The Nervous Pathways for Poisoning, Eating and Learning in Octopus

When a strange figure falls in the visual field of an octopus the animal usually attacks it after a considerable delay. If the attack yields food then on a subsequent appearance of the figure the attack will be repeated. Within a few trials the octopus learns to attack the figure regularly and quickly. If conversely no food results, attacks become slower and soon cease (Young, 1956, 1958). The present work was under-taken to find the pathway by which the signals that teach the animal to attack reach the memory, which is in the optic lobes. There seem to be three possible pathways (labelled 1, 2 and 3 in Text-fig. 1): (1) from the lips that surround the beak, via the labial nerves and superior buccal lobe ; (2) from the receptors in the buccal mass or farther down the gut, via the sympathetic nerves, inferior buccal ganglia, interbuccal connectives and superior buccal lobe; (3) from the arms, via the brachial nerves and either the cerebro-brachial connective or by a more posterior pathway, the brachio-optic lobe tract (Boycott & Young, unpublished).

The resultant attack might be due to impulses in a combination of these pathways. It has not been possible to investigate them all equally thoroughly, but evidence will be given that the learning can take place if the lip has been removed or the labial nerves severed, but not if the interbuccal connectives have been cut. Pathways from the arms or lips alone are not therefore able to produce the learned responses.

The details of the connectives of the inferior frontal and superior buccal systems are considered elsewhere (Young, 1965a, b). Besides analysing the pathways mentioned above, the present experiments also serve to demonstrate that the bundles of fibres originating from the back of the superior buccal lobe carry the impulses for stimulation of the posterior salivary gland secretion which poisons the prey (crabs) (Ghiretti, 1960). Further it is shown that the connectives joining the superior buccal lobe to the inferior buccal ganglia must be intact if the animal is to open the crab properly and remove the meat from the skeleton.

The plan of the experiments has been to interrupt the pathways at a series of levels, from the peripheral nerves in front, backwards through the superior buccal and posterior buccal lobes.

Octopuses were isolated and kept in asbestos tanks 100 × 60 × 40 cm. with circulating sea water and a home of bricks at one end (see Boycott & Young, 1955). Crabs or white horizontal or vertical rectangles were introduced at the end of the tank opposite the home and the time taken for the octopus to attack was recorded with a stop-watch. The food reward for attacking the horizontal rectangle was a small piece of fish given on the end of a wire. An electric shock (12V. a.c.) was given if the vertical rectangle was attacked.

Trials were given in sessions in the morning and evening, each session consisting of five to ten trials, trials being separated by about 5 min. Operations were performed under urethane anaesthesia (3 % in sea water) and sections were prepared with Cajal’s silver method as previously described.

The lip is a ring of tissue, around the beak, surrounded by the arms. It contracts when touched, even under light anaesthesia, and can only be removed if the animal is deeply anaesthetized. The operation was performed on seven animals and none showed any obvious aberrations of behaviour or feeding thereafter. They attacked crabs seen at a distance and seized, poisoned and ate them apparently like normal animals.

The attack behaviour of animals was tested at three sessions before and three after operation by showing crabs, which they were not allowed to eat. The proportion of attacks was as great after operation as before in several animals, but in a few it fell slightly giving the means shown in Text-fig. 2.

A white horizontal rectangle was then shown to each octopus and a small piece of fish was given whether or not the object was attacked. The effect of giving food after the first trial increased the probability of attack at the next one, as in normal animals (Young, 1956). All the animals came out several times during the first sessions of eight trials and several continued to attack regularly thereafter (Text-fig. 2). In the animal of Text-fig. 3 a further extension of the experiment was to show the animal a vertical rectangle and to give shocks when this was attacked. Food continued to be given at alternate trials when the horizontal rectangle was shown. Attacks on the latter at first decreased sharply but by the fifth session the discrimination was made accurately.

It is concluded that the lip is not essential for the promotion and maintenance of attacks at an unfamiliar visual figure.

Nerve trunks of four types arise from the front of the superior buccal lobe (Text-fig. 1). (1) there are some fifteen labial nerves; (2) at the sides there is a pair of nerves that run directly to the posterior salivary glands ; (3) a pair of interbuccal connectives join the superior buccal lobe to the inferior buccal ganglion; (4) a pair of cerebro-subradular connectives run close to the interbuccal connectives but by-pass the inferior buccal ganglion and run direct to the subradular ganglion, which controls the papilla at the end of the posterior salivary gland duct (Young, 1965b).

