1. When the proboscis of a hemichordate worm is prodded, two or more nerve pulses travel along the ventral cord to trigger the large, compound muscle potential that precedes the startle response.

  2. Nerve pulses also may be compound or they may be all-or-none spikes. Compound pulses decay in size as they travel from their point of origin, but spikes are generally through-conducted to the posterior end and are seen most often during the repetitive discharge evoked by a strong stimulus.

  3. Repetitive discharge produces facilitation of conduction velocity so that a burst of closely spaced spikes is generated.

  4. The same pulses that trigger the startle response initiate waves of retreat peristaltic contractions, but in the latter activity there is less summation of the spike-like muscle potentials.

  5. Although conduction of the waves toward the anterior end depends on the presence of the ventral cord, waves that follow the initial contraction are not preceded by nerve spikes.

In burrowing or tube-dwelling animals one type of behaviour that can be evoked consistently is a rapid withdrawal from a strong stimulus. In many species this startle response is associated with giant fibres in the nervous system (Bullock & Hor- ridge, 1965). Some hemichordates have giant fibres, but members of the enteropneust family Ptychoderidae have none; large fibres grade into small. Nevertheless, these worms exhibit the same rapid contractions as do other hemichordates, and studies have shown that the ventral cord is a major conduction pathway in the initiation of this behaviour (Bullock, 1940; Knight-Jones, 1952). A few years ago a technique was developed to record potentials from the ventral cord and underlying muscles (Pickens, 1970). What follows is an analysis of the part these potentials play in the startle response and other retreat behaviour.

Specimens of Ptychodera flava, a tropical hemichordate found in Kaneohe Bay, Oahu, Hawaii, were examined at the Institute of Marine Biology, University of Hawaii, or shipped to the University of Arizona. Prior to the recording of potentials a worm was transferred to a flat pan with sea water at 22·27 °C and a harness of pins was placed across its collar in order to eliminate excessive movement of the electrodes. Gross differences in startle behaviour between free-moving and restrained animals were looked for but none was found even though sensory inputs must have been different in each case. Nerve and muscle potentials from whole animals or from worms that were slit along the dorsal cord were recorded with suction electrodes, a.c.-pre-amplifiers, and other standard instruments (Pickens, 1970).

Retreat behaviour

When a worm is lying in a pan and is tapped on the proboscis, the anterior third of the trunk, from the posterior edge of the collar back to the hepatic region, contracts rapidly. Two to six seconds after this startle response retreat peristalsis begins. The latter consists of a series of waves, called antiperistaltic waves, that first start in the hepatic region and pass toward the collar. Subsequent waves begin at the posterior end. In contrast, a worm in its burrow does not begin retreat peristalsis until 30–60 sec after the startle response and the waves are initiated at the posterior end rather than in the middle of the animal. Retreat peristalsis results from the sequential contraction of fibres in the longitudinal muscle layer, the only one that is well developed in enteropneusts.

Light taps given to the proboscis are the weakest mechanical stimuli that will evoke startle responses and retreat peristalsis. Much stronger prods have to be applied to the collar and anterior trunk in order to produce the same effects. The proboscis is sensitive to electric shocks as well, but voltages required are generally ten times greater than those that evoke nerve spikes when the stimulating electrode is on the ventral cord. No matter what the stimulus, the recovery period for the startle response is about 5 min (Bullock, 1940). A second stimulus of the same strength given before complete recovery will initiate retreat peristalsis, local movements of the proboscis, or on rare occasions, another startle response.

Nerve spikes

When the proboscis is tapped and a recording is made of activity in the ventral cord and underlying muscles, one or more spikes are seen to precede a large, compound potential (Fig. 1). The latter appears to originate from muscle, but the initial spikes are considered to be nerve pulses because they can be recorded only from the ventral cord and not lateral to it (Fig. 5 A), are not conducted beyond a cut in the cord, are seen during fatigue or habituation of the startle response when there is no movement around the recording electrode (Fig. 4), are slow to disappear if the worm is anaesthetized, and resemble nerve spikes evoked by electrical stimulation of the ventral cord (Fig. 1C and Pickens, 1970).

Fig. 1.

