A recent claim that the giant fibre of the hermit crab excites its contralateral motor giant neurone through a chemical rather than an electrical synapse (Stephens, 1986) was re-examined. We found that the reported increased latency (relative to the electrical ipsilateral synapse) was postsynaptic in origin, as was the increased spike ‘jitter’. There was no difference in synaptic latency between the electrical synapse and the supposed chemical one. We did not find a consistent resistance to N-ethylmaleimide (an uncoupler of electrical synapses) by the supposed chemical synapse, but the synapse was resistant to 2 mmol l−1 cadmium, which blocks known chemical synapses in the system. Sub-threshold depolarizing current passed from the presynaptic giant fibre to the postsynaptic contralateral motor giant, and hyperpolarizing current passed antidromically. We conclude that the svnapse is electrical and not chemical in nature.

A striking feature of the neuronal circuitry underlying the escape tail-flip of crayfish is that the giant fibres (GFs) activate all of the excitatory abdominal flexor motor neurones by way of electrical synapses (for a review see Wine & Krasne, 1982). Hermit crabs too have an escape behaviour (rapid withdrawal into their mollusc shells) which is homologous in many respects to that of the crayfish (Chapple, 1966). There is a single pair of GFs which are homologous to the crayfish medial giant (Chapple & Hearney, 1974; Stephens, 1985). Each GF drives the ipsilateral member of a pair of large neurones, the segmental giants (SGs), through a rectifying electrical synapse (Heitler & Fraser, 1986). The SG in turn excites the fast flexor (FF) motor neurones, also probably through electrical synapses. There is one large flexor motor neurone which usually receives suprathreshold input directly from the ipsilateral GF via a rectifying electrical synapse. This is the homologue of the crayfish motor giant (MoG) neurone (Umbach & Lang, 1981). The hermit crab MoG, unlike that of the crayfish, also receives input from the SGs (like the unspecialized FF neurones), but this input is usually subthreshold (Heitler & Fraser, 1986).

It has recently been reported that in the hermit crab the GF excites the MoG contralateral to itself by way of a chemical synapse (Stephens, 1986). In our own previous work we noted the existence of this contralateral excitatory path, but tacitly assumed that the excitation was mediated electrically. If the excitation is indeed chemical, it would be the only known case of functional excitatory chemical output from a crustacean GF, and the only known case where an FF motor neurone received its primary GF excitation via a chemical synapse.

The evidence adduced to support the claim of chemical transmission was fourfold.

  1. The contralateral GF-MoG synaptic delay was longer than the ipsilateral GF—MoG delay.

  2. There was greater latency jitter with contralateral excitation and, with repeated stimulation, the contralateral excitation failed much more quickly than the ipsilateral excitation.

  3. Application of N-ethylmaleimide (NEM; an uncoupler of electrical synapses; Spray el al. 1984) disrupted ipsilateral but not contralateral excitation.

  4. The dye Lucifer Yellow was transferred from the GF to the ipsilateral MoG but not the contralateral one.

This evidence is certainly consistent with chemical transmission, but is by no means conclusive proof. We have re-examined transmission between the GF and the contralateral MoG and, although some of our results are similar to those of Stephens, we suggest that they are as consistent with electrical as with chemical transmission. We support this by presenting additional data, including demonstration of the direct passage of current between the two neurones. We conclude that the MoG receives excitation from the contralateral GF through an electrical synapse which is similar to, but weaker than, the electrical synapse which connects it to the ipsilateral GF.

Hermit crabs (Eupagurus bemhardus) were collected locally from St Andrews bay and kept in tanks of circulating sea water. Prior to dissection a crab was removed from its mollusc shell by carefully crushing the latter in a vice. The abdomen was cut from the anterior body, and the fourth ganglion was dissected free from the abdominal ventral nerve in association with its roots and interganglionic connectives. The nerve cord was pinned dorsal surface upwards on a Sylgard platform, and submerged in saline (Atwood & Dorai-Raj, 1966, but with Tris substituted for NaHCO3). The saline was kept chilled to 8°C. Where required, 5 mmol l−1.N-ethylmaleimide (NEM) and 2mmoll−1 cadmium chloride were added to the saline without adjusting the other ionic components.

