The Effect of Splitting part of the Brain or Removal of the Median Inferior Frontal Lobe on Touch Learning in Octopus

The object of these experiments was to get further information about the parts played by the various supraoesophageal lobes in touch learning. The anatomy is described elsewhere (Young, 1964, 1969). We have compared the rate of learning a simple successive tactile discrimination task in animals with (1) no central lesion (‘normals’), (2) the whole supraoesophageal brain divided in the mid-line (‘half-brains’), (3) the whole inferior frontal system divided (‘anterior splits’), (4) the rest of the supraoesophageal lobes divided, leaving the inferior frontal system intact (‘posterior splits’), and (5) with the median inferior frontal lobe removed.

The animals were trained on one side and in most of them the opposite side was later tested for its capacity to discriminate. It has already been shown that in animals trained on one side after complete supraoesophageal section in the mid-line the untaught side does not discriminate (Wells & Young, 1966). The present experiments show that the passage for side-to-side transfer is mainly through the inferior frontal system, and particularly through the median inferior frontal lobe. The vertical lobe system possibly provides a channel for some lateral transfer. The experiments also show that after removal of the median inferior frontal lobe touch-learning is grossly impaired

The octopuses were between 250 and 450 g., obtained from the Bay of Naples and kept separately. They were tested with crabs on the days after capture and operated when they were feeding regularly. Under urethane anaesthesia all were blinded by bilateral section of the optic nerves. In those to be ‘normal’ controls the cranium was then opened and the jelly removed. In the ‘operated’ animals bisections were performed with a fine scalpel blade. Scalpel and very fine scissors were used to remove the median inferior frontal lobe.

In the days after operation each animal was tested 8 times morning and evening with a small piece of sardine, given on a wire. Training was begun only with animals that were feeding regularly. Rough and smooth plastic spheres were used as already described (Wells & Young, 1966). There were two training sessions each day. At each there were eight trials with the positive and eight with the negative sphere, given alternately with positive first. Food was given only when the positive sphere was passed towards the mouth or when it was still being held by the suckers after 20 sec. A shock (10 V. a.c.) was given when the animal similarly took the negative sphere or held it for 20 sec. Trials were at intervals of about 5 min. All training was on the right side, using the second arm whenever possible.

At the end of the experiment the brains were fixed in formalin, silver-stained in block and serially sectioned (Stephens, 1969).

The temperature of the sea water varied from 23° to 26° C.

Normal octopuses have a greater tendency to take smooth than rough objects and this difference is reversed after some lesions (Wells & Young, 1968). In the present experiment, therefore, half the animals with each operation were trained in each direction, selected so far as possible randomly out of each batch operated.

The normal animals showed a clear preference for smooth (Figs. 1, 2). The details of their performance are recorded elsewhere (Wells & Young, 1968). The split animals trained with smooth positive began at a lower level of smooth preference than normals (52 % against 64% takes of smooth in first 16 total takes) but thereafter the two groups learned at essentially the same rates. The half-brains trained rough-positive began with a somewhat greater tendency to take rough than normals and this difference remained throughout. The main difference between normal and split-brain animals is thus that the former have a smooth preference but not the latter. The rate of improvement in discrimination is similar in both. There was, however, a somewhat greater tendency to take both objects in the half-brain animals, perhaps because they lack the vertical and subfrontal lobes of one side. There is evidence that these lobes have in general an inhibitory action. In tactile experiments animals typically begin by taking both of the objects to be discriminated. Training usually results in a diminution of the proportion of objects taken, at first indiscriminate but later limited to the negative object. Removal of the vertical and subfrontal lobes tends to lead to an increased proportion of takes.

These results are in apparent disagreement with some previously reported in which half-brain animals trained with smooth positive were considered to learn more slowly than normals (Wells & Young, 1966). However, in that experiment, as in the present one, the half-brains took fewer smooths than did the normals over the early trials, the effect of a shift in untrained preference that is associated with the splitting operation and which was only detected subsequent to the 1966 experiments (Wells & Young, 1968). When the results with animals trained S+ and R+ are considered together, normal and split-brain octopuses show a similar rate of improvement in this simply discrimination.

Eleven animals were satisfactorily trained on the right side with the inferior frontal system split (Table 1). There was no serious damage to the right subfrontals and only in one (slight) to left subfrontal. The median inferior frontal and subfrontal commissures were completely cut except for some fibres of the latter in one animal.