Nerve spikes, startle potentials and a local muscle potential. When the proboscis of a hemichordate worm is tapped (A) or receives an electric shock (B), brief nerve pulses appear in the ventral cord and are followed by prolonged, compound potentials originating from adjacent muscle fibres. Similar potentials occur when shocks are applied directly to the ventral cord (C). The positions of stimulating and recording electrodes (in mm from the proboscis tip) are indicated on the line drawing of Ptychodera, Paired traces show activity recorded by one proximal and one distal electrode. The latter was located in the hepatic region of the trunk where nerve spikes travelling posteriorly appear with variable latency. Startle contractions are absent in this region so that retreat peristalsis does not begin until a few tenths of seconds after the spikes arrive, e.g. shortly after the first arrow in (A). The stimulus artifact is present in (C) but occurred in (B) 120 msec before the first spike in the upper trace and is not shown. The local muscle potential (Pickens, 1970) following the nerve spike in (C) precedes a non propagated contraction and is always smaller than the muscle potential recorded during a startle contraction (A, B). Nerve pulses other than the first one or two are not easily identifiable because they sum with muscle potentials. However, two delayed spikes in response to a strong prod are indicated by arrows in (A).

Fig. 1.

Nerve spikes, startle potentials and a local muscle potential. When the proboscis of a hemichordate worm is tapped (A) or receives an electric shock (B), brief nerve pulses appear in the ventral cord and are followed by prolonged, compound potentials originating from adjacent muscle fibres. Similar potentials occur when shocks are applied directly to the ventral cord (C). The positions of stimulating and recording electrodes (in mm from the proboscis tip) are indicated on the line drawing of Ptychodera, Paired traces show activity recorded by one proximal and one distal electrode. The latter was located in the hepatic region of the trunk where nerve spikes travelling posteriorly appear with variable latency. Startle contractions are absent in this region so that retreat peristalsis does not begin until a few tenths of seconds after the spikes arrive, e.g. shortly after the first arrow in (A). The stimulus artifact is present in (C) but occurred in (B) 120 msec before the first spike in the upper trace and is not shown. The local muscle potential (Pickens, 1970) following the nerve spike in (C) precedes a non propagated contraction and is always smaller than the muscle potential recorded during a startle contraction (A, B). Nerve pulses other than the first one or two are not easily identifiable because they sum with muscle potentials. However, two delayed spikes in response to a strong prod are indicated by arrows in (A).

Under most conditions a just-suprathreshold stimulus produces two nerve spikes (Fig. 1 B). A single, large compound spike is rarely seen (Fig. 1 A). A tap weak enough to elicit only one small spike evokes no startle response (Fig. 2A). Strong prods elicit several spikes, some of which appear to originate in conduction pathways that fire repetitively. In many recordings the interval between successive spikes decreases (Figs. 2D, 3) resulting in a burst of closely spaced spikes, and this suggests that facilitation of conduction velocity has taken place. On the other hand, during repetitive firing facilitation of spike amplitude is rare (Fig. 2D) and pulses normally have the appearance of all-or-none spikes. A burst usually produces a larger, more prolonged contraction, although the difference in size between a compound muscle potential following two spikes and one following a burst is not great. Nevertheless, larger muscle potentials appear to be associated with a decrease in the intervals between nerve spikes and longer contractions with an increase in the number of spikes.

Fig. 2.

Response to mechanical stimuli prior to and after removal of the proboscis and collar. A weak prod given to the proboscis evokes no startle response (A). A very strong prod given to the proboscis produces several nerve spikes and a muscle potential (B). After the proboscis and collar are removed, a stimulus of the same strength applied to the anterior trunk evokes a shorter burst of nerve spikes and a much smaller muscle potential (D). Typical nerve and muscle potentials are seen when the ventral cord is given an electric shock either before or after removal of the anterior end (C). The recording electrode was 15 mm from the posterior end of the collar. Note the decrease in interval between spikes in (D) and the small increase in spike height.

Fig. 2.

Response to mechanical stimuli prior to and after removal of the proboscis and collar. A weak prod given to the proboscis evokes no startle response (A). A very strong prod given to the proboscis produces several nerve spikes and a muscle potential (B). After the proboscis and collar are removed, a stimulus of the same strength applied to the anterior trunk evokes a shorter burst of nerve spikes and a much smaller muscle potential (D). Typical nerve and muscle potentials are seen when the ventral cord is given an electric shock either before or after removal of the anterior end (C). The recording electrode was 15 mm from the posterior end of the collar. Note the decrease in interval between spikes in (D) and the small increase in spike height.