These bundles of nerves can be reached where they leave the brain by opening the cranium from above, removing the jelly and extending the incision forwards, if necessary until it enters the buccal venous sinus. The labial nerve bundles near to the mid-line can be cut easily, but it would be difficult to cut those at the sides without damaging the salivary nerves or the connectives. Conversely, the salivary nerves and connectives have not been cut without damaging the more lateral labial nerves, but the more central labial nerves can be left intact after the salivary nerves and connectives have been cut.

In several operations attempts were made to cut the salivary nerves but leave the connectives, or vice versa. This was successful only in a few animals, which are sufficient however to establish the functions of the nerves. No attempt has been made to cut the interbuccal and cerebro-subradular connectives separately.

The operation of opening the buccal sinus and gaining access to the nerves does not in itself impair the power of the animal to attack, kill and eat crabs, nor to learn to attack an unfamiliar figure. In several animals the operation had failed to cut the required trunks and these activities were unimpaired (Table 1). In octopus MCN there was no damage to the salivary nerves or connectives but all the labial nerves in the mid-line had been cut, sparing only the labial nerves that run with the connectives at the side (Pl. 1, fig. 1). This animal attacked, killed and cleaned crabs on the day after operation and later consistently attacked a horizontal rectangle.

In octopus MCM one interbuccal connective was cut and also the labial nerves at the sides, but the salivary nerves were intact. This animal behaved exactly as the previous one. In MCL one posterior salivary nerve had been damaged though not completely cut (Pl. 1, figs. 2, 3). On the day after operation it caught crabs but did not poison them. On the second and subsequent days, however, the crabs were poisoned and then fully cleaned. The animal responded regularly to the rectangle, although many of the labial nerves had been cut.

In two further animals, MDP and MDQ, one salivary nerve and one interbuccal connective was left intact (or nearly so). They both showed some imperfections in killing and eating crabs. MDP only succeeded in poisoning the crab on about half of the occasions tested, but was able to remove the meat well from those it did kill. MDQ (Pl. 1, fig. 4) poisoned the crabs regularly but sometimes left their carapaces imperfectly cleaned. Both animals eventually attacked the horizontal rectangle, although never so regularly as the first three mentioned (Table 1). A normal octopus leaves the carapace and endophragmal skeleton of the crab quite free of flesh. These operated animals frequently left some meat in the anterior (enclosed) part of the carapace and they failed to remove the gills from the endophragmal skeleton (Pl. 2, fig. 9). Moreover, normal octopuses separate the joints of the larger appendages of the crabs and remove their contents, presumably by suction and/or with the radula. The operated animals left the appendages unseparated.

Octopus MCX (Pl. 1, figs. 5–7) was an animal in which both posterior salivary nerves were cut and the interbuccal and subradular connectives were cut on one side. It was unable to poison the crabs but would attack and catch them and could kill them, apparently by tearing off the carapace. The meat was then removed reasonably efficiently, though sometimes parts of the skeleton were left uncleaned. This animal learned to attack the horizontal rectangle when shown. Moreover, it proved to be able to discriminate between horizontal and vertical rectangles :

This animal with one interbuccal connective intact could therefore learn, although it was unable to poison crabs.

In five other octopuses the posterior salivary nerves and interbuccal connectives were cut on both sides (Table 1). Many of the labial nerves were also cut but some near the mid-line remained intact (Pl. 1, fig. 8). All of these animals attacked crabs when shown but none was able to poison them. Sometimes the crab was killed, by pulling it apart, but the meat was never properly removed. These animals did not eat any of the fish given to them. None of these animals learned to come out to attack the rectangle.

Octopus MDO was kept alive for 15 days and tested on several occasions by repeated presentation of crabs at 5 min. intervals. At the beginning of such a session it would emerge rapidly and seize the crab, but would then drop it and return home. At later trials of the session it came out more slowly and ultimately failed to attack. Thus in one session the times taken to attack were 6, 15, 18 sec., no attack, no attack. When tested again 5 hr. later attacks were made after 4, 20, 5 sec., no attack, 20 sec. On each occasion when the octopus emerged it swam the whole length of the tank, touched the crab and returned home without seizing it. Presumably the optic lobe system was operating to produce an attack but the apparatus for completing the seizure, killing and swallowing was defective. Accordingly no signals indicating the results of the attack reached the optic lobes and the tendency to attack crabs became extinguished.