The right sides all learned well, with correct response means of 71·5% S+ and 80·0% R+ over 160 trials. In 80 further training trials they gave 77·8% S+ and 93·2% R+ (Table 1). These animals therefore prefer rough (Figs, 1, 2). It would be interesting to know whether these half-inferior-frontals learn at a different rate from normals, but comparisons are of course made difficult by the fact that the half-brains have a preference opposite to the normals. If we take the mean correct response for rough-positive and smooth-positive together we obtain 75% over 160 trials for the 23 normal animals already reported (Wells & Young, 1968) and 76% for the present 11 half-inferior-frontals. Of course such a comparison implies that the preferences are equally strong in the opposite directions, which is unlikely to be true. Even so it is clear that no great deficit in discrimination learning is involved by halving this part of the system. This is confirmed by comparing the scores achieved in each successive series of 16 takes by the different groups of animals (Fig. 1). The differences in preference appear throughout but the rate of gain of information is in general similar in normal animals, in those with the supraoesophageal lobes wholly split and in those with the anterior part only divided. The course of learning was reasonably consistent in the different animals with anterior splits (though two out of the eleven animals showed a clear preference from the start: one for S +, one for R +), which was not modified by subsequent training.

The animals with anterior splits, when tested on the left, showed only faint signs of correct response. Ten out of the eleven gave 50% correct or more, but only three were over 60% correct. The means were 57% for R+ and 56% for S +. The small sign of capacity for correct response on the untrained side can perhaps be taken to show that some transfer takes place through the posterior part of the brain. It must be remembered, however, that the risks of contamination are great. If a sphere is carried rapidly towards the mouth it is difficult and sometime impossible to prevent it touching arms on the other side near the mouth. In a previous series of experiments using exactly the same training techniques there were, however, no signs of transfer at all after complete splitting (Wells & Young, 1966). This previous study also included five animals with the inferior frontal system alone split and they showed even more correct response on the (untrained) left side than the present series (73 %, all with smooth-positive). It seems likely, therefore, that some transfer can take place in these animals, with the inferior frontals split, presumably through the vertical lobe system.

An interesting feature of the animals with anterior splits is the very low number of takes of the ‘positive’ on the left compared with the very high number on the right. Clearly there is ‘positive’ learning by the trained side as a result of the food taken, and this has not been transferred. Every single animal shows much higher takes of the positive on the right than left (Table 1). Conversely, in all animals except two there were more takes of the negative on the left. The contrast is very striking and this is excellent evidence of learning on both directions. Very little of either can transfer through the vertical lobes or be used by the opposite side through them.

There are eight of these: six S + and two R+ (Fig. 2.). In general these learned on the right side to levels of correct response lower than the anterior splits (Figs. 1,2). The difference is due to a greater tendency to take both positive and negative in the animals with posterior splits.

The difference in attacks at the negative is significant at 5% level (Wilcoxon t = 53, n₁ = 8, n₂ = 11). This is good evidence that removing half of the vertical lobe causes increased takes especially of the negative object. Actually removal was of more than half, since many of the cuts lay to the right of the mid-line (see Table 2). The large number of takes of both spheres continued in the second period of training throughout the period when the animals were also getting tests on the untrained side.

Tests on the left side showed more correct responses on the untrained side than with the anterior splits (67-3% correct S+ and 57-0% R +). There were, however, three octopuses out of the eight posterior splits tested that made less than 60% correct response and the mean difference between anterior and posterior split-brain animals is not significant at the 0-05 % level. There are reasons for the three failures to show good response on the untrained side (Table 2). Octopus PGI learned very poorly on the right and should probably be excluded altogether for that reason. PGG was a very strong taker and correspondingly poor discriminator. PGW was such a strong taker on the left that it took nearly everything. We can therefore be confident in saying on the basis of the remaining six animals that with a split vertical-lobe system correct response can take place on an untrained side. Therefore one must suppose either that the memory record is transferred in the more anterior lobes, through one of the two inferior frontal commissures, or that the untrained side uses the record in the trained side through one or both of these (see Wells & Young, 1966).

This transfer, or capacity to read out from the trained side, affects both positive and negative learning. In the animals with posterior splits the takes of the positive were only slightly lower on the left than on the right and very much higher than in the animals with anterior splits. This difference is significant at 0·05% (if we exclude PGF, which is quite anomalous). Evidently the ‘positive’ memory record, ensuring take, can be transferred (or that on the other side used) even with a split vertical lobe system.