Typically, nerve spikes are large and compound when recorded near the anterior end and small near the hepatic region or posterior end (Fig. 1). The decrease in size is due partly to the break-up of the compound pulses into smaller ones as they travel at different velocities and partly due to lability of transmission, particularly in the hepatic region. However, some pulses at the anterior end of the cord are all-or-none and are conducted without decrement. On rare occasions two such spikes of distinctly different shape may be seen (Fig. 3). The anatomical correlates of these spikes are unknown as the cord has a rather uniform appearance under the electron microscope and there are no glial cells to insulate the fibres from one another (Dilly, Welsch & Storch, 1970).

Fig. 3.

Activity in two independent conduction pathways recorded during a startle response. A prod given to the proboscis evoked potentials that were picked up by a pair of electrodes placed 17 mm apart on the ventral cord. The fourth spike (arrows) is of shorter duration than the first three and is considered to arise in a different conduction pathway. The conduction velocity in this tract is 30 cm/sec. It is apparently not influenced by conduction in a nearby pathway which is facilitated as a result of repetitive firing; that is, the first spike is conducted at 30, the second at 50, and the third at 90 cm/sec. Smaller compound muscle potentials than usual are seen because a prod was given prior to complete recovery of the startle response.

Fig. 3.

Activity in two independent conduction pathways recorded during a startle response. A prod given to the proboscis evoked potentials that were picked up by a pair of electrodes placed 17 mm apart on the ventral cord. The fourth spike (arrows) is of shorter duration than the first three and is considered to arise in a different conduction pathway. The conduction velocity in this tract is 30 cm/sec. It is apparently not influenced by conduction in a nearby pathway which is facilitated as a result of repetitive firing; that is, the first spike is conducted at 30, the second at 50, and the third at 90 cm/sec. Smaller compound muscle potentials than usual are seen because a prod was given prior to complete recovery of the startle response.

Fig. 4.

Activity in anterior and posterior parts of the ventral nerve cord elicited by prodding the proboscis at 30 sec intervals. Only responses to the 5th, 7th, 9th, 12th, 13th, and 14th prods are shown. The 1st prod evoked a typical startle response and subsequent prods produced few or no muscle potentials. Nerve spikes in the posterior cord (lower trace of each pair) vary in number and time of appearance, suggesting that integration is occurring in the hepatic region separating anterior and posterior portions of the cord. Electrodes were 20 mm apart, equidistant from the centre of the hepatic region.

Fig. 4.

Activity in anterior and posterior parts of the ventral nerve cord elicited by prodding the proboscis at 30 sec intervals. Only responses to the 5th, 7th, 9th, 12th, 13th, and 14th prods are shown. The 1st prod evoked a typical startle response and subsequent prods produced few or no muscle potentials. Nerve spikes in the posterior cord (lower trace of each pair) vary in number and time of appearance, suggesting that integration is occurring in the hepatic region separating anterior and posterior portions of the cord. Electrodes were 20 mm apart, equidistant from the centre of the hepatic region.

Fig. 5.

Bioelectric activity occurring during retreat peristalsis. Nerve and muscle potentials accompanying a startle response (A) are followed by muscle potentials associated with a forward-travelling, antiperistaltic wave. The upper trace is a recording from an electrode on the ventral cord, the lower from another placed 1 mm lateral to it. Nerve spikes are absent from the lower trace. In both traces there is less summation of individual muscle potentials during retreat peristalsis than during the startle contraction, but otherwise there are very few muscle spikes in one trace that can be matched up with spikes in the other. Retreat peristaltic potentials follow the startle response and travel forward at 1–2 cm/sec as shown in (B) in which a pair of electrodes were placed 20 mm apart on the anterior ventral cord. The first burst is similar to (A) but compressed in time. The second wave occurs after 5 sec and the interval between subsequent waves increases. Note that the peak of the potentials during the startle contraction in (B) seems to occur sooner at the posterior electrode (lower trace) than at the anterior.

Fig. 5.

