Motor Control in Soft-Bodied Animals: The Octopus
Abstract and Keywords
Motor Control is essentially the computations required for producing coordinated sequences of commands from the controlling system (i.e., nervous system) to the actuation system (i.e., muscles) to generate efficient motion. The level of motor control complexity depends on the number of free parameters (degrees of freedom) that have to be coordinated. This number is much smaller in skeletal animals because they have a rather limited number of joints. In soft bodied animals, like the octopus, this number is virtually infinite. Here we show that the efficient motor control system of the octopus uses solutions that are very different from those of articulated animals, and it involves embodied co-evolution of the unique morphology together with the organization of the nervous and muscular systems to enable control strategies that are best suited for a highly active soft-bodied animal like the octopus.
The mollusc Octopus vulgaris displays one of the richest motor repertoire observed among all animals. We start by giving, in a nutshell, a taste of this richness to those who are not familiar with octopuses (for more reading, see Wells, 1978; Hanlon & Messenger, 1996; Huffard, 2007; Mather, 2008). Figure 21.1 shows the uncommon morphology of the Octopus vulgaris and the main external parts of its body.
Octopuses can swim by either using their arms and web to row or by using their siphon, which regularly serves for breathing, to propel the body with a jet of water. The siphon can also be used as a tool for emitting a strong jet of water (sometimes mixed with ink) at an intruder, to clean the den, or to remove sand from a shell to ease carrying it (Finn et al., 2009). Octopuses can use their arms as a very powerful tool for opening large clams, for removing heavy lids of their aquarium to escape, or for opening a jar plunger to obtain the crab inside (Fiorito et al., 1990; Anderson & Mather, 2010). Octopuses can drill through clam shells and inject paralyzing materials secreted from their salivary glands (Fiorito & Gherardi, 1999). They have several patterns of locomotion driven by the arms, such as walking, crawling, and climbing (see “Conclusion”). Some octopus species can use their arms for mimicry (Norman et al., 2001), shaping them into algae-like branches for example, while using two arms for a bipedal walking (Huffard et al., 2005). They can also use their arms for stereotypical goal-directed movements like reaching to a target (see “Peripheral Control of Arm-to-Arm Interactions”). They hunt by “netting” with the web that runs between the arms and fetch food to the mouth while standing on the rear arms. About 300 suckers are aligned along each arm and serve as very sophisticated fingers. The arms can make what seem to be autonomous movements that do not involve the body, such as probing and searching the environment aided by the millions of chemical and tactile receptors that are embedded all over the skin (with (p. 496) higher density on the sucker rims), collecting stones in a “conveyor belt”–like maneuver for building or blocking their dens, and for grooming (documentary videos are available at [http://octopus.huji.ac.il]). If all these are not enough, male octopuses use their third right arm (hectocotylized arm) for introducing sperm into the mantle cavity of the female (Wells, 1978). We will not deal here with the separate chromatophore, iridophore, and subcutaneous musculature systems that are responsible for the amazingly fast body communicational and camouflaging abilities of the modern cephalopods (Chiao et al., 2015).
Controlling such a variety of motor actions in a soft and flexible body is obviously very complex and, thus, octopuses are favorable biological inspiration for the new field of soft robotics (Kim et al., 2013). On the other hand, this flexibility and the unusual morphology have endowed the Octopus vulgaris with its amazing behavioral abilities described in this chapter, enabling it to survive and satisfy all of its needs, even though it is a solitary animal (Wells, 1978). These needs are tremendous; due to the high energy demand and their fast growth (Packard, 1972), octopuses need to efficiently gather high-energy food sources, which they do by active hunting. This way of life also demands efficient defensive mechanisms such as camouflage and intelligence supported by a very good learning and memory system (see chapters in this volume by Turchetti-Maia et al. and Hanlon; and Hanlon and Messenger, 1996). In summary, this active hunting and other behaviors of the octopus require a very efficient motor system guided by sensory input from the highly evolved visual, tactile, and chemical senses. In this chapter we attempt to answer the question of how the motor control system of the octopus is organized to enable such a repertoire of successful behaviors.
