Centres for Tactile and Visual Learning in the Brain of Octopus

Octopuses can be taught to discriminate between things that they see, and many experiments have been made to test the effect of brain lesions on performance in visual discriminations. It appears that the optic, superior frontal and vertical lobes together constitute a mechanism for learning to recognize situations by sight (references, see Young, 1961). The whole of this system can be removed without abolishing the capacity to learn to recognize objects by touch ; and still further extension of the lesion, beyond the limits of the visual learning system, has no effect upon tactile discrimination until the inferior frontal or subfrontal lobes are included, whereupon touch learning fails (Wells, 1959b). From this it would appear that visual and tactile learning are carried on in separate parts of the brain. The experiments reported so far do not, however, eliminate the possibility that with the visual system intact, an octopus might still be able to learn tactile discriminations after removal of the inferior frontal system. The present account shows that tactile learning can be abolished by lesions limited to the inferior frontal system and that such lesions do not prevent visual learning.

Octopuses of between 250 and 1000 g. from the Bay of Naples were caught and kept as described by Boycott (1954). Parts of the supra-oesophageal brain mass were removed before, or during, the course of training experiments. Most of the animals used in tactile experiments were also blinded by peripheral section of the optic nerves. Operational techniques are described in Wells & Wells (1956, 1957b).

In the text, figures and tables, individual octopuses are identified by a number, consisting of a prefix, A, B, C, …, F indicating the year in which the experiment was made (1954–59, respectively) and a yearly serial number.

Brain lesions were checked from serial sections. Lobes of the brain mentioned in the text are shown in Fig. 1, part of which is also used as the standard diagram in which lesions are plotted for Figs. 2 and 3. Some further details of the anatomy of the inferior frontal system are given in Wells (1959b).

E7 had only the buccal lobe removed.

The lesions made for the experiments to be described do not prevent octopuses from moving about in a normal manner, and it is generally only when the animals are trained that it becomes apparent that they are in any way different from controls. An exception is that animals blinded by section of the optic nerves or removal of the optic lobes tend to move about their aquaria with the arms more outspread than usual, cease to use the ‘home’ of bricks provided at one end of the tank and prefer to remain in the angle provided by the tank side and water surface when at rest. These animals are, of course, also readily distinguishable because they do not react to visual stimuli.

A variety of training procedures was used. These can be classed as setting problems of two distinct orders of complexity, as follows.

In these experiments the animals were required to stop reacting positively in a situation normally eliciting a positive response. As a visual learning test they were trained not to attack crabs, by giving them small electric shocks when they did so. As a test of capacity for touch learning, the octopuses were repeatedly presented with a small cylindrical object and trained not to take it by giving them shocks if it was passed under the web to the mouth. Unoperated controls typically learned not to attack crabs or not to take the test object after three or four shocks. In other tactile tests no shocks were given, the animals being allowed to examine and reject the test object regardless of how long they took to do so or whether the test object was passed in towards the mouth during the course of the examination; again, control animals learned after three or four trials to reject the test object without first passing it to the mouth.

Here the octopuses were required to learn to distinguish between two situations, to react positively in one and negatively in the other. As a visual test they were trained to discriminate between a rectangle shown vertically and the same rectangle shown horizontally, and as a tactile test to discriminate between two Perspex cylinders, one of which had a smooth surface and the other a surface roughened by grooves. The animals were rewarded with pieces of fish for correct responses in the ‘positive’ situation, and given electric shocks for errors in the ‘negative’. The two discriminations are about equally difficult for Octopus, and in both cases controls typically made 75 % or more correct responses after the first twenty or thirty trials.

An untrained octopus will come out of its home of bricks and attack crabs that it sees at the other end of its tank. In the experiments a crab was shown, suspended on a line by which it was kept moving about until the octopus attacked. The crab was then snatched away at the last possible moment, and the octopus given a 6–9 V. a.c. shock through electrodes attached to the end of a probe. The shock causes the animal to retreat into its shelter and the crab can be replaced, the process being repeated until the end of the trial period. In the present series, each trial lasted for 2 min., and typical animals attacked two or three times during the first trial. There were four trials per day, not less than 1 hr. apart.