Bioelectric activity occurring during retreat peristalsis. Nerve and muscle potentials accompanying a startle response (A) are followed by muscle potentials associated with a forward-travelling, antiperistaltic wave. The upper trace is a recording from an electrode on the ventral cord, the lower from another placed 1 mm lateral to it. Nerve spikes are absent from the lower trace. In both traces there is less summation of individual muscle potentials during retreat peristalsis than during the startle contraction, but otherwise there are very few muscle spikes in one trace that can be matched up with spikes in the other. Retreat peristaltic potentials follow the startle response and travel forward at 1–2 cm/sec as shown in (B) in which a pair of electrodes were placed 20 mm apart on the anterior ventral cord. The first burst is similar to (A) but compressed in time. The second wave occurs after 5 sec and the interval between subsequent waves increases. Note that the peak of the potentials during the startle contraction in (B) seems to occur sooner at the posterior electrode (lower trace) than at the anterior.

Excitation triggering the startle response in the anterior half of the worm is transmitted to the posterior half of the ventral cord. In worms that have not been disturbed for several minutes the pulses are through-conducted to the posterior end on what appear to be a one-to-one basis. With continued stimulation different numbers of pulses occur in the posterior cord at irregular intervals after the stimulus (Fig. 4). Since it is not possible to produce the same force each time in applying prods, some of the variability is due to the inequality of mechanical stimuli. Nevertheless, there is indication that the hepatic region of the cord has integrative properties, although the evidence is not as clearcut as when the ventral cord is stimulated electrically (Pickens, 1970).

There are at least two situations in which a burst will not trigger a startle response. The first is during the period after a startle response when the animal is refractory (Fig. 4). The second is when subthreshold prods are applied and their strength gradually increased. If this experiment is performed carefully, the number of spikes per prod increases, but no startle response occurs. In both cases the muscle fibres are not fatigued because they contract if an antiperistaltic wave passes forward. Consequently, the rise in threshold probably takes place in the nerve fibres of the cord if the assumption is made that there is not dual innervation of muscle fibres.

Nerve spikes elicited by a prod to the proboscis are only rarely recorded from the dorsal cord, and a cut across the cord does not alter the latent period for the burst of muscle potentials. The normal conduction pathway for the initiation of the startle response, therefore, is from collar cord to ventral cord. This confirms deductions based on behavioural evidence (Bullock, 1940; Knight-Jones, 1952). Stroking the proboscis is ineffective ; the surface must be deformed. Hence proprioceptors, not tactile receptors, trigger the response. By slicing away pieces of the proboscis it can be shown that the low threshold for the response is still present when nothing more than the proboscis stalk remains in front of the trunk. Once the stalk is removed, the threshold increases markedly. Receptors in the trunk can still provide appropriate input for a modified startle response, so that a worm without a proboscis and collar will contract rapidly in response to a prod, but the response is weak (Fig. 2D).

The nerve pulses preceding the startle response are usually no more than half the size of those evoked by a strong shock to the cord (Fig. 2). Again, this seems to indicate that excitation triggering a startle response does not spread throughout all the fibres of the cord. On the other hand, when a shock is applied to the cord and a maximum nerve spike is evoked, the local muscle potential is not as large as the compound muscle potential that precedes the startle response (Fig. 1). It is not known why electrical stimulation of the cord will not produce the large muscle potential associated with a propagated muscle contraction.

Muscle potentials

Compound potentials with numerous peaks can be recorded from any part of the trunk when muscles contract, but not when they relax. These are considered to be muscle action potentials. They can be distinguished from nerve potentials in that they disappear in a short period of time when the proboscis is prodded continuously or when a spontaneously active worm is anaesthetized with magnesium chloride. The individual pulses that make up the compound potentials have a spike-like appearance (10–15 msec in duration compared to 3–7 msec for nerve spikes), but it is not known whether they are all-or-none or graded potentials (Fig. 5). No two groups of muscle potentials recorded sequentially from the same spot exhibit the same detailed pattern, nor will two electrodes placed 1 mm apart record pulses of similar height and duration.