Hook and pin electrodes were used to record extracellularly and to stimulate the connectives and the first and third roots of the fourth abdominal ganglion. Intracellular recordings were made with glass microelectrodes (initial resistance 20–40 MΩ) filled either with potassium acetate or with 5 % Lucifer Yellow dissolved in 1 mol l−1 lithium chloride. Penetrations were all made from the dorsal aspect of the ganglion into axonal or neuropile processes. Prior to penetration the ganglionic sheath was first softened by applying a few specks of protease (Sigma Type XIV) for about 60s, and then partially removed. In some experiments a computer-based signal averager was used to reduce noise (mainly spontaneous activity) in the response to stimulation. Neurones were injected with Lucifer Yellow using 0·5-s negative current pulses of 10–100 nA delivered at 1 Hz for a period of 30 min to 3 h. Tissue was fixed in 4 % formaldehyde in saline, dehydrated in alcohol and cleared in methyl salicylate. Preparations were examined in whole mount with epi-illumination (Zeiss filter set 06), and drawn using a camera lucida attachment.

Abbreviations

Abbreviations used in the text are as follows: GF, giant fibre; MoG, giant motor neurone; SG, segmental giant neurone; G4, fourth ganglion of the abdominal nerve cord; AC, interganglionic connectives anterior to G4; PC, interganglionic connectives posterior to G4; Rl, first root innervating swimmerets and containing the SG axon; R3, third root innervating the flexor muscles and containing the MoG axon; i, ipsilateral; c, contralateral; 1, left; r, right; x, extracellular stimulation (see below).

In the text, laterality of neurones will usually be referenced as ipsilateral or contralateral to the axon of the MoG in question (e.g. iGF for the GF ipsilateral to the MoG axon and contralateral to its cell body, cGF for the other GF). In the figure legends an exact designation (left or right, e.g. 1AC, rAC) is given where possible. No consistent difference was observed between the neurones on the left side of the ganglion and those on the right.

Figure conventions

To avoid confusion, certain conventions have been adhered to throughout the figures and their legends. The artefacts caused by extracellular stimuli are all marked with filled curved arrows, while the duration of intracellular current pulses is marked with dots. Where a giant fibre is activated by extracellular stimulation of its axon in one of the connectives, this is termed a stimulus applied to the GFx.

Current injection

The neurones into which current was injected had a low input impedance, and the Lucifer Yellow-filled microelectrodes had a tendency to block during injection of depolarizing current, which meant that in order to pass sufficient current to achieve significant changes in membrane potential we often had to bypass the recording amplifier and apply voltages (10–100 V) directly into the microelectrode. Thus, in these cases we were not able to monitor the membrane potential of the injected neurone. The bath potential was stabilized using a virtual-ground current monitor with separate voltage-recording and current-passing indifferent electrodes (Purves, 1981). With this configuration there was little or no artefactual voltage change in the recording microelectrode when large currents were passed, other than transients at the onset and offset of the pulse. (See Fig. 7F for physiological evidence of the lack of coupling artefact.)

A current pulse is described as bracketing an event when the event fell within the duration of the current pulse.

Is there a connection between the GF and contralateral MoG?

In an earlier study (Stephens, 1985) no connection was observed between the cGF and the MoG, although a connection was described between the MoG and a contralateral non-GF through-conducting axon (Stephens, 1985). This conclusion was not in agreement with our own preliminary work (which was a factor in initiating the present study), but was subsequently modified when a chemical synapse from the cGF to the MoG was described (Stephens, 1986). The first task in the present report is thus to confirm the existence of the cGF—MoG connection.

Physiology

Extracellular recording shows that in most preparations stimulation of the cAC induces a MoG spike at exactly the cGF threshold (Heitler & Fraser, 1986). Intracellular electrodes were used to test the necessity and sufficiency of a cGF spike for MoG activation (Fig. 1). Depolarizing current injected into a GF would cause it to spike, and this usually resulted in a spike in both MoGs. Conversely, if the GF was induced to spike by extracellular stimulation of the AC, a pulse of negative current injected into the GF could block its spike. When the GF spike failed, the MoG spikes also failed. Thus a cGF spike is sufficient to activate the MoG, and there is no axon in the AC which activates the MoG and has a threshold less than or identical to that of the GF.

