Tactile sensing in the octopus

From Scholarpedia
Frank Grasso and Martin Wells (2013), Scholarpedia, 8(6):7165. doi:10.4249/scholarpedia.7165 revision #150456 [link to/cite this article]
Jump to: navigation, search
Post-publication activity

Curator: Frank Grasso

Figure 1: A blind octopus grasps a test object [P1, see Figure 1] with its suckers prior to making a decision to take or reject it.

All animals must have a sense of touch, if only to avoid damage. The problem is to discover how much information about their environment the animals can derive from this sense. Octopus vulgaris has proved to be a useful tool for the investigation of how much a soft-bodied invertebrate animal can derive from contacts with its environment because it learns rapidly in the laboratory. The limits of its ability to discriminate can be deduced from the results of training experiments.

Observations of the animal in the laboratory and in its natural habitat show that it is sensitive to contact all over its body surface. It uses its long flexible arms to investigate objects and to grope below and around surfaces as it moves about. It gathers much of its food in this way. The hundreds of suckers in particular are used to move across and to grasp objects that it encounters. Males use the tip of the third right arm, the so-called hectocotylus, to insert packets of sperm into the oviducts of females.


Sense organs in the skin

Figure 2: a The position of sense organs and an encapsulated neurone in the rim of a sucker. Sense organs of types 1 and 3 are presumed to be mechanoreceptors while the more numerous type 2, which reach to the surface and sometimes occur in clusters seem likely to be chemoreceptors. b shows more detail of a type 2 receptor. a After Graziadei (1962). b after Graziadei (1964).

Graziadei and Gagne (1973) have reviewed light and EM studies of sensory cells in the suckers. There are enormous numbers of these, a single sucker of 3 mm diameter (from halfway along the arm of an animal of 250 g) will have several tens of thousands of receptors with single axons running inwards from the single-cell thick columnar epithelium of the rim of the sucker. Similar but more widely spaced receptors are found all over the skin surface. In addition, see below, there are receptors within the muscles. The eight arms together would carry an estimated 2.4 × 108 receptors (Graziadei 1971). Of these the overwhelming number are long narrow cells with a ciliated tip at the surface, assumed to be chemoreceptors. A second category is formed by rounded cells, near the base of the epithelium, presumed mechanoreceptors. Two further sorts of receptors are found, one is fusiform, like the presumed chemoreceptors, but terminating in a pore with stiff immobile cilia, again possibly a form of chemoreceptor. Finally there are flattened cells spread close to the base of the epithelium, which seem likely to be a further category of contact receptor (see Figure 2).

Figure 3: Presumed proprioceptors in the muscles of octopuses. a A cell in the mantle musculature of Eledone. b Fine branches from a similar cell in a small strip of muscle (outlined by dots). c shows a receptor cell in one of the intramuscular nerve cords in the arm of Octopus with dendrites extending into an oblique muscle. The monopolar elements in the cord are assumed to be motor neurones. The relationship between the intramuscular cords and other structures in the arm are shown in Figure 3. a and b after Alexandrowicz (1960), c after Graziadei (1965).
Figure 4: Nerves and muscles in the arm of an Octopus. Nerves and muscles in the skin have been omitted outside the rim and infundibulum of a sucker. After Graziadei (1965).

As well as receptors in the epithelium there are stellate cells among the muscles of the arms and suckers (see Figure 3 and Figure 4). These are presumed to be proprioceptors, signalling muscular stretch. Degeneration experiments show that at least some of the primary receptors send axons direct to the ganglia of the nerve cord running down each arm. Others run to the acetabular ganglion at the base of each sucker. The majority of the axons from the presumed chemoreceptors run to encapsulated neurones in the connective tissue between the sucker epithelium and sucker musculature (Figure 2). This is the first stage in an integrative process that terminates in a mere 30,000 neurones running up into the brain from the tens of millions of receptors found in each arm. The tiny axons from the receptors in the suckers have proved too small for electrical recordings to be made. Rowell (1966) managed to record from nerves running into the arm nerve cords. Recordings were dominated by rapidly adapting neurones signalling mechanical distortion of the suckers. Surprisingly, no response to chemical stimulation (acetic acid, to which the animals were known to react behaviourally) was found. Recordings from the arm nerve cords showed rapid habituation to repeated peripheral stimulation with apparent novelty units responding when stimulation was changed. It appears that some learning occurs even at this level in the octopus nervous system.