Differences between compound potentials preceding a startle response and those preceding retreat peristalsis, based on the characteristics of individual muscle spikes, have not been found, and it is concluded that the same muscle fibres contract in both types of behaviour. However, the compound potentials occurring during the startle response are large and arise simultaneously in all parts of the anterior trunk while those preceding retreat peristalsis are of smaller size and occur sequentially from hepatic region to collar (Fig. 5). The differences in the two types of contractions are believed to be due to differences in the organization of nerve fibres within the cord. Dual innervation of muscle fibres appears unlikely as very few nerve fibres have been seen to run from cord to muscles (Bullock, 1945; Dilly et al. 1970). Consideration has been given to the possibility that contractions in these two types of behaviour differ only in the rate of recruitment of muscle fibres; in other words, the startle response is an antiperistaltic wave that travels very rapidly from hepatic region to collar. Violent, antiperistaltic contractions such as this have been seen in the larva of the hemichordate, Saccoglossus (Burdon-Jones, 1952), but proving that the same sort of contraction occurs in an adult Ptychodera has been difficult as only a small number of recordings (Fig. 5B, for example) show a compound potential that reaches its peak in the hepatic region before it reaches its peak close to the collar. Most muscle potentials occur simultaneously in all parts of the anterior trunk or, in worms that have not fully recovered from a startle response, near the collar before they occur near the hepatic region (Fig. 3). Furthermore, the maximum velocity for an antiperistaltic wave is about 4 cm/sec, much slower than the 1 or 2 m/sec required for the nearly synchronous contraction seen during the startle response.

As has been shown before (Crozier, 1915; Knight-Jones, 1952; Bullock, 1940), a cut across the ventral cord blocks both the startle response and retreat peristalsis. In contrast to the earlier results, however, it was found that antiperistaltic waves can travel around a cut under certain conditions, presumably by way of the lateral nerve net. The first wave reaching a cut stops, but the next crosses it after a short delay.

The conduction velocity through this alternate pathway is one-fifth of that along the cord (0·4 vs. 2·1 cm/sec in one experiment). Each succeeding wave crosses after shorter and shorter delays until retreat peristalsis progresses as if no cut existed. The facilitating effect through the alternate pathway is labile and when retreat peristalsis is initiated again after a quiescent period the delay at the cut is apparent once more. There may be some polarity in this pathway because peristaltic, or backward travelling, waves were never seen to cross the cut. Differences between these and earlier results may be due to differences in the extent of the cuts across the ventral cord.

Although the ventral cord is the normal pathway for conduction of excitation during retreat peristalsis, no recordable nerve spikes precede this excitation. The velocities of conduction of these waves range from 0·1 to 4 cm/sec, but velocities during any one recording period do not seem to change appreciably; that is, velocity is independent of the interval between waves. The intervals range from 3 to 23 sec and may be regular or irregular. Shocks to the cord during retreat peristalsis decrease the interval and increase the regularity, but the waves cannot be driven at fixed intervals by electrical stimulation.

In a whole worm retreat peristalsis usually continues for about 30 sec after the startle response. Then the proboscis resumes its exploratory or burrowing activity with the result that retreat peristalsis stops first in the anterior and then in the posterior part of the worm. In worms that are restrained the waves may continue for more than 1 h at an average rate of 3–4 waves/min. Cuts that either remove the proboscis and collar completely, or separate the collar nerve cord from the trunk cords, cause contractions in the trunk to stop. If retreat peristalsis is initiated again by prodding, the waves continue for longer periods of time than when the cords are intact (5 min compared to 30 sec in one experiment). This suggests that there is some inhibitory control of the trunk cords by the nervous system of the proboscis or collar. The ventral cord appears to be important also in co-ordinating slow activity in all parts of the trunk. For example, a wave can be made to move toward the hepatic region from each end by shocks applied at the centre of the trunk, but the wave travelling forward dominates after the two waves meet and only retreat peristalsis is seen from that point on. However, if the ventral cord is cut beneath the stimulating electrodes before the shock is given, waves travelling in opposite directions may persist for some time.

Responses to a predator

Ptychodera comprises a large percentage of the diets of two species of gastropods, Conus lividus and C. leopardus, living in the sand flats of Kaneohe Bay (Kohn, 1959). Accordingly, a number of observations were made of the interaction between C. lividus and Ptychodera. The difference in behaviour between predator and prey is striking. The cone appears to be keenly aware of a hemichordate on the opposite side of the aquarium and moves steadily toward its prey. Its activity is increased when mucus from the worm is placed near its siphon. When the cone reaches Ptychodera, it explores the worm with its proboscis and attempts to engulf the collar or trunk, avoiding the mucus-covered proboscis. Normally it does not shoot out a toxic dart, but was seen to do so on one occasion. The dart had no effect on the worm even though it can paralyse other marine animals.