Anatomy

Lucifer Yellow stains of the MoG (Fig. 2) usually revealed two groups of short dendrites sprouting from either side of the neurite, which were suitably located to make contact with the GFs on their respective sides. The dendritic group on the side of the MoG axon was usually slightly posterior to the other group. Thick sections of preparations in which the MoG was stained with horseradish peroxidase (HRP) showed that the two groups of dendrites did indeed make close contact with their respective GF axons.

We did not observe dye-coupling to either MoG after injecting Lucifer Yellow into a GF axon, even after injection times of several hours. However, the staining was not as intense as that produced by a similar regime in a crayfish GF, in our experience.

Latency of GF-MoG interaction

When the two GFs were stimulated with extracellular electrodes which had been carefully positioned on the ACs to be equidistant from the ganglion, there was a, greater latency of response recorded extracellularly in the R3 contralateral to the stimulated AC compared to that in the ipsilateral R3, as reported previously (Stephens, 1986). However, intracellular recording from the MoG showed that much of this latency difference was caused by a change in the conduction latency between the peak of the spike recorded intracellularly, and the MoG spike recorded extracellularly in R3 (Fig. 3A,B). This suggests that an iGF spike initiates a spike in the MoG at a more distal location than a cGF spike, thus shortening the conduction distance to the extracellular recording electrodes. Since in all experiments the microelectrodes were located close to the synaptic sites this in turn suggests that the iGF excites the MoG more powerfully than the cGF. Further evidence for this was obtained by simultaneous intracellular recordings from the MoGs. When care was taken to penetrate these neurones at exactly the same anterior-posterior location there was no detectable difference in the latency from a stimulus exciting either AC to the start of the MoG response (Fig. 3C,D). There was a slight difference in the latency from the stimulus to the peak of the intracellular spike, with the MoG spike stimulated by the cAC delayed relative to that stimulated by the iAC. This is clearly due to the faster rise time of the latter, again suggesting that the MoG receives more powerful iGF than cGF excitation. This was confirmed by stimulating the AC at high frequency to induce MoG spike failure, and examining the amplitude of the residual EPSP. No matter which AC was stimulated, the MoG ipsilateral to the stimulus had a larger EPSP than that contralateral to the stimulus (Fig. 3E,F).

Thus we conclude that the latency difference between the extracellular response of the MoG to iGF and cGF stimulation is due to the excitation that the MoG receives from the iGF being more powerful and having a faster rise time than that it receives from the cGF, and is not due to a difference in the actual synaptic delay.

Synaptic jitter

When either GF was stimulated extracellularly at relatively high frequency (50–80 Hz) there was a gradual increase in the latency of MoG spikes recorded extracellularly in R3, followed by eventual spike failure, as described above. This occurred much more rapidly, and at lower frequencies of stimulation, for the cGF than for the iGF (Fig. 4). The effect has been noticed previously (Stephens, 1986) and attributed to the MoG receiving chemical excitation from the cGF and electrical excitation from the iGF. Our intracellular recordings from the MoGs during high-frequency AC stimulation showed that the increasing latency was mainly due to a progressive delay in the time of occurrence of the spike relative to the start of the underlying EPSP (Fig. 4A,B). The difference in latency jitter with iGF and cGF stimulation can thus be explained by a gradually rising spike threshold in the MoG combined with the difference in amplitude and rise time of the underlying EPSPs. In our experiments the effect was reversible, provided the stimulation was not too prolonged (paceStephens, 1986).

