Split-Brain Preparations and Touch Learning in the Octopus

In the study of learning, interpretations of experiments involving brain lesions are often suspect because of the complexity of the changes produced by the operations. Sensory inputs may be interrupted or the capacity to carry out motor acts impaired, causing changes in the performance of the animals that are difficult to separate from changes arising from intereference with the storage systems themselves.

In the tactile learning system of the octopus the situation seems to be simpler than usual. The sensory input to the parts concerned comes from below, where damage as a result of surgery in the supraoesophageal regions is unlikely (Young, 1965). The movements made in handling objects are known to be set up mainly within the axial nerve cords of the arms ; even preparations from which the brain has been removed can carry out most of the relevant movements. Failures to discriminate cannot therefore be attributed to failures in the motor response system (Wells, 1959a, 1963). There remains, however, the problem of the level of responsiveness. The proportion of objects taken not infrequently alters as a result of the operations, and it is difficult to assess this. Is the change directly due to interference with afferent or storage elements within the tactile learning system, or is it the result of less direct physiological effects? Is the animal, for example, less in need of food because it is less active? Changes in responsiveness can clearly be due to so many different causes that one can never be certain that any two animals are responding at the same level for the same reasons.

This paper reports experiments made in the course of evolving a technique for overcoming this difficulty. Octopuses were used following operations in which parts or the whole of the supraoesophageal brain were split by a vertical longitudinal cut. The animals can still be trained to make tactile discriminations after the operation, and because both sides can now be trained independently the preparation opens up wide possibilities for the investigation of brain lesions. One side of the animal can be used as a control for the other.

Octopus vulgaris Lamarck was used throughout. The animals came from the Bay of Naples and weighed from 300 to 700 gr. They were kept in individual tanks and operated upon as soon as they were seen to be feeding and apparently healthy.

Brain lesions were made following anaesthesia with 3 % urethane, parts being cut away from the supraoesophageal brain mass and splits made with scalpel or scissors.

Octopuses in which the operation did not involve disconnexion of the optic lobes were blinded by section of the optic nerves. The animals were allowed to recover from the operations for a few days during which they were fed upon crabs and/or pieces of fish. Any that did not feed regularly were discarded. The rest were trained to discriminate between a pair of Perspex cylinders, one smooth, the other roughened by vertical grooves cut into the surface. This is an easy discrimination for normal octopuses, and has formed the basis of a number of previous experiments (see Wells, 1965). In training, the animals were rewarded with a piece of fish for passing one of the two objects under the interbrachial web to the mouth, and punished by means of a 6–9 V. a.c. shock (given by touching the animal with a pair of electrodes on a probe) for taking the other. So far as possible the cylinders were always presented to the same arm (or arms, if both sides were being trained) of the octopus, and were pulled away at the moment of transfer beneath the interbrachial web. Rewards or shocks were given immediately afterwards. There were normally twenty trials per day, in two groups of ten. Positive and negative trials alternated. There is no evidence that octopuses can learn to discriminate from the alternation sequence alone. The first trial was positive at the first group of ten trials, negative in the next, and so on. A few animals, specified in the text or figures, had 12–18 trials in a group. Trials within a group were at 5-min. intervals, with a period of at least 6 hr. between groups.

Tests on the untrained side of the animal were carried out in some experiments, exactly as in training, but without rewards or punishments.

In many of the animals all the supraoesophageal centres were removed, leaving only the inferior frontal system and the buccal lobe. Such preparations may lie motionless on the bottom of the tank, with the arms irregularly arranged and without posture. They are then easy to test by lowering objects to touch a selected arm. Some of the preparations show a different and less favourable behaviour. On the slightest stimulus they begin to spin about their axis and may continue to do so for several minutes. Other preparations assume the ‘defence posture’, with the tips of the arms turned back and the suckers outwards, whenever they are touched. This makes them very difficult to test.

The differences in behaviour are connected with the extent to which connexions between the optic and magnocellular lobes are left intact. In the best preparations they were completely cut.

Brain lesions were checked from serial sections, prepared by a modification of Cajal’s silver technique given in Sereni & Young (1932).