These results thus give a clear view of the functions of the various nerves involved.

Innervation of the posterior salivary glands is shown to be necessary for poisoning but not for learning (octopus MCX). Learning is possible providing that one interbuccal connective is intact (octopuses MDQ and MCX), but not otherwise, even though some labial nerves are intact (octopus MDO, etc.). Severing most of the labial nerves does not interfere with learning, but there is no animal in which they have all been cut without damage to other nerves.

This operation was performed on four animals, to test the hypothesis that the act of poisoning crabs is initiated through one or both of these pathways (Pl. 2, figs. 10, 11, octopus NAA) The characteristic syndrome after such an operation was that the octopus would readily attack and seize crabs but then neither poison nor release them. The crabs were handled clumsily and the failure to poison them may have been due to wrong positioning, but probably mainly to absence of the stimulus to eject the poison in spite of the fact that the posterior salivary nerves were intact (Pl. 2, figs. 10, 11). The animals did sometimes accept fish when given and make one or more bites in it.

After this operation the suckers are very ‘sticky’, they do not readily let go. This phenomenon was noticed by von Uexküll (1895), who called such animals ‘Greiftiere ‘.

The visual response to crabs gradually became extinguished, even though the crabs were captured. Thus 2 days after the operation octopus MEC made a series of attacks on crabs, caught and handled them moderately efficiently but never poisoned them. The crabs were given at 10 min. intervals and attacks were completed in the following times (seconds) :

When the octopus was killed it was found to be carrying a bundle of five live crabs. There was no damage to the superior buccal lobe and both posterior salivary nerves were intact throughout. The cerebro-brachial and buccal-brachial connectives were completely severed on the right side. On the left the inferior frontal-brachial and posterior buccal-brachial tracts were intact (Pl. 2, fig. 12). These are therefore not the pathways for poisoning, since this was absent in this animal. The superior buccal-brachial tracts were completely cut on the right, and only a few fibres remained on the left (Pl. 2, fig. 13).

It therefore seems that the superior buccal-brachial tracts are necessary for poisoning and for eating even fish meat.

After cutting the cerebro-brachial and buccal-brachial connective on one side only, an octopus can attack, poison and clean a crab, although sometimes leaving more meat in the exoskeleton than a normal animal would.

These operations were performed after opening the cranium and removing the jelly. Cuts were usually made with scissors, beginning dorsally and proceeding down to the level of the dorsal surface of the oesophagus.

In three octopuses (MDI, MCR and MCS) the cut passed down between the median superior and inferior frontal lobes and behind the posterior buccal lobe (Table 2; Text-fig. 4 a, b). On the day after the operation all three animals attacked crabs, though one failed to poison them. Subsequently all attacked and poisoned crabs and cleaned them well. After several days, however, the tendency to attack crabs declined and then disappeared altogether, presumably because no signals of the arrival of food were reaching the optic lobes. Through as many as eight sessions each of five trials with the horizontal rectangle none of these animals came out to attack it. This was all the more remarkable because they accepted and ate the fish that was given on each occasion after showing the rectangle. It was noticeable, however, that they would not come to take the piece of sardine on a wire as normal octopuses do, but treated it visually as an object to be avoided. Indeed they would often dash around the tank when it appeared. As soon as the fish touched the arms, however, it was accepted. Yet even after this long course of positive training the rectangle was treated as an object to be avoided, the octopus moving away when it approached. These animals therefore attacked, poisoned and ate crabs and ate fish, but did not learn to attack a strange figure.

In one animal (octopus LTF) the cut was somewhat farther forward, in front of the median inferior frontal lobe and passed through the posterior buccal lobe. The animal attacked, poisoned and ate crabs (Text-fig. 4 c). No training with a rectangle was given.

If the lesion passed through the back or middle of the superior buccal lobe the animal still attacked crabs seen at a distance but did not poison them (octopuses MAI, MAK, LZQ, MDN). This can be correlated with obvious degeneration of the posterior salivary nerve (Text-fig. 5a–c). These animals sometimes killed a crab, apparently by pulling it apart, and they cleaned it, though not always perfectly. The reflexes for eating are therefore operated through the front part of the superior buccal lobe. The eating reflexes do not necessarily involve the connective between the superior buccal and brachial lobes. In octopus MAK this was severed on both sides but both interbuccal connectives were mainly intact. The octopus did not poison crabs, but was several times found to have thoroughly cleaned crabs that it had caught and pulled apart.