A second effect was noticed with high-frequency cGF stimulation. If the stimulation was maintained after MoG spike failure, there was a progressive decrease in the amplitude of the residual EPSP to a baseline minimum. There was also some decrease in response to iGF stimulation, but this was much less marked (Fig. 4A,B). Prominent anti-facilitation of this sort is generally characteristic of chemical synapses, where it is a presynaptic phenomenon caused by a reduction in transmitter release. However, such effects could also be generated postsynaptically, for instance, by a gradual decline in membrane resistance. To gain further information on this point the two GFs were stimulated alternately with pulse trains of gradually increasing frequency. Eventually spike failure and a reduction in EPSP amplitude occurred in the MoG, similar to that occurring when the cGF alone was stimulated. Using these paired stimuli the cGF frequency at which failure occurred was about half that required when the cGF alone was stimulated. The stimulation to the iGF was then switched off, so that the total stimulation received by the MoG was halved, but the frequency of cGF-MoG synaptic activation did not change. This caused a rapid increase in the amplitude of the EPSP in the MoG, and a return of the MoG spike (Fig. 4C). The effect was the exact reverse of that seen with high-frequency cGF stimulation. This strongly suggests that the decline in EPSP amplitude with high-frequency stimulation is not due to anti-facilitation at a chemical synapse, since it can be induced by paired stimuli to different presynaptic neurones, and reversed without any change in presynaptic spike frequency of the supposed chemical synapse. Furthermore, addition of 2 mmol l−1 cadmium ions to the bathing medium significantly increased the duration and/or frequency of presynaptic stimulation required before MoG spike failure and EPSP anti-facilitation occurred. This was true for both cGF stimulation alone, and for the paired cGF-iGF stimuli, suggesting that a common mechanism underlies the anti-facilitation under the two stimulus conditions. This is certainly not what would be expected at a chemical synapse (see below for more details).

XEM application

A’-ethylmaleimide (XEM) is reputed to be an uncoupler of electrical synapses (Spray et al. 1984). Like Stephens (1986), we found that XEM caused overall depolarization, and both the GF and MoG neurones eventually ceased to respond to direct extracellular stimulation. However, in some cases the EPSP caused by iGF stimulation was reduced in amplitude and eventually abolished while the GF spike was still maintained (Fig. 5). In these cases the EPSP caused by cGF stimulation decremented in a parallel fashion. We found no consistent difference in the effects of XEM on the response of the MoG to iGF and cGF stimulation, unlike Stephens (1986).

We do not wish to place too much emphasis on these results because of the variability introduced by the direct effects of XEM on membrane potential and spike viability, but in our experiments the evidence, in so far as it goes, supports the proposal that both MoGs receive their input from the GF by electrical synapses.

Effects of cadmium ions

If the MoG receives its input from the cGF through a chemical synapse, the input ought to be blocked by substances, such as cadmium ions, which block calcium channels and hence presynaptic transmitter release. A convenient control exists in this system because there are axons in both the ACs which have thresholds close to the GF threshold and which elicit depolarizing IPSPs in the MoG. It seems reasonable to assume that these are homologous to the IPSPs induced in the crayfish MoG (Wine, 1977), which are chemical in origin and caused by an increase in chloride conductance (Ochi, 1969), although this has not been tested explicitly in the hermit crab.

To test the effects of cadmium ions on cGF-MoG transmission the iAC was stimulated at an intensity which induced a depolarizing IPSP in the MoG. In different preparations this intensity could be either above or below the GF threshold. In a preparation where the IPSP was elicited with an iAC stimulus just below the iGF threshold, the cAC was also stimulated, this time at exactly the cGF threshold. When this stimulus was timed to coincide with the peak of the IPSP, the cGF-induced MoG spike was abolished, leaving only a small residual EPSP (Fig. 6A,B). Saline containing 2 mmol l−1 cadmium was then applied to the system. Over a period of about 10min the IPSP (induced by iAC stimulation) diminished in amplitude, while the EPSP (induced by cGF stimulation) grew, and eventually gave rise to an MoG spike (Fig. 6C,D). Thus the chemically induced inhibition was abolished, while the cGF activation was not. This strongly suggests that the cGF activates the MoG through an electrical synapse rather than a chemical one.