Where cell counts are given the numbers of cells remaining were computed by counting the cells in a sample area and multiplying by the total area found in the serial sections. Where numbers were very small one of us (J. Z. Y.) made direct counts of all the cells visible. The figures given here are means from our two independent estimates.

Two sorts of experiment were carried out. In the first the animals were trained to discriminate, always using the arms on one side of the body, and then the performance on the other side was tested by trials without rewards. These tests were made during sessions in which training was continued as before. In the second type of experiment each animal was trained to discriminate in the opposite sense on the two sides of the body.

In both series split-brain and control animals were used. In nearly all of them parts of the supraoesophageal brain were removed. The lesions were commonly larger on one side than the other, either by design or because the longitudinal split was not median. It is thus possible to compare the effect of unequal lesions on the performances of the two sides of the same animal.

Octopuses trained by means of trials limited to the arms on one side of the body were known to perform correctly when tested on the other (Wells, 1959b), and only two such experiments were made in the present series. From one of the animals, NDM, the upper part of the supraoesophageal brain was removed, leaving the basal lobes, the buccal lobe and the inferior frontal system intact (see Text-fig. 1). Such animals learn simple tactile discriminations without difficulty (Wells, 1959a) and NDM made 89 % correct responses in the first 240 trials of training; 9 out of the 20 takes of the negative object were in the first 20 trials (Text-fig. 2).

In the course of the next 80 presentations there were 40 tests in which the cylinders were presented to the untrained side of the animal. The positive cylinder was taken 20 times and the negative 8 times in 20 trials with each (80 % correct responses).

A second animal, NEG, falls into the ‘unsplit’ category, although it was intended to be a split-brain preparation. In this octopus the vertical division of the brain was incomplete in that it did not extend right through the inferior frontal system to the level of the gut. There were broad areas of contact between the two sides of the subfrontal and in the posterior buccal. The upper parts of these lobes and the inferior frontal lobe were divided, with some damage to the right-hand side of the upper part of the subfrontal. The animal was trained on the left side and made 71 % correct responses in 180 trials. In 60 tests on the right side it scored 82% correct (Fig. 2).

In octopuses NEN and NEX the lesions were comparable with the unsplit NDM, with the addition of a longitudinal split and some slight damage to the median inferior frontal and subfrontal lobes. They discriminated well on the trained side, making 82 % (NEN) and 78 % (NEX) correct responses respectively in 300 and 200 training trials. Their trained performance was thus comparable with that of the unsplit NDM.

In tests on the untrained side, however, the performance of the split and unsplit octopuses was markedly different. NEN and NEX showed little or no transfer (Text-fig. 3); NEN scored 52 % correct and NEX 61 % correct, in 100 test presentations to the untrained side.

In the remaining animals in the series, there was injury to the inferior frontal system as well as a central split. In octopuses NDO, NDL and NEF the lesions were large, including the median inferior frontal and the greater part of the subfrontal lobes on the trained side, but in each of them a proportion of the small amacrine cells forming the wall between the subfrontal and posterior buccal lobes remained. These animals learned but showed little or no signs of correct performance by the untrained side (Text-fig. 4). In three further octopuses, NGD, NPM and NEY, the lesions were similar except that no amacrine cells remained. Two of these animals failed to learn. The third, NEY, made 70% correct responses on the trained side (Text-fig. 5). In this animal, however, the difference in response actually decreased during ‘training’ and it may have been simply a manifestation of the preference for the rough object (see section 2b, below). With this possible exception the quality of performance appears to be related to the amount of amacrine tissue remaining (Table 1). Thus NDO, with an estimated 45,000 amacrines remaining on the trained side, made 80% correct responses in training. NDL, with 16,500 and NEF, with 10,600, each made 74% correct responses. NGD and NPM, lacking such cells, respectively scored only 53 % and 60 % correct.

In two octopuses the supraoesophageal lobes were removed, leaving the inferior frontal system intact and unsplit. They were then trained to take rough + /smooth − on one side, and smooth + /rough − on the other. With this procedure the responses were thoroughly confused. NLQ failed to learn altogether and NNN showed only a transient preference for the rough object (Text-fig. 6,a). In these animals the lesions were comparable with those of NDM (Text-fig. 2) and their failure can only mean that the experience of the arms on one side normally influences decisions made about the activities of the arms on the other.