Ptychodera apparently does not recognize its predator until it is attacked. Mucus from the cone elicits no change in the worm’s behaviour nor does the worm retreat when Conus is exploring the surface of its body. In fact, Ptychodera will explore the shell and soft parts of the cone with its own proboscis until the first bite is taken. As soon as this occurs, the anterior half of the worm contracts rapidly once and then retreat peristalsis begins at the tip of the trunk. Recordings show that the muscle potentials evoked during these two events have the same appearance as they do in an animal restrained by pins. Antiperistaltic waves occur every 3 sec and continue even after proboscis, collar, and one half of the trunk have been engulfed by the cone.

A startle response is triggered when proprioceptors in the proboscis are distorted. Excitation is transmitted to muscles presumably by at least one set of internerurones interposed between sense cells and motoneurones. The pulses recorded from the ventral cord during the startle response are thought to arise from these internerurones. Similar pulses can be evoked without triggering a startle response and these produce either retreat peristalsis or no movement at all. Because retreat peristalsis depends on the integrity of the cord and, in all probability, the same muscle fibres are used in both aspects of retreat behaviour, it is concluded that neuromuscular junctions are not refractory right after the startle response, but rather it is the junctions between internerurones and motoneurones that are probably the most labile of any along this conduction pathway. It is most likely that habituation of the startle response takes place at these junctions and that facilitation may be a requirement as well (Pickens,1970)

The appropriate code for pulses triggering the startle response is not known as it is difficult to evoke a typical response by applying to the ventral cord electric shocks that closely simulate the intervals between and duration of the natural pulses. However, it is known that one pulse is insufficient, that two are effective, and that more than two elicit a slightly greater or more prolonged contraction. Strong stimuli evoke a surfeit of pulses and it appears unlikely that all contribute to the startle response. Nerve pulses that precede a startle response may be compound or all-or-none; they are either through-conducted or die out with distance from the point of origin; and they will be repeated if the stimulus is intense. The decay of a compound pulse is not unusual and occurs as the pulse breaks up into smaller components that travel at different conduction velocities. It could be due also to the presence of synapses along the ventral cord that fail to transmit under certain conditions. All-or-none spikes, whether they occur singly or during repetitive firing, are usually through-conducted to the posterior end of the worm even though the startle contraction occurs only in the anterior half. They must trigger the first antiperistaltic wave from the posterior tip of the trunk. No further nerve pulses are seen preceding each antiperistaltic wave but the interval between waves can be quite regular, suggesting that pacemakers are present in the cord. Through-conduction, decay of excitation with distance, repetitive firing, facilitation and pacemaking are all properties of nerve nets and the fact that these are also properties of the ventral cord supports the suggestion that the cord is a net that has been compressed or collapsed (Horridge, 1968). However, the cord differs from a nerve net in that there does not appear to be diffuse spread laterally within the cord:repetitive pulses remain at the same size when they are conducted down the cord, facilitation of conduction velocity in one channel does not affect the velocity in another, and one tract may fire repetitively while another does not. Therefore, it appears that the presence of a cord in a simple nervous system not only decreases the conduction time for information passing from one point to another but also through some unknown mechanism allows functionally separate tracts to exist side by side.

The anatomical basis for such tracts is not known although there is no evidence that the tracts are localized (Pickens, 1970). Similar conclusions have been reached with regard to the radial cords of echinoderms (Sandeman, 1965; Millott & Okumura, 1968), a group that is functionally if not phylogenetically closely related. Cobb (1970) proposes a model for the nervous system of asteroids and echinoids in which the radial nerve is made up of tracts that form connexions between areas of neuropile. Although neuropile has not yet been located in the cords of hemichordates, Cobb’s model could be applied to the hemichordate nervous system as well. Particularly attractive is his proposal that loose, relatively non-specific synapses are formed between ectoneural nerve fibres and muscles that are separated by a connective tissue sheet. If a similar situation exists in hemichordates, this might explain why very few nerve fibres have been seen to cross the basal lamina between nerve cord and muscle cells (Bullock, 1945; Dilly et al. 1970).