Current coupling

Perhaps the most direct physiological test for electrical coupling is the ability to pass current from one neurone to the other. Depolarizing current injected into the cGF does indeed depolarize the MoG (Fig. 7B), while hyperpolarizing current has no effect. Depolarizing current injected into the MoG has no effect on the GF. The simplest explanation for these results is that the two neurones are connected by a rectifying electrical synapse. However, interpretation is complicated by two factors. First, both the cGF and the MoG tended to spike as a result of the current injection (Fig. 7C). Second, a similar effect could be obtained at a chemical synapse if the presynaptic neurone released transmitter in a graded manner in response to subthreshold depolarization. A more satisfactory test is to hyperpolarize the postsynaptic MoG, and see whether the hyperpolarization spreads to the presynaptic cGF. The experimental procedure was to stimulate the cGF extracellularly in the cAC, and bracket the stimulus with a pulse of hyperpolarizing current injected into the MoG. With gradually increasing levels of current the first effect observed in extracellular recordings was the failure of the MoG spike in iR3 (Fig. 7D). This was followed by the simultaneous failure of the cGF spike recorded extracellularly in the PC (not shown), and the cMoG and cSG spikes recorded extracellularly in cR3 and cRl, respectively (Fig. 7E). Simultaneous intracellular recording from the cGF showed that as the increasing hyperpolarizing current was injected into the MoG, the cGF hyperpolarized until its spike eventually failed (Fig. 7D-F). The absolute value of the peak voltage of the cGF spike remained approximately constant as the cGF membrane potential was increased, indicating that this hyperpolarization was not a coupling artefact.

The passage of hyperpolarizing current from a postsynaptic neurone to a presynaptic one strongly suggests that the two are connected by electrical synapses. However, it does not necessarily mean that the connection is monosynaptic. There is a disynaptic electrical pathway from the cGF to the MoG via the cSG, since the latter receives electrical input from its ipsilateral GF and makes electrical output to both MoGs (Heitler & Fraser, 1986, and see below). Thus hyperpolarizing current could pass from the MoG to the cSG and hence to the cGF. There are a number of reasons why this is unlikely to be a major pathway. First, the electrical EPSP produced in the MoG by cSG stimulation is nearly always subthreshold, and usually quite small (2–8mV; Heitler & Fraser, 1986), suggesting that this pathway is not strong. Second, when hyperpolarizing current was injected into the MoG, the cSG spike did not fail at lower current levels than the cGF spike, as might be expected if the current was passing through the cSG on its way to the cGF (Fig. 7D,E). Third, when sufficient hyperpolarizing current was injected into the MoG to block the cGI and cSG spikes, a simultaneous recording from the cSG showed that it only hyperpolarized by about 8 mV, compared to the 40 mV hyperpolarization recorded in the cGF in another experiment (Fig. 8A compared to Fig. 7F). This 8mV was insufficient to block a cSG antidromic spike initiated by stimulating cRl (Fig. 8B), which required at least 20 mV hyperpolarization of cSG. Finally, in one aberrant preparation, the iGF failed to connect to the MoG, leaving only the disynaptic pathway iGF–iSG–MoG (Fig. 8C). The iSG-MoG connection was strong, producing an EPSP of about 13 mV, but even 1000 nA of hyperpolarizing current injected into the MoG failed to block the iGF spike, and only produced an iGF hyperpolarization of about 7 mV (Fig. 8D). Since the iSG—MoG synapse is usually stronger than the cSG—MoG synapse, these results suggest that the disynaptic cGF—cSG—MoG connections are not a significant pathway for the passage of hyperpolarizing current from the MoG to the cGF.

The synapse between the GF and MoG in crayfish is a classic example of an electrical synapse; indeed it was the first such synapse ever demonstrated (Furshpan & Potter, 1959). The suggestion that in hermit crabs the cGF makes a chemical synapse (Stephens, 1986) with the MoG is therefore of considerable interest. The aim of this report is twofold; first, to confirm that the hermit crab cGF does indeed activate the MoG, since there is some doubt in the literature about this (Stephens, 1985, 1986; Heitler & Fraser, 1986) and, second, to determine whether this activation is by a chemical synapse as claimed by Stephens (1986). The evidence for cGF activation of the MoG is clear-cut. Depolarizing current injected into a cGF to induce it to spike almost always induces a MoG spike. Conversely, if the cGF is induced to spike by extracellular stimulation and then injected with sufficient hyperpolarizing current, its spike is blocked and there is simultaneous spike failure of the MoG.