The performance of NPE can also be considered here. In this octopus the attempted split was incomplete and the commissure joining the two halves of the posterior buccal lobe remained intact. The preparation learned rough + /smooth − well on one side and smooth +/rough − on the other (Fig. 6 b). This result suggests that this commissure is not responsible for the transfer of information between the two sides.

Four animals were trained in the opposite sense on the two sides following a central longitudinal split through the supraoesophageal lobes. In one of these, NQA, the supraoesophageal lobes were otherwise intact. In the others (NHK, NHL and NNA) all the lobes behind the inferior frontal system had been removed. The superior frontal/vertical lobe system, which is known to play a part in touch learning (Wells, 1959a, 1965—learning is more rapid and accurate when it is intact), was present in NQA, but not in the others.

NQA learned to take rough on the right-hand side, rejecting smooth, and to take smooth, rejecting rough, on the left-hand side, despite a period in the middle of training during which it refused both objects indiscriminately at most trials (Text-fig. 7).

The performance of the other three split-brain animals is interesting, and brings to light a further complication that must be taken into account in drawing any conclusions from these experiments. NHK, NHL and NNA all learned rough +/smooth− on the left, but failed to discriminate on the right with smooth +/rough − (Text-fig. 8).

The experiments show, therefore, that it is possible to establish different responses on the two sides of the same animal after splitting the brain, but give no evidence that the two sides can be trained to discriminate rough/smooth in the opposite sense when only the inferior frontal system remains. They show that the rough object is in some way more attractive than the smooth; even at the start of these experiments it was taken more often than the smooth cylinder. Smooth +/rough − is apparently the more difficult of the two possible discriminations.

Five octopuses were trained to discriminate rough/smooth in the opposite sense on the two sides after operations dividing the brain into two with unequal damage on the two sides. The performance of these animals is summarized in Text-fig. 9. In four of the animals (NKB, NOQ, NOR and NPL) the subfrontal lobe was destroyed on one side only; the other side was complete, or very nearly so. The scores on the intact sides were: NKB, 120 trials, 68% correct; NOQ, 120 trials, 76% correct; NOR, 96 trials, 63% correct; NPL, 120 trials, 80% correct (NOQ and NPL were trained with smooth +, the more difficult discrimination, on this intact side). Corresponding scores for the side with the subfrontal removed and no amacrines remaining on the wall of the posterior buccal were 60, 60, 46 and 54% (Table 1). In the remaining animal, NLP, there were an estimated 5000 amacrine cells left on the ‘subfrontal removed’ side. This is less than half the number found in the most extensive of the ‘train and transfer ‘lesion series (see section 1 c above) and is apparently insufficient to permit learning; NLP scored 71 % correct with smooth +/rough — on the ‘intact’ side and only 50 % correct with the easier discrimination on the side with the lesion.

In these two octopuses the brains were split with lesions removing the amacrine tissue on one side. Both learned on the intact side (NGF 63 %, NNM 71 % correct) but failed to discriminate effectively on the side with the lesion (60 and 55 % correct, Text-fig. 10).

The performance of three animals with split brains and extensive lesions is summarized in Text-fig. 11. NIK, the only animal with amacrine tissue remaining (84,000 cells on left, 41,000 on right) learned to take the smooth cylinder on one side and reject it on the other. The other two, NIL and NHZ, failed to learn.

These experiments show that octopuses can be trained to make a tactile discrimination after the supraoesophageal part of the brain has been split by a longitudinal cut. Provided that the split is median, both sides can learn. When presentations were limited to one side of the animal during training, subsequent tests on the untrained side showed no discrimination; the split prevents transfer from one side of the animal to the other.