These animals with lesions in the superior buccal lobe failed in up to eight sessions to learn to attack the horizontal rectangle. In the later sessions they frequently refused to accept the fish offered as food and towards the end of the experiment they refused also to attack crabs, moving away when the crab approached.

With the transection somewhat further forward the animals did not eat. This may occur when the cut passes through the centre of the superior buccal lobe (Text-fig. 5b) (octopuses, MCY, MCT, LZO), but the exact boundary has not been determined.

In the animals with damage to the superior buccal lobe various abnormalities appeared. The octopuses often failed to attack crabs from a distance. Even if they killed and ate them the empty shells were held by the arms for a long time.

If the transection was made through the front of the superior buccal lobe the animals neither poisoned crabs nor ate them (octopuses MDL and MDM). After such operations some octopuses seldom attacked crabs when seen at a distance but others continued to do so. As with all the other transections described the animals did not learn to attack a rectangle, even if food was given after every presentation for many sessions. Two of these animals had been taught before operation to attack the horizontal rectangle vigorously and also the vertical one. An attempt was then made to train the animal (Text-fig. 6) to continue to attack the horizontal (rewarded with food) but to avoid a vertical rectangle (shocked). At first both shapes were attacked by this animal but subsequently there were only occasional attacks at one or the other figure throughout seven sessions of ten trials each. The octopus learned not to attack the vertical figure, indicating that the pain pathway was intact, but was unable to continue attacking the horizontal one. After an interval of 2 days the horizontal rectangle was again shown and food given, it was sometimes attacked but these attacks gradually declined. The other animal (MDM) showed essentially similar behaviour but with fewer attacks.

These animals therefore show that if a representation has been established in the visual memory system it can continue to produce its effects even if after operation signals cannot reach the system from the mouth and gut. However its effectiveness in producing attacks is not sufficient to allow differentiation between two figures. The effect of the representation persists for some trials but ultimately undergoes extinction.

This waning of the effect of a representation established before operation is clearly shown even for attacks on crabs. The four animals shown in Text-fig. 7 attacked crabs regularly before operation. Cuts were then made between the superior and posterior buccal lobes. All the animals continued to attack on the day following operation, (but none at more than 6/8 trials). On the second day there was a further fall in attacks by all the animals, with a maximum of 4/8 trials. Tests with the horizontal rectangle and food then failed to produce any attacks in five sessions. These animals had not of course been trained before operation to attack the rectangle.

A further four octopuses were given food with the rectangle before operation (Text-fig. 8). Afterwards they attacked more often than the animals not so pretrained, but very irregularly. There was no further learning to attack the rectangle during seven sessions. At the end of the experiment the animals were tested with crabs and all attacked at the majority of trials (Text-fig. 8). The failures to attack the rectangle were therefore not due to ill-health but to the presence before operation of a ‘weak’ representation, which could not be reinforced because of severance of taste pathway.

These experiments have shown decisively that the efferent pathways for poisoning crabs and for cleaning them are distinct. Animals MCX (Table 1) and MAI (Table 2) were unable to poison, but cleaned the crabs adequately once they had killed them. Moreover, the results confirm that the nervous pathway to the posterior salivary glands begins at the back of the superior buccal lobe, a conclusion previously reached by anatomical methods (Young, 1965b). Unfortunately there has been no animal in which the interbuccal connectives (eating pathway) has been interrupted alone, leaving the poisoning pathway. This separation clearly cannot be achieved by any cut within the brain. Neither has the cerebro-subradular connective been cut independently of the other nerves, so it remains uncertain how the control of the salivary papilla, through the subradular ganglion, is effected.

The discharge of poison is presumably initiated by impulses arriving at the back of the superior buccal lobe, probably from the arms via the superior buccal-brachial connective, and the animals in which this connective was cut on both sides were indeed unable to poison or eat crabs or to eat fish. The ejection of the poison from the posterior salivary glands is presumably begun before whatever activation there may be of the salivary papilla, which is probably effected by cells lying more anteriorly in the superior buccal lobe. It is indeed possible that the operations of the papilla are largely controlled by reflexes through the subradular ganglion, under little central control.