Training experiments

Figure 5: a Perspex cylinders used in tactile training experiments with Octopus. In each case the percentage of the surface cut away in grooves is shown. The screw on top of each was for the attachment of a monofilament line, so that the trainer could recover the object after acceptance or rejection. b Each point shows the probability of error by one animal in 40 trials on days 8-12 of training at 8 trials (4+,4-) per day. From Wells (1978).
Figure 6: The progress of some typical training experiments using the objects shown in Figure 5a. From Wells and Wells (1957a).
Figure 7: Further Perspex objects used in tactile training experiments with Octopus. All drawn to the same scale. Simple and compound cylinders are shown in cross-section only, all were 50 mm long. Octopuses in the size range used in the experiments have suckers from 10 mm downwards in diameter as measured when applied to a flat glass surface. From Wells (1964).
Figure 8: The proportion of errors made in training varies with the difference in surface curvature of the objects to be distinguished. Points plotted 0 and 0 show results with simple and compound cylinders (Figure 7) respectively, the stated diameter being that of the units not the overall diameter of the compound bundle, which appears to be unrecognised by the animals. From Wells (1964a).

Octopus vulgaris adapts very quickly to life in aquaria, rapidly becoming tame enough to associate people with food. There is a large literature on the results of experiments in which the animals were shown pairs of cut-out shapes in a semi-random sequence, rewarded for attacking one with a fragment of fish and punished with a small electric shock for attacking the other. Under these conditions Octopus learns simple discriminations in ten or twenty trials. There is even some evidence that the animals can learn by observing the performance in training of other individuals (Fiorito and Scotto 1992). Octopus' visual acuity is similar to ours. They cannot see colours, but can detect the plane of polarisation of light (for reviews of visual learning see Wells 1978, Budelmann 1996, Hanlon and Messenger 1996).

Tactile discrimination can be investigated by similar means. It is convenient to use animals with the optic nerves cut so that they cannot see the trainer or the objects to be distinguished. Such animals tend to sit on the sides of their aquaria with the arms extended along the walls around them. Objects can be presented by touching them gently against the side of an arm. The almost invariable response is for the arm to twist and grasp the object with three or four of the nearest suckers. If the octopus considers that the object is or might be edible the arm bends and passes it under the interbrachial web towards the mouth. If the object is of no interest or distasteful it is dropped or thrust away. These behaviours can be accentuated by giving a small piece of fish impaled on the end of a fine wire to the suckers if the 'positive' object is taken, and a small [12v.dc] electric shock from a probe if the 'negative' is passed below the web. Because the trainer usually wants to present a succession of objects it is convenient to suspend each on a monofilament line so that it can be recovered. This is easy enough if the object has been rejected, more difficult if the object has been passed below the web. In the latter case it can generally be recovered by a quick jerk on the line; if the pull is developed slowly the animal will tighten its grip and refuse to let go. See Figure 5 and Figure 6 show the results of a typical series of training experiments with Perspex cylinders differing in the number and orientation of right-angled grooves cut into their sides. Individual performances vary, but it is immediately obvious that the orientation of the grooves in relation to the shape of the cylinders is irrelevant; what matters is the proportion of each surface cut away. Grooves in the surface will distort the rims of the suckers in contact. One would expect maximum stimulation of mechanoreceptors buried deeply in the rims to occur when the surface of the object grasped is 50% grooved. In contrast a smooth cylinder will only slightly distort the suckers in contact, the degree of distortion depending on the diameter of the cylinder. A very narrow cylinder should develop signals approaching those caused by an abrupt corner.