The hepatic region might be one area of the ventral cord in which neuropile is prevalent. Excitation is conducted through this region without modification, or changed prior to transmission, or blocked completely. Conditions for the transmission of pulses are difficult to define, but the hepatic region is well defined in terms of certain types of behaviour. For example, the startle contraction takes place anterior to this region, retreat peristalsis first starts here in many cases, antiperistaltic waves begin at the posterior end several seconds after they do in this area, and peristaltic or antiperistaltic waves can occur in the anterior trunk without similar activity occurring in the posterior trunk, and vice versa. On the other hand the hepatic part of the cord is not indispensable for the initiation of antiperistaltic waves, since a cut across the ventral cord anterior to it produces a new pacemaker just in front of the cut and retreat peristalsis will start from this point after a short delay. This apparent plasticity is probably due to the presence of a multiplicity of pacemakers along the cord with the more posterior ones having the lower thresholds.

It has been postulated that the same muscle fibres contract during the startle response and retreat peristalsis. While most muscle potentials accompanying each aspect of behaviour are variable in size and duration, many are of the same height and have a spike-like appearance. It is tempting to suggest that these effectors, like photocytes (Baxter & Pickens, 1964), may react in an all-or-none manner.

This study was initiated at the Institute of Marine Biology, University of Hawaii, where Drs Vernon Brock, Philip Helfrich, and Ernst Reese were kind enough to provide space and equipment. My thanks are extended also to several others in Hawaii who collected and shipped worms to the University of Arizona. Dr Alan Kohn was helpful in identifying the species of Conus used and providing considerable data on the biology of this snail.

Baxter
,
C. H.
&
Pickens
,
P. E.
(
1964
).
Control of luminescence in hemichordates and some properties of a nerve net system
.
J. exp. Biol
.
41
,
1
14
.
Bullock
,
T. H.
(
1940
).
The functional organization of the nervous system of Enteropneusta
.
Biol. Bull. mar. biol. Lab., Woods Hole
79
,
91
113
.
Bullock
,
T. H.
(
1945
).
The anatomical organization of the nervous system of Enteropneusta
.
Q.J. microsc. Sci
.
85
,
55
111
.
Bullock
,
T. H.
&
Horridoe
,
G. A.
(
1965
).
Structure and Function in the Nervous Systems of Invertebrates
.
San Francisco
:
W. H. Freeman and Co
.
Burdon-Jones
,
C.
(
1952
).
Development and biology of the larva of Saccoglossus horsti (Enteropneusta)
.
Phil. Trans. R. Soc. B
236
,
553
90
.
Cobb
,
J. L. S.
(
1970
).
The significance of the radial nerve cords in asteroids and echinoids
.
Z. Zellforsch. mikrosk. Anat
.
108
,
457
74
.
Crozier
,
W. J.
(
1915
).
The behavior of an enteropneust
.
Science, N.Y
.
41
,
471
2
.
Dilly
,
P. N.
,
Welsch
,
U.
&
Storch
,
V.
(
1970
).
The structure of the nerve fibre layer and neurocord in the enteropneusts
.
Z. Zellforsch. mikrosk. Anat
.
103
,
129
48
.
Horridge
,
G. A.
(
1968
).
The origins of the nervous system
.
In The Structure and Function of Nervous Tissue
, vol.
1
(ed.
G. H.
Bourne
).
New York
:
Academic Press
.
Knight-Jones
,
E. W.
(
1952
).
On the nervous system of Saccoglossus cambrensis (Enteropneusta)
.
Phil. Trans. R. Soc. B
236
,
315
54
.
Kohn
,
A. J.
(
1959
).
The ecology of Conus in Hawaii
.
Ecol. Monogr
.
29
,
47
90
.
Millott
,
N.
&
Okumura
,
H.
(
1968
).
The electrical activity of the radial nerve in Diadema antillarum Philippi and certain other echinoids
.
J. exp. Biol
.
48
,
279
87
Pickens
,
P. E.
(
1970
).
Conduction along the ventral nerve cord of a hemichordate worm
.
J. exp. Biol
.
53
.
515
28
.
Sandeman
,
D. C.
(
1965
).
Electrical activity in the radial nerve cord and ampullae of sea urchins
.
J. exp. Biol
.
43
,
247
56
.