Animals without brain lesions and animals with parts of the inferior frontal system removed learned quickly under these conditions, and very few of them attacked more than once at the second trial (Table 1). At the third trial only about half of the animals attacked at all, and after that attacks were infrequent. Most individuals will make occasional attacks on crabs even after long training not to do so. This implies that the negative training overlies rather than replaces the effects of a long previous experience of crabs as food, a matter that is discussed more fully elsewhere (Young, 1961).

The behaviour of octopuses with vertical lobe lesions under these conditions is in any case quite distinct; they seem to be unable to learn to leave crabs alone. The animals continue to attack despite shocks, and show very little, if any, improvement in performance as training continues (Table 1).

A summary of these results, with maps showing the range of lesions concerned, is given in Fig. 2. It should be noted that the animals with vertical lobe lesions were the same octopuses as the controls, being trained both before and after operation ; removal of the vertical lobe abolishes all trace of previous learning as well as making it difficult or impossible to retrain the animals.

Octopuses were required to learn to reject a Perspex cylinder repeatedly presented to them. Experiments of this sort with control animals and with octopuses having vertical lobe lesions have been described elsewhere (Wells, 1959a, b); with trials at 3 or 5 min. intervals both learn quickly and after taking the object, and receiving shocks for doing so three or four times, begin to reject the cylinder by thrusting it away when it is presented. Some typical performances drawn from two longer series given in Wells (1959a, b) are shown in Table 2, where they may be compared with the results of similar experiments made on octopuses after removal of parts from the inferior frontal system.

In these experiments the behaviour of animals with inferior frontal lesions is so curious that it must be considered in some detail. At the first trial an octopus such as E9 (a map of the lesion is included in Fig. 3) behaves quite normally, grasping and passing the object under the interbrachial web just like an unoperated animal. But whereas an unoperated octopus will stop taking a test object after three or four trials, an animal with an extensive inferior frontal lesion will continue to take the object even though it receives an electric shock on every occasion. On the face of it, this indicates no learning. But an octopus with the inferior frontal system removed is evidently able to recognize some elements of the test situation, even though it performs incorrectly by taking the test object, for at nearly every trial after the first three or four, it grasps the cylinder and moves away, dragging it around the tank before passing it in towards the mouth. After a number of trials, presentation of the cylinder is regularly followed by headlong flight, but the octopus does not let go and eventually (generally as soon as it stops moving) passes the object under the interbrachial web to the mouth (Table 2). The performance of animals with lesions limited to the inferior frontal system is thus quite different from that of animals with large lesions including the vertical and basal lobes as well as the inferior frontal, for while the latter also continue to take the test object and receive shocks, they do so without any apparent change in their other behaviour (Wells, 1959b). Animals with the basal lobes destroyed cannot, of course, run away, because the lesions interfere with mechanisms co-ordinating movement; but they are at least potentially able to depress the head and draw away from a stimulus (Boycott & Young, 1950), and they do not. The difference between the animals which learn to run away after making contact and those that do not change their behaviour is presumably due to the integrity of the visual learning system in the former, and it is significant that the only animal from the present series which failed to learn either to reject the object or to run away had the superior frontal lobe, a part of the visual learning system, destroyed (Ei, Table 2 and Fig. 3).

In most of the animals used for the experiments summarized in Table 2 the brachio-cerebral tracts were cut on both sides of the brain. It has generally been assumed that these predominantly sensory tracts are the main channel for tactile sensory input to the upper part of the brain. Yet shocks apparently have an effect upon the visual learning system whether the brachio-cerebral tracts are cut ot not, which implies a second sensory input channel, through the basal lobes or the brachio-optic tract. It is not known which receptors are concerned when an octopus reacts to an electric shock.

For these the same inedible cylinder was presented repeatedly, but there was no punishment for a positive response. An unblinded control animal will typically reach out to take a test object as it is dropped into the tank, and pass it at once under the interbrachial web to the mouth. After struggling to break it apart, and biting it for a minute or two (Wells & Wells, 1956), the octopus will reject the object, passing it out along the suckers and finally pushing it away with an armtip. After two or three such attempts to eat it, the cylinder is rejected when presented, without first being passed under the web to the mouth. The time taken to reject the object is somewhat variable, since it depends upon whether the octopus thrusts it away or merely removes its own arm after feeling the object over; commonly it falls from a minute or more to less than five seconds within five or six trials (Table 3).