The question of whether the cGF—MoG synapse is chemical or electrical is more complex. Stephens worked on the first abdominal ganglion of Pagurus pollicaris (which is fused to the thoracic ganglionic mass), while we have been using the fourth (free) abdominal ganglion of Eupagurus. We do not think that this difference in preparations is significant, however, since our raw data are largely in agreement with those of Stephens. The difference lies in our interpretation of these data. We have extended the observations made by Stephens, and also performed new experiments, which lead us to conclude that the synapse is in fact electrical, not chemical.

Is the synapse electrical?

The most convincing physiological proof of the existence of an electrical synapse between two neurones is the demonstration of the direct passage of current from one neurone to the other. We show that depolarizing current injected into the cGF spreads to the MoG, while hyperpolarizing current injected into the MoG spreads to the cGF. If sufficient hyperpolarizing current is injected into the MoG it can actually cause cGF spike failure. The only known disynaptic path from the cGF to the MoG is via the SG, and simultaneous microelectrode penetrations have shown that this pathway is not strong enough to account for the observed effects.

Supplementary evidence for the nature of a synapse can be obtained from its response to synaptic blocking agents. The cGF–MoG synapse is resistant to the application of 2 mmol l−1 cadmium ions. In contrast, the depolarizing I PSP which is activated in the MoGs by extracellular stimulation of unidentified units in the anterior connectives is abolished by this treatment, showing that cadmium does indeed block at least some chemical synapses in this system.

NEM, which is a blocker of electrical synapses, was applied to the preparation to try to distinguish between cGF and iGF activation of the MoG. In contrast to Stephens we found that NEM could block the cGF–MoG synapse, although the results were hard to interpret because of the direct effects of NEM on membrane potential and spike production. We found no consistent resistance of the cGF-initiated EPSP compared to that initiated by the iGF, and on at least some occasions the responses declined in parallel.

Thus, on the basis of direct current coupling, resistance to known blockers of chemical synapses and susceptibility to blockers of electrical synapses, we conclude that there is a strong electrical component to the cGF–MoG synapse.

Is the synapse also chemical?

The first line of evidence adduced to suggest that the cGF–MoG synapse is chemical was that the MoG spike recorded extracellularly has a greater latency to cGF stimulation than to iGF stimulation. We confirm that this latency differential exists, but show that it is due to a difference in delay between the initiation of the EPSP in the MoG and the passage of the spike past the extracellular recording electrodes on R3. This is partly due to the slower rise time of the cGF-initiated EPSP causing a delay in MoG spike initiation, and partly due to the cGF-initiated MoG spike having a greater conduction delay to the extracellular recording electrodes (presumably due to differences in the site of spike initiation). The overall latency difference is not due to a difference in synaptic latency.

We also confirm that high-frequency cGF stimulation causes greater MoG spike jitter and earlier spike failure than iGF stimulation. This is explained in part by the difference in amplitude and rise time of the EPSP underlying excitation from the two sources, and in part by the greater degree of anti-facilitation shown by the cGF-induced EPSP. This sort of anti-facilitation is a common characteristic of chemical synapses. However, similar anti-facilitation can be produced by paired stimuli to both the GFs. Under these conditions the cGF needs only to spike at about half the frequency that is required to produce the effect when the cGF alone is stimulated. This suggests that the important factor is the net stimulation received by the MoG, rather than the cGF spike frequency perse. An alternative possibility is that two separate phenomena are occurring — anti-facilitation when the cGF alone is stimulated, and some sort of presynaptic inhibition of the cGF by the iGF when paired stimuli are used. It seems unlikely that these two processes would exactly mimic each other, and it also seems unlikely that they would both be substantially reduced by the presence of cadmium ions. In particular, if the anti-facilitation is the result of a chemical synapse, the cadmium ions should tend to reduce the amplitude of the EPSP, rather than augment it as actually happens.

The cause of the decrement has not been investigated in detail, but one possibility is that it is caused by a reduction in postsynaptic membrane resistance due to corollary discharge neurones making chemical inputs to the MoG which are blocked by cadmium. Another possibility is that there is normally a cumulative increase in postsynaptic calcium-activated potassium conductance which is blocked by cadmium.