To see if these predictions were correct, a series of objects was developed as shown in Figure 7. If the distortion of mechanoreceptors in the rims of the suckers is all that the octopus recognises, a smooth sphere should yield the same signals as a flat surface. A cube is mostly flat surface but will differ from a sphere because any sucker that happens to bend around a corner will produce a distortion signal. Objects with the same ratio of flat surface to corner [such as objects PC and PSL in fig.Figure 7] should appear identical to the octopus.

The hypothesis that Octopus does not take into account the relative position of suckers in contact appears to be upheld by the experimental results (Figure 8, Figure 9 and Figure 10). This means that the animal cannot be taking into account any input from proprioceptors in the muscles. This observation is confirmed by experiments in which attempts were made to train animals to distinguish between objects with the same surface texture, but different weight. The arms quite obviously react when a heavy object is grasped and the trainer relaxes the line on which the object is suspended; The muscles at once tighten up to take the weight (Figure 11) but the animal never learns that such an object is in any way different from its light alternative. The overall conclusion from all of the experiments outlined above must be that octopuses cannot tell shapes by touch.

Figure 9: Plots show the course of four series of experiments using the Objects shown in Figure 7. P1 vs P4 is a rough-smooth discrimination, PC a cube and PS a sphere. PC and PSL are a cube and a slab with the same ratio of flat surface to length of right-angled corners. 19 vs 19 is an impossible discrimination between cylinders of the same diameter. From Wells (1964a).
Figure 10: Summarises the results of the last 80 trials of training to distinguish between a Cube and a Sphere. Training was then continued with the cube replaced by a similar cube with rounded corners, evidently more difficult to distinguish from the sphere. Substitution of the narrow rod PR for PC2 restored performance while replacing the sphere with PR revealed that PR was in fact more cube-like than a cube. Training was continued with PS and PC. Finally the animals were operated upon, lesions being made in the Inferior Frontal system of the supraesophageal brain. After this the ability to discriminate broke down completely. From Wells (1964b).
Figure 11: Attempts to teach octopuses to distinguish between objects of different weight failed unless there was also a difference in surface texture. Light objects weighed 5 g. Heavy objects drilled out and filled with lead weighted 25 (P8H) or 45 g (P4H1). From Wells (1961).

Taste by touch

The presence of so many chemoreceptors in the rims of the suckers implies that it should be possible to train octopuses to make chemotactile discriminations. This is easily shown to be so. Objects with spongy surfaces soaked in solutions of presumed different taste can be used in training just as readily as tasteless objects. Using the crude human classification of tastes as Sweet (sucrose), Sour (hydrochloric acid) and Bitter (quinine sulphate) it emerges that the octopus chemotactile sense, for these substances in seawater is at least 100 times as sensitive as our own tongue for tastes of the same substances in distilled water (Wells 1963). The animal can detect small differences in concentration and differing dilutions of seawater. Acids appear to be distinguished on a basis of pH, with acetic acid apparently tasting more acid than hydrochloric or sulphuric acids at the same pH. (Wells, Freeman and Ashburner 1965).

Further investigation of the chemotactile sense is hampered by our own inability to imagine suitable experiments. The few results available, however, make it quite clear that one must be very wary of interpreting responses to naturally occurring objects such as shellfish - an apparent shape or surface textural discrimination is just as likely to be being made on a basis of taste. In this respect it is encouraging to find that some of the Perspex objects used in the experiments outlined above were in fact indistinguishable to the octopus; there was always the possibility that the positive object would become contaminated and identified on a basis of the taste of the fish given as rewards.