The Difficulty Involved in Controlling Behavior of a Flexible Body
Any motor control system needs to integrate external and internal (e.g., proprioceptive) sensory information together with experience (memory) to enable the choosing of which motor action to (p. 497) generate in order to serve the animal’s needs. It is believed that this complicated dynamics is achieved by interfacing the sensory and motor information through some sort of central representations of the two systems. One prominent example of such an organization is the somatotopic representation (i.e., with respect to the body parts) of sensory and motor information in the primary sensory and motor cortices of vertebrates. In humans, these representations have been first described as the sensory and motor “homunculi” (Penfield & Rasmussen, 1950; Kandel et al., 2012).
The flexibility and the unusual morphology of the octopus body pose several major difficulties for accomplishing motor control that is based on the body representation scheme discussed earlier. First, octopus arms are unsegmented, slender, and long, and can deform at any point along their length. Each arm can, at any point along its length, bend in any direction, elongate, shorten, and twist either clockwise or counterclockwise (see “The Unique Organization of the Octopus Nervous System”). Thus, the control system of the octopus must handle a virtually infinite number of degrees of freedom (DOFs). This means that its motor control system cannot rely on maps, as no biological system (to distinguish from computer control) would be able to deal with such a huge number of parameters that would be sufficient to represent its muscular system. Second, and from the same reason, this hyperredundancy of the body and arms requires also special ways to deal with the practically endless number of spatial parameters that are required to represent the sensory information gathered by the body. To summarize, this hyperredundancy of the octopus and the lack of a limited number of skeletal joints make the representation of information in body coordinates practically nonrealistic.
In this chapter we provide a concise description of how the octopus solved the difficulties involved in controlling its hyperredundant, flexible body by evolving novelties at all biological levels. All these novelties together, including also the evolution of the octopus unique morphology, have led to the emergence of an efficient embodied solution (Zullo & Hochner, 2011; Hochner, 2012, 2013). The essence of the embodied evolution of the octopus is that it overcame the problem of representation by enabling, maybe, “intelligence without representation” as suggested by Brooks (1991) in the field of Artificial Intelligence.
The Unique Neuromuscular System of the Octopus Arm
The neuromuscular system of the octopus arm differs dramatically from those of skeletal animals, the vertebrates and arthropods (Matzner et al., 2000; Rokni & Hochner, 2002; Gutfreund et al., 2006; Feinstein et al., 2011). These differences most likely reflect the adaptation of the “mechanical system” to its ecological niche to facilitate and simplify the control of the physical interactions of the flexible arms with the environment. This neuromuscular system, which includes the intrinsic musculature of the arms and the sophisticated muscular system of the ~300 suckers (per arm), lies at the bottom level of a control system hierarchy where interactions with the environment are mediated by the lowest neuronal stage; the ~3 million motor neurons that are distributed along each arm. According to embodied evolution principles, the uniqueness of the octopus arm neuromuscular system, as described in this section, is likely to reflect upward onto the organization of the higher control levels, where the adaptive neural organization evolved to simplify the control and to exploit the high redundancy of the arms.
The octopus arm, like other cephalopod tentacles, the elephant trunk, and vertebrate tongue, lacks a rigid external or internal skeleton. Instead, stiffening the muscles in these organs provides skeletal support against which muscle contractions generate the movements. Kier and Smith (1985) have termed this type of structure a “muscular hydrostat” because it is mainly composed of closely packed, incompressible muscle tissue organized in three orientations—longitudinal, transverse, and oblique (Fig. 21.2a). Due to the constant volume constraint (Fig. 21.2b), these muscle groups work against each other in the same way antagonist muscles function in skeletal animals. Due to the constant volume of the arm, any change in one dimension leads to an antagonistic change in other dimensions. Reduction of arm diameter by contraction of the transverse muscles causes the arm to elongate, and shortening of the arm by contraction of the longitudinal muscles causes the arm to thicken. The fundamental constraint of constant volume is also important for stiffness control, as coactivation of antagonistic muscle groups causes stiffening even without change in shape. More specific stiffening of the dorsal (aboral) muscle groups, for example, creates a dynamical skeletal structure on the dorsal side of the arm, serving as leverage for downward arm bending, which could be generated (p. 498) by contraction of the longitudinal muscles on the ventral (oral) side of the arm. In this case the dorsal stiffening resists arm shortening induced by the ventral groups of muscles, thus enabling forceful arm bending. Contraction of the oblique muscles causes torsion of the arm, either clockwise or anticlockwise, depending on which oblique muscle group is activated (see Kier & Stella, 2007; Yekutieli et al., 2009; Feinstein et al., 2011, for more details).