The pathway by which impulses indicating arrival of food are able to teach the octopus to attack a given visual shape passes through the interbuccal connectives, and the superior buccal and posterior buccal lobes. Learning is not possible if there has been a bilateral interruption anywhere along this course. That this is not due to general depression as a result of operation is shown by the animals of Table 1, where the lesions to the salivary nerves or labial nerves do not prevent such learning. There is no similar control in the operations within the brain, since the pathway was interrupted in all of them (Table 2). However, the animals were all active and many of them attacked and ate crabs and fish, but yet could not learn.

The site of the receptors can be approximately identified. They do not lie in the lips, which can be removed without preventing learning. Neither are they in the oesophagus. Afferents have been shown to pass from the oesophagus in the sympathetic nerves, but animals with both the nerves cut can learn to attack a rectangle (Young, unpublished).

The afferent fibres responsible for the learning must therefore lie within the buccal mass, perhaps in the lateral buccal palps that form the floor of the mouth. Fine fibres have been seen near the surface of these palps and impulses from them would pass via the inferior buccal ganglion, and interbuccal connectives, which have been shown to be the relevant pathway (Young, 1965b).

The buccal tract is a bundle of fibres that leaves the interbuccal connective and passes to the posterior buccal and ultimately to the subvertical lobes (Pl. 1, figs. 3, 7; Text-fig. 4a). This may well be the central course of the fibres. It is not clear whether they reach the optic lobes direct (via the inferior frontal-optic tract) or through the inferior frontal-lateral superior frontal tract (perhaps by both routes).

The interruption of this pathway prevents the learning to attack some previously unfamiliar figure, but after this operation objects that the octopus had previously learned to attack will still produce a response. However, in all the animals that could not learn, the capacity to attack crabs gradually died away. This is not surprising for the animals in Table i, since these animals could not eat. However, after section behind the posterior buccal lobe we find animals that can eat but cannot learn, and in these also the tendency to attack crabs also died away, presumably by extinction of the processes that initiate it in the optic lobes if the impulses acting as signals of results do not reach the lobes.

Octopus LTF with a cut between the superior and posterior buccal lobes was able to poison and eat crabs efficiently. This gives us the important information that the posterior buccal lobe is not essential for these processes.

I am very grateful to Dr P. Dohrn, the Director, and the staff of the Zoological Station at Naples for their assistance. Also to Mrs M. Nixon for her help in preparing the MS for publication. Help with the experiments was also given by Messrs J. Bendall, M.C. Bishop, E. W. Maclarty, C. J. Mitchell and A. J. Sweatman.

The work has been sponsored in part by the Air Force Office of Scientific Research under Grant AF EOAR with the European Office.

Fig. 1. Horizontal section through top of superior buccal lobe, the labial nerves having been cut 9 days previously (octopus MCN).

Fig. 2. Sagittal section showing posterior salivary nerve damaged 9 days previously (octopus MCL).

Fig. 3. The opposite side of the animal shown in figure 2; the nerves are undamaged.

Fig. 4. Horizontal section of superior buccal lobe 10 days after the posterior salivary nerve and interbuccal connective had been cut on one side (octopus MDQ).

Fig. 5. Sagittal section showing lesion interrupting the posterior salivary nerve and interbuccal connective (13 days, octopus MCX).

Figs. 6, 7. Sections of the opposite side to that shown in fig. 5. The posterior salivary nerve has been cut, but the interbuccal connective is intact.

Fig. 8. Horizontal section showing the majority of the labial nerves cut seven days previously, but a few near the mid-line intact (n. lab.) (octopus MAL).

Fig. 9 A. Carapace and endophragmal skeleton of a crab as they appear after being cleaned and eaten by a normal octopus. B. A crab with much meat left after being attacked and taken by an octopus with the superior buccal lobe damaged.

Figs. 10, 11. Sagittal sections from the two sides of octopus NAA in which both cerebro-brachial connectives were completely cut. The animal could not poison or eat crabs in spite of the fact that both posterior salivary nerves and interbuccal connectives were intact.

Figs. 12, 13. Transverse sections of octopus MEC in which both buccal-brachial connectives were damaged but on one side (L) the inferior frontal-brachial and superior buccal-brachial tracts were intact. Notice degeneration throughout the lateral inferior frontal lobe on right but not left,.