Stephens (1986) found Lucifer Yellow dye-coupling between the iGF and the MoG but not between the cGF and the MoG, while in our experiments we were unable to obtain dye-coupling between either GF and the MoG. However, even if present, differential dye coupling could be explained by the difference in the strength of electrical coupling which we report, and not necessarily by a difference between electrical and chemical coupling. The absence of dye-coupling is a notoriously unreliable indicator for the absence of electrical coupling (Audesirk, Audesirk & Bowsher, 1982; Powell & Westerfield, 1984).

We conclude that there is no positive physiological evidence for a significant chemical component in the cGF-MoG synapse. It is, of course, extremely difficult, if not impossible, to prove that a minute chemical component does not exist, but, in the absence of any evidence that it does exist, we suggest that at present the data indicate that the cGF-MoG synapse is purely electrical in nature.

This work was supported by a SERC grant to WJH. We thank Dr R. Pitman for helpful discussion and critically reading the manuscript. We thank Dr W. Stewart for his gift of Lucifer Yellow.

Atwood
,
H. L.
&
Dorai-Raj
,
B. J.
(
1966
).
Tension development and membrane responses in phasic and tonic muscle fibers of a crab
.
J. cell. comp. Physiol
.
64
,
55
72
.
Audesirk
,
G.
,
Audesirk
,
T.
&
Bowsher
,
P.
(
1982
).
Variability and frequent failure of Lucifer Yellow to pass between two electrically coupled neurons in Lvninaea stagnalis
.
J. Neurobiol
.
13
,
369
375
.
Chapple
,
W. D.
(
1966
).
Asymmetry of the motor system in the hermit crab Pagurusganosimanus Stimpson
.
J. exp. Riol
.
45
,
65
81
.
Chapple
,
W. D.
&
Hearney
,
E. S.
(
1974
).
The morphology of the fourth abdominal ganglion of the hermit crab: a light microscope study
.
J. Morph
.
144
,
407
420
.
Furshpan
,
E. J.
&
Potter
,
D. D.
(
1959
).
Transmission at the giant motor synapses of the crayfish
.
J. Physiol., Lond
.
145
,
289
325
.
Heitler
,
W. J.
&
Fraser
,
K.
(
1986
).
The segmental giant neurone of the hermit crab F.upagurus bertthardus
.
J. exp. Biol
.
125
,
245
269
.
Ochi
,
R.
(
1969
).
Ionic mechanism of the inhibitory postsynaptic potential of crayfish giant motor fiber
.
Pflügers Arch. ges. Physiol
.
311
,
131
143
.
Powell
,
S. L.
&
Westerfield
,
M.
(
1984
).
The absence of specific dye-coupling among frog spinal neurons
.
Brain Res
.
294
,
9
14
.
Purves
,
R. D.
(
1981
).
Microelectrode Methods for Intracellular Recording and lonophoresis
.
London
:
Academic Press
.
Spray
,
D. C.
,
White
,
R. L.
,
Campos De Carvalho
,
A.
,
Harris
,
A. L.
&
Bennet
,
M. V. L.
(
1984
).
Gating of gap junction channels
.
Biophys. J
.
45
,
219
230
.
Stephens
,
P. J.
(
1985
).
Morphological and physiological properties of the giant interneuron of the hermit crab (Pagurus pollicaris)
.
J. Neurobiol
.
16
,
361
372
.
Stephens
,
P. J.
(
1986
).
Electrical and chemical synapses between giant interneurones and giant flexor motor neurones of the hermit crab (Pagurus pollicaris)
.
J. exp. Biol
.
123
,
217
228
.
Mbach
,
J. A.
&
Lang
,
F.
(
1981
).
Synaptic interaction between the giant interneuron and the giant motorneuron in the hermit crab, Pagurus pollicaris
.
Comp. Biochem Physiol
.
68
,
49
53
.
Wine
,
J. J.
(
1977
).
Neuronal organization of crayfish escape behavior: inhibition of giant motoneuron via a disynaptic pathway from other motoneurons
.
J. Neurophysiol
.
40
,
1078
1097
.
Wine
,
J. J.
&
Krasne
,
F. B.
(
1982
).
The cellular organization of the crayfish escape behaviour. In The Biology of Crustacea
, vol.
111
,
Neural Integration
(ed.
H.
Atwood
&
D.
Sandeman
), chapter 15.
New York
:
Academic Press
.