Finally, in five out of eight animals with posterior splits the takes of ‘negative’ were less on the left than the right, whereas in anterior splits they were nearly always more. There is no special reason why they should be other than equal in the posterior splits, but the difference from the anterior splits emphasizes that the latter had not learned not to take the negative on the left and that the left side could not use the record of the right.

After this operation six animals were trained with smooth and seven with rough positive (Table 3). The animals showed a marked preference for rough (Figs. 1, 2). Of those trained with smooth positive only two showed progressively more correct responses as training proceeded (460 = PFM and 461 = PFN, the mean for %C being 75 %); the performance of four others became more indiscriminate as training proceeded (468 = PFO, 469 = PFR, 313 = PFP and 314 = PFQ, the mean for %C being 61 %).

Of the seven trained with rough positive five showed a strong rough preference from the start. Two of these failed to improve in performance. A third animal (PEF) took nearly all of the objects presented to it and thus showed no preference and no improvement. The learning of these animals was, in short, strikingly slower than that of either normal or split-brain animals. A good control is provided by two octopuses in which the operation removed only a small amount of the median inferior frontal; these two learned well with smooth positive (Fig. 2, ‘dummy ops’).

Tests on the untrained (left) side showed some accuracy in two of the animals trained with smooth positive (Table 3), but in both of these a few intact median inferior frontal fibres were found in the mid-line. In the other four animals there was no sign of transfer and in none of them were there intact commissural fibres. The animals trained with rough positive mostly appeared to discriminate on the left, as was to be expected from the initial preference for rough.

It is clear that the median inferior frontal has a very important part in the tactile learning process. The operations had mostly been very successful in removing only this lobe, leaving the subfrontal intact (Table 3). Of course the subfrontals could not have been operating normally, however, since they were deprived of all input from above. For the present, therefore, all we can say is that interference with the inferior frontal system in this way profoundly disturbs tactile learning, reversing the preferences and delaying but not preventing the setting up of long-term memory records.

These results contain several examples of the reversal in preference that has been described elsewhere (Wells & Young, 1968). The animals with brains split, whether completely, or at the front or back, all showed a less than normal preference for smooth, sometimes producing an actual excess of takes of rough. Removal of the median inferior frontal also produced this effect. In most animals this change of preference is accompanied by an excessive tendency to take objects, including both positive and negative. The association of the two effects may of course be fortuitous, but it supports the suggestion already put forward that the rougher objects produce the greater number of signals that tend to produce a take but that this tendency is inhibited, perhaps by those parts of the system containing amacrine cells, namely the vertical and subfrontal lobes. It is presumably the reduction in the number of these cells that leads to the increased take after the operations and perhaps also to the greater tendency to take rough.

The experiments show that the commissural fibres through which tactile training on one side allows correct response on the other are mainly in the median inferior frontal lobe. No effective transfer by way of the subfrontal commissure is possible after the median lobe has been removed, and only at most a little through the vertical lobe system. Further experiments are needed to show whether the transfer through the median inferior frontal involves establishment of a representation in the memory regions of the untrained side or use (through the median inferior frontal) of that on the trained side.

When tested on the untrained side the animals with the inferior frontal system divided showed a much lower tendency to take the positive object as well as a greater tendency to take the negative than on the trained side. This is good evidence that training consists both in learning to take the one object and to reject the other. Neither form of association could form on the untrained side in the divided animals. This phenomenon did not appear after division only of the posterior (vertical lobe) region.

It is surprising that elimination of half of the brain by splitting produces so little alteration in the performance of the trained side. Perhaps this is because with the comparatively small number of animals used, small effects do not appear.

The discrimination was in any case a very simple one. As with the vertical lobe in visual learning one might expect the effect of removing a proportion of the available tissue to be most apparent in difficult discriminations. This matter has yet to be investigated with touch learning and tasks of graded difficulty.

We are grateful to the Director and Staff of the Zoological Station at Naples for the facilities that made the work possible. The tanks and water supply were provided by a grant from the Science Research Council. The work was assisted in part by the Air Force Office of Scientific Research through the European Office of Aerospace Research, OAR, United States Air Force, under grant AF EOAR 66–53, and in part by the Science Research Council of Great Britain. We are grateful to Mr G. D. Sanders and Mr D. Gaffan for discussion of the results and assistance in preparing the MS.