. The large facilitating excitatory synapses made with the F cells B8 and B81 are not reported from T. diomedea but there is a biphasic synapse on the motoneurone B5, an F cell which has the ability to initiate cyclic feeding activity in this species.
Relationships with the FPG network
The nature of the synaptic contacts of the W cells on the buccal motoneurones of T. hombergi is at first sight inconsistent with their observed firing during phase II–III in some spontaneous buccal cycles. At this time they are co-active with the R cells (which they inhibit) and out of phase with the P cells (which they excite). However their spike frequency during these cycles does not exceed 2 s−1, which might be expected to only marginally influence the excitability of the motoneurones other than B8 and B81. From these observations one might conclude that the principal action of the W cells is on the interneurones of the FPG network.
Unitary synaptic inputs from the FPG interneurones cannot be resolved in the W cells, but the simultaneous ramp depolarizations that are seen in non-spiking W cells during phase I of spontaneous feeding cycles (Fig. 9) are not seen at any other time, and appear to be derived from phase I interneurone activity, indicating some form of positive feedback from the FPG network. This input can give rise to occasional spikes in the W cell during phase I and may be followed by similar activity in phase III as a result of feedback from some F cells.
In some preparations the W cells are more active than in others and exhibit a high level of synaptic activity and spontaneous spiking associated with continuous feeding movements in the buccal mass. In other quiescent preparations the W cells are silent but depolarizing currents of around 1 nA injected into the soma lead to spiking at around 2 s−1 which eventually recruits the partner in the opposite ganglion, followed by activation of the common inhibitory input. This sequence is associated with the excitation of the FPG network monitored by the characteristic synaptic inputs of these interneurones in the motoneurones. A similar transition from quiescence to synaptically driven activity is also seen in Pleurobranchaea white cells following the stimulation of the oral veil with appetitive substances (Gillette & Gillette, 1983).
A model summarizing the main features of the W cell interactions is given in Fig. 11. The reciprocal excitatory connection between the W cells is provided by the centres ‘e’, two of which are provided as both W cells must be active to excite the common inhibitory input The activity of the inhibitory input to the W cells (‘i’) is phase-linked to the FPG interneurone 1A, but is separated by the finding that spontaneous FPG cycles are not accompanied by phase I inhibitory input in silent W cells. It is envisaged that the centres ‘e’ would gate the action of ‘i’.
The W cells may excite the FPG interneurones through two routes. Excitation of interneurone 1A could result from the activation of ‘i’, but a more probable route, suggested by the fact that the FPG inputs are normally elicited in their proper temporal sequence, is by the excitation of interneurone 1. If this connection was reciprocal it would also provide the source of the early phase I depolarization seen in inactive W cells (see Fig. 9).
One pair of white cells in T. diomedea (Bl 1) contain large amounts of a small cardioactive peptide (SCPB) together with acetylcholine (Masinovsky et al. 1985). The other pair (B12) only contain the peptide. It may be anticipated that the W cells in T. hombergi are also peptidergic, but the anatomy suggests a closer homology to the B12 neurones which have a branch to the contralateral ganglion. No release sites for the peptide have been identified in the ganglionic sheath or along the gastro-oesophageal nerve and their action outside the ganglion is not known.
In Pleurobranchaea, appetitive substances applied to the oral veil result in depolarization and prolonged burst episodes in the ventral white cells (VWC) which drive the buccal motor network (Gillette & Gillette, 1983). Direct stimulation of the VWC also produces a transient contraction of the longitudinal muscle of the oesophagus. Gillette & Gillette suggest that the VWCs have a command role in the feeding behaviour of Pleurobranchaea, the criteria for which are not met in the present observations on T. hombergi.
There is considerable functional resemblance between the W cells of T. hombergi and the slow oscillator (SO) neurone in the buccal ganglion of Lymnaea (Rose & Benjamin, 1981). Direct synaptic input from the SO to the N1 neurones excites the FPG interneurones in their proper sequence, while the SO activity becomes phase-locked to the buccal cycle by inhibitory feedback from the interneurones. In Lymnaea the inhibition of SO occurs during the retraction phases of the cycle. The SO also makes monosynaptic connections to buccal motoneurones.
The feeding movements of the buccal mass of Tritonia are more vigorous, and the period between individual cycles is least when the W neurones are co-active with the FPG network. Thus the W neurones can both initiate and modulate feeding activity, but their activity is not necessary for its expression in the isolated buccal mass. At present, no special function, apart from their motor activity, can be assigned to the buccal neurones B8 and B81. Their unique synaptic relationship to the W cells argues for some special role in the neural circuit associated with the behaviour, and will be the subject of further investigation.