Learning to make skilled movements

Octopus arms are flexible and very strong. Each can extend and contract, twist and bend, independently of the others. Each sucker can move independently, grasp, extend and contract, and exert suction. In total a very remarkable apparatus potentially capable of performing very intricate tasks. But the animal never builds a shelter more elaborate than walls of debris pulled towards it to block the entrance to the hole in which it is living. It never assembles traps for its prey, makes or dismantles anything. Even Aristotle's contention that Octopus would carefully place a stone between the valves of shellfish to prevent their closing seems never to happen in the laboratory. It has sometimes been said that octopuses will learn to open screw-top jars to get at crabs seen inside, which would suggest learning to repeat a relatively complex series of movements, but again there is little more than anecdotal evidence of results that could be achieved by random pulling.

The problem is that muscular stretch receptors cannot signal position achieved independently of contractive force exerted. To do that an animal needs joints and further receptors that can signal the angles of joints. Vertebrates and Arthropods have joints and can repeat movements, hence honeycombs and spiders webs, bird's nests and bicycles. The natural world is sharply divided into two categories, with the Octopus clearly demonstrating that whatever a soft-bodied animal's intelligence and muscular equipment, there are things that it will never be able to do.


  • Alexandrowicz, J S (1960). A muscle receptor organ in Eledone cirrhosa. Journal of the Marine Biological Association of the United Kingdom 39: 419-431.
  • Budelmann, B U (1996). Active marine predators: The sensory world of cephalopods. Marine and Freshwater Behaviour and Physiology 27: 59.
  • Fiorito, G and Scotto, P (1992). Observational learning in Octopus vulgaris. Science 256: 545-547.
  • Graziadei, P (1964). Electron microscopy of some primary receptors in the sucker of Octopus vulgaris. Zeitschrift fur Zellforschung und mikroskopische Anatomie 64: 510-522.
  • Graziadei, P (1965a). Electron microscope observations of some peripheral synapses in the sensory pathway of the sucker of Octopus vulgaris. Zeitschrift fur Zellforschung und mikroskopische Anatomie 65: 363-379.
  • Graziadei, P (1965b). Muscle receptors in cephalopods. Proceedings of the Royal Society of London B 161: 392-402.
  • Graziadei, P (1971). The nervous system of the arms. In: J Z Young (Ed.), The Anatomy of the Nervous System of Octopus vulgaris (pp. 44-61). Oxford: Clarendon Press.
  • Hanlon, R T and Messenger, J B (1996). Cephalopod Behaviour. Cambridge: Cambridge University Press.
  • Rowell, C H F (1966). Activity of interneurones in the arm of Octopus in response to tactile stimulation. Journal of Experimental Biology 44: 589-605.
  • Wells, M J (1961). Weight discrimination by Octopus. Journal of Experimental Biology 38: 127-133.
  • Wells, M J (1963). Taste by touch: Some experiments with Octopus. Journal of Experimental Biology 40: 187-193.
  • Wells, M J (1964a). Tactile discrimination of surface curvature and shape by the octopus. Journal of Experimental Biology 41: 433-445.
  • Wells, M J (1964b). Tactile discrimination of shape by Octopus. Quarterly Journal of Experimental Psychology 15(2): 156-162.
  • Wells, M J (1978). In: Octopus: Physiology and Behaviour of an advanced invertebrate (pp. 417). Chapman and Hall.
  • Wells, M J; Freeman, N H and Ashburner M (1965). Some experiments of the chemotactile sense of octopuses. Journal of Experimental Biology 43: 553-563.
  • Wells, M J and Wells, J (1957). The function of the brain of Octopus in tactile discrimination. Journal of Experimental Biology 34: 131-142.

Recommended reading

  • Hanlon, R T and Messenger, J B (1996). Cephalopod Behaviour. Cambridge: Cambridge University Press.
  • Wells, M J (1978). Octopus: Physiology and Behaviour of an advanced invertebrate. Chapman and Hall.
  • Young, J Z (1971). The Anatomy of the Nervous System of Octopus vulgaris. Oxford: Clarendon Press.

See also

Brain, Neuron

Personal tools

Focal areas