The muscle cells of the arm are small (~1,200 × 10 µm) and electrically compact (isopotential) (Matzner et al., 2000). Each muscle cell is innervated by three distinct types of motor neurons, forming a single excitatory cholinergic synaptic junction at the center of each cell. Due to the electrical compactness of the cells, these synaptic inputs control the membrane potential of the entire muscle cell and, thus, fast sodium spikes for spreading the electrical signal along the cell (like those in vertebrate muscle cells) are unnecessary. Indeed, only fast calcium action potentials that probably activate the contractile machinery were found in octopus muscle cells (Rokni & Hochner, 2002).
The neuromuscular synaptic inputs lack short-term synaptic plasticity (no short-term facilitation or depression) and also lack postsynaptic inhibition (Matzner et al., 2000). The lack of these synaptic properties, which are often found in the neuromuscular systems of other invertebrates (Bullock & Horridge, 1965; Atwood & Karunanithi, 2002), suggest that the octopus neuromuscular system uses a more linear transformation of neuronal activity into muscular actions, more similar to the transformation found in vertebrate neuromuscular junctions. Such an organization might better fit feedforward motor commands like those found in octopus reaching movement (see “Peripheral Control of Arm-to-Arm Interactions”). The muscle cells in the antagonistic longitudinal and transverse muscles all show similar morphological and physiological properties. This unification of muscle cell properties in all muscle groups possibly simplifies the neural organization of motor programs and emphasizes the importance of the highly ordered morphological organization of the arm musculature and connective tissues (Fig. 21.2a) in determining the biomechanics of the arm as a muscular hydrostat (Kier & Stella, 2007; Feinstein et al., 2011).
The intrinsic muscles of the arms, which generate both the arms’ stiffness and their movements, are innervated by an extremely large number of motor neurons, distributed along the arm nerve cord (~4 × 105 motor neurons of the intrinsic musculature per arm; Young, 1971). A rough calculation based on the muscle cell size concludes that a motor unit (all muscle cells that are innervated by a single motor neuron) comprises about 3,300 muscle cells (Feinstein et al., 2011), occupying a (p. 499) volume of only about 0.2 mm3 of the arm intrinsic musculature. This innervation pattern and density may achieve both a highly localized and continuous neural control of this nonsegmented arm musculature, and it is likely ideal for generating the waves of muscle activation typically observed in octopus arm movements (Gutfreund et al., 1998; Sumbre et al., 2006).
The Unique Organization of the Octopus Nervous System
The exceptional anatomical organization of the octopus nervous system was described by J. Z. Young, M. J. Wells, E. G. Gray, and their colleagues (Young, 1971). As shown schematically in Figure 21.3, the nervous system of the octopus (and other modern cephalopods) is divided into three main parts: a central brain surrounded by a cartilaginous capsule; two large optic lobes connected to the retinae of the highly developed, camera-like eyes; and third, the peripheral nervous system of the arms that contains about two thirds of the total ~500 million nerve cells of the octopus nervous system. These researchers deduced from the relatively few afferent and efferent fibers interconnecting the three main parts (Fig. 21.3) that much of the processing of motor and sensory information is performed in the peripheral nervous system and in the optic lobes, with the central brain fulfilling more cognitive and executive functions like motor coordination, decisionmaking, and learning and memory.
The physiological results described in this chapter emphasize the tight relationship between the special anatomy of the nervous system and the unique distribution of control labor between the central and peripheral nervous systems. This anatomical organization contributes importantly to the embodied organization of the behavior because it simplifies the control by enabling the processing of the sensory information in the periphery and storing the details for the execution of various motor program in the form of “motor primitives” (Flash & Hochner, 2005) at the peripheral neuromuscular system.
Organization of Higher Motor Control Centers in the Central Brain of the Octopus
The essential functionality of all central nervous systems is to interface the external and internal sensory information with preprogrammed (innate) motor programs and the stored information (experience) to execute appropriate motor actions. It is generally accepted that this interfacing involves some sort of representation of the sensory and motor information in dimensions that enable appropriate computation algorithms to make optimal decisions. As we discussed in the introduction, the octopus is a special case, because its unique body plan with its high redundancy poses difficulties in representing the sensory and motor information in maps based on body coordinates as is universal among skeletal vertebrates.
In all animals we can distinguish between different motor control