Animals with the vertical lobe removed behave essentially as controls do in these experiments, the most notable difference being that they take approximately six times as long as controls to examine and reject the object on the occasion of its first presentation (Wells & Wells, 1957c). In general they are somewhat slower to learn than controls, taking the test cylinder slightly more often and retaining it on average for a little longer than controls do.

The behaviour of animals with lesions in the inferior frontal system is again quite distinct from that of controls, or of octopuses with lesions in the visual learning system. Out of the six animals tested, only one (B101) learned to reject the test object within the first twenty trials and four of the animals (A131, B11, B13 and B102) showed little or no improvement in performance, although two of them eventually ceased to react, apparently having learned to recognize the test object by sight. Considered as a whole the group with inferior frontal lesions not only passed the test object under the web to the mouth much more often than the others but also kept it there for longer on each occasion. Thus they took the test object at 67 % of the trials, twice as often as controls (28%) and animals with vertical lobe lesions (34%). They retained the object for 2 min. or more on 65 % of these occasions, compared with 22 % by octopuses with vertical lobe lesions–controls never kept the object for as long as this after the first trial, which is not included in calculating the figures. Indeed, the difference in performance of animals with inferior frontal lesions on the one hand, and controls or animals with vertical lobe lesions on the other, is even greater than these percentages would suggest. For while controls and animals with vertical lobe lesions never retain the object for longer than a few minutes after the first trial, octopuses with lesions in the inferior frontal system quite often keep the cylinder for half an hour or more (Table 3). During these long periods, they make movements which suggest that they are struggling to wrench the cylinder apart as they might the valves of a lamellibranch. From time to time these struggles cease and the object is rejected to the edge of the web, or even part of the way along an arm, only to be retaken and the performance repeated. Indeed it appears that animals such asB11 or B13 (Table 3) only succeed in rejecting the test object at all when by chance they fumble and lose contact with it.

The poor tactile performance of these animals does not, however, seem to interfere with their learning to recognize the test object by sight. Thus tests with B11 had to be stopped after the seventeenth trial because the animal refused to reach out an arm and grasp the cylinder when it was dropped in, while a second poor performer (B 24) would not take the object after the seventeenth trial unless this was made to roll about the floor of its tank by the experimenter; the cylinder was evidently recognized visually, but was, nevertheless, passed to the mouth when touched. In controls, and in animals lacking the vertical lobe, learning by touch regularly precedes learning by sight under these conditions.

A great many experiments have been made upon the capacity of octopuses to learn to distinguish between figures before and after vertical lobe removal. For a summary of these experiments see Young (1961). Visual discrimination and learning after inferior frontal and subfrontal lesions has, in contrast, scarcely been investigated. Boycott & Young (1955) cite two instances (without, however, giving any details, or maps of the lesions concerned) in which removal of the inferior frontal had no effect upon the performance of octopuses that had already been trained to distinguish between the situations ‘crab’ and ‘crab plus white plate’; after vertical lobe removal, octopuses are unable to learn to make this particular discrimination and continue to attack in both situations.

In the present instance animals with lesions to the inferior frontal system were trained to discriminate between two 2x10 cm. rectangles, one being shown horizontally and the other vertically. This discrimination is quickly learned by controls, which regularly make 75 % or more correct responses after the first twenty or thirty trials ; octopuses with the vertical lobes removed can also be taught to make it, but require three or four times as many trials as controls to reach the same standard of performance (Young, 1961).

For the experiments summarized in Table 4 seven unoperated octopuses were trained to distinguish between horizontal and vertical rectangles; these animals together made 69% correct responses in training periods of up to fifty-six trials. Four of them then had parts of the inferior frontal system removed; they continued to discriminate after the operation, the best animal (E32) making 70% correct post-operational responses. The other three (E7, E34 and F25) had the buccal lobe damaged or removed in the operation and were subsequently unable to eat their rewards ; these octopuses nevertheless showed that they were still able to discriminate, together attacking the positive shape sixteen times and the negative only five times before ceasing to attack altogether. Three further octopuses (F13, F15 and F28) were operated upon without pre-training. These were all able to eat their rewards post-operationally and together totalled 74% correct responses over the whole period of training; in the last forty training trials these animals made 88% correct responses. Details of all these experiments are given in Table 4.