Removal of the supraoesophageal lobes lying behind the buccal/inferior frontal system does not prevent learning, but extension of the lesion into the median inferior frontal and subfrontal lobes eventually does so. Effective discrimination ceases with removal of all of the amacrine cells that lie in the wall between the subfrontal and posterior buccal lobes. Experiments on this point were carried out with eighteen octopuses. Only one (NEY) showed some doubtful signs of learning on one side of the animal in the apparent absence of amacrine cells. There seems to be a correlation between the quality of performance in discrimination experiments and the number of amacrines left in the posterior buccal region (Table 1). If this number falls below about 10,000 the preparation fails to learn to discriminate between rough and smooth Perspex cylinders. Thus NLP, with only about 5000 cells present (out of a total of 2·5 × 10⁶ on each side in the intact animal), did not learn to distinguish the rough and smooth cylinders; while NDL and NEF, with only two or three times as many amacrine cells, made 74 % correct responses. These results confirm the main finding from the experiments of Wells (1959a) on tactile discrimination following brain lesions. They show that some amacrine cells from the subfrontal/posterior buccal region must be present if learning is to occur. They extend the earlier results by demonstrating that effective discrimination can continue in octopuses with only a few thousand amacrines present. The earlier estimate gave figures of the order of a quarter of a million amacrines necessary for discrimination of rough and smooth cylinders, with a few thousand sufficient only for the simpler task of learning to reject an object repeatedly presented.

In summary, the present series of experiments has shown that the split-brain preparations survive, feed and can be taught to discriminate by touch. Their general level of response appears to be normal, with the animals taking most of the objects presented at the start of training experiments. If the split is complete down to the level of the gut the two sides of the animal can be taught independently. Discrimination learning then follows an apparently normal course on both sides. There is little or no side-to-side transfer of the capacity to respond correctly if the split is complete. The main object of the investigation was thus achieved ; split-brain preparations are a satisfactory method of studying the effect of brain lesions to the tactile learning system, with the performance of one side of the animal used as a ‘level of response ‘control for the effects of a lesion on the other.

The observation that some of the animals preferred the rough to the smooth cylinder is interesting because the preference seems to occur only in octopuses with very large brain lesions. The bias did not, for example, appear in NQA, in which the supraoesophageal lobes were split but otherwise intact, and it has not hitherto been noticed in experiments with control animals or with octopuses having only the superior frontal/vertical lobe system removed. In Text-fig. 12 the performance of 42 control animals and 35 without vertical lobes is compared in terms of the number of trials required to reach a criterion of 75 % correct responses in the rough/smooth discrimination. There is no indication that either class of animal learns rough +/smooth − more rapidly than smooth + /rough −. Yet the present results show that animals with parts removed from the inferior frontal system discriminate rough + /smooth − more readily than the reverse. The bias is not seen in preparations with the whole inferior frontal system removed, which take everything (Wells, 1959a). Moreover in these preparations neither object seemed obviously difficult for the octopus to grasp or let go. One can only suppose that the innate structure of the tactile learning system is in some way biased in favour of grasping and pulling in objects with a rough surface ; a similar and steady reading from all the suckers is relatively unattractive. In the intact animal this marginal preference rarely expresses itself because the animal learns so rapidly. In the creature with a drastically reduced touch-learning system it becomes important and must clearly be taken into account in any further studies with such preparations. Since the median split is now known to divide the brain effectively it will presumably be unnecessary to train the sides of the animals in the opposite sense in future experiments. Instead the rough/smooth discrimination situation can be used to provide two problems of different difficulty.

Our thanks are due to Dr P. Dohrn and the staff of the Zoological Station at Naplest Also to Miss P. Stephens for histological assistance. The work was supported in par. by the Air Force Office of Scientific Research under Grant AF EOAR 64-42, with the European Office of Aerospace Research, United States Air Force (J. Z.Y.), and partly by a grant from the Rockefeller Foundation (M.J.W.).

Fig. A. Transverse section of inferior frontal region. Cajal’s stain, l.i.f., Lateral inferior frontal; m.i.f., median inferior frontal ; p.buc. posterior buccal; subfr., subfrontal.

Figs. B, C. Horizontal section through the remaining parts of the supraoesophageal lobes of octopus NLP, whose performance is shown in Text-fig. 9. Note the complete bisection. Some amacrine tissue remains on the left, but not the right and there was discrimination only on the left, am., amacrine tissue; lab., labial nerves; sup.buc., superior buccal lobe.