Clearly removal of large parts from the inferior frontal system does not prevent visual learning, nor does it abolish memories of learned visual discriminations. Indeed, the performance of the four animals that were still able to eat post-operationally (E32, F13, F15 and F28 together made 74% correct responses in training −84% in the last party trials) strongly suggests that inferior frontal removal is entirely without effect upon the capacity to learn to discriminate by sight. What the operation does sometimes do is alter the chances of the animal attacking at all. Thus F13, which discriminated well, attacking the positive shape forty-one times and the negative only three times in ninety-six trials with each, nevertheless scored only 70% correct responses because of the very large number of failures to attack in the positive situation (Table 4). A possible explanation of this post-operational reluctance to attack, which was also apparent in the case of the four octopuses trained before operation, is dealt with in the discussion.

Blinded octopuses can readily be trained to discriminate between grooved and smooth Perspex cylinders, controls regularly making 75% or more correct responses the end of the first twenty or thirty trials (Wells & Wells, 1957a,b). If the vertical lobe is removed learning is slowed or prevented. The animals tend to take any object that is presented and most of them fail to learn to discriminate when training is at a rate of eight trials (4 +, 4 − ) per day. Increasing the rate of training to forty trials per day, in two sessions of twenty trials (each 10+, 10 − ), however, depresses the overriding tendency to react positively and under these conditions it becomes possible to teach octopuses lacking the vertical lobe, the animals requiring only about 50 % more trials than controls to attain standards of 75 or 85% correct responses (Wells & Wells, 19576). Once the vertical lobe has been removed, the rest of the supra-oesophageal brain can be taken out without further effect upon touch learning, provided that the inferior frontal system and the buccal lobe are left intact (Wells, 19596).

Octopuses with extensive inferior frontal lesions do not learn to discriminate by touch, whatever the rate of training. With infrequent trials (eight per day) some of the animals (B24 and B25 in Table 5) take almost every object that is presented and in this resemble animals with vertical lobe lesions ; an increase in the rate of training to forty trials per day is, however, followed by a progressive indiscriminate decline in the proportion of objects that is taken (D35, D36, D37–Table 5), so that there is no decrease in the number of errors made.

A possible explanation of the poor performance of operated animals in repeated presentation experiments is that damage to the inferior frontal system prevents octopuses from letting go. This would be compatible with von Uexküll’s (1895) observation, using Eledone, that a transverse vertical cut through the middle of the supra-oesophageal brain produces octopuses that have difficulty in releasing things that they have grasped. Such a cut would damage or disconnect the interior frontal from the rest of the supra-oesophageal brain. Von Uexküll postulated separate centres for prehension and release of the suckers, separated by the cut, but it has not subsequently been possible to trace these centres and this is probably the wrong explanation of his results (see below, and Boycott, 1961). There is certainly, however, very often a notable ‘stickiness’ about the suckers of animals with inferior frontal lesions, though the effect is curiously irregular. It is evident when the animals are touched and allowed to grasp the hand or a probe; having done so they commonly seem unable to let go, though apparently making every effort to escape by discharging jets of water and ink. In contrast there is no trace of interference with normal locomotion as the animals wander about their tanks. They do not remain stuck to surfaces unless an attempt is made to pick them up or to pull objects away from them.

These actions are explicable in terms of a dual control of arm movements, for which there is already some evidence. Thus the arms are known to be controlled partly by centres in the basal lobes, without which there is no co-ordination of their movements, no walking and no swimming (Boycott & Young, 1950). Presumably, there is a proprioceptive sensory input to these centres. Superimposed on this mechanism for the regulation of locomotion is a second system, acting almost independently on a basis of surface tactile information and normally concerned in taking or rejecting potential food objects. Isolated arms can be stimulated by touch or taste to perform all of the movements used in grasping and passing small objects to the mouth. Whether they do so or not is regulated by the inferior frontal system, for with this intact octopuses still perform satisfactorily in tactile discrimination experiments even when all of the other supra-oesophageal lobes have been removed. Without it they normally take all objects presented, despite the fact that they are potentially able to make rejection movements, which also seem to be regulated at the arms level (Wells, 19596). Even in intact octopuses the decision whether to take or reject an object touched is made entirely on a basis of surface tactile information, for octopuses cannot be taught to make tactile discriminations that would require their taking proprioceptive information into account (Wells & Wells, 1957a; Wells, 1961).

In the present series of experiments octopuses had the inferior frontal system damaged or destroyed, with the motor control regions of the basal lobes left intact. Thus in repeated presentation experiments they could run away from a contact, but had no means of inhibiting the arm reflexes that cause them to pass the test objects in towards the mouth.

A similar failure to inhibit reflex grasping by the suckers may well be responsible for the failure to attack in some visual experiments and for the ‘stickiness’ of animals with inferior frontal lesions when picked up. In these situations there must be a sudden increase in activity within the arm nerve cord as the creature prepares to rush forward or away, and in the absence of any higher control from the inferior frontal system this would perhaps incidentally enhance the activity of the reflexes controlling prehension, making it difficult or impossible for the animals to let go.

The experiments described above show that visual and tactile learning are carried out in different parts of the brain of Octopus. The optic lobes seem to be exclusively concerned with visual learning and the inferior frontal and subfrontal with learning to recognize situations by touch. The vertical lobe alone plays some part in both tactile and visual learning. If it is removed, animals that have learned to make discriminations before the operation show little trace of their training, and they can only be retaught with difficulty. The immediate effect of removal suggests that it functions at least partly as a store ; but it may also serve in establishing memory traces elsewhere in the brain for discussions of the probable functions of the vertical lobe see Young, 1961 ; Wells, 1962).

Separation of so much of the mechanism concerned in learning into two distinct regions would seem, on the face of it, to be a curious way of organizing any nervous system. But it could have arisen because the ancestors of the present-day octopods were pelagic and predominantly dependent upon vision. We know at least that cretaceous octopods had both fins and an axial skeleton, like modern decapods (Roger, 1944) and that, in modern decapods, the inferior frontal system is poorly developed (there is, for example, no subfrontal lobe (Wirz, 1959)). Presumably such animals depend on touch only in as much as it signals the success or failure of actions that have been set in motion by visual stimuli, and it may well be that as the octopods took to the bottom in cretaceous times the visual system was already too specialized for the analysis of visual stimuli to be useful as an extractor of the relevant information from an increasingly complex tactile input.

Thus the results of visual discrimination experiments (Sutherland,1) and the structure of the optic lobes (Young, 1960) both suggest that the mechanism in the brain of Octopus for analysing visual stimuli is organized to extract ratios of extents from patterns of stimulation of the retina (Young, 1961). A mechanism capable of filtering inputs in this way might be ill-adapted for dealing with tactile information where the spatial patterning of the neurones that are firing is of secondary importance to variations in their rate of firing. Octopuses distinguish objects that they touch by their texture and taste, not by their shape (Wells & Wells, 1957a). A separate mechanism for tactile learning may have developed in the inferior frontal lobe, not so much because there is any advantage in separating tactile and visual information stores, as because of an incompatability in the way that visual and tactile inputs must be coded for storing. The stimuli affecting the receptors are in both instances patterned in distribution and intensity. But whereas in the optic lobes the extraction process preserves as relevant the spatial distribution of the stimulus pattern, in the arms and inferior frontal system it is the degree of stimulation that is the more important element and its spatial distribution is discarded by the filter.

So far we know very little about the anatomy of the inferior frontal system. All that can be said at present is that there is little trace of the elaborate arrangements reported by Young (1960) for the optic lobes. The inferior frontal and subfrontal, on the other hand, do have a good deal of structural similarity to the superior frontal and vertical lobes. This, and the fact that the vertical lobe acts as a store both for tactile and for visual memories, could mean that the arrangements found in the optic lobes serve to convert patterned visual information into a form similar to that in which tactile information is stored, that is, perhaps, into trains of nerve impulses with frequencies characteristic of particular patterns of stimulation. In this situation storage might be envisaged as a property of groups of cells, so interconnected as to discharge collectively only in response to a limited range of input frequencies, rather than by ‘typing’ of cells responding to particular spatial patterns of input, as has been suggested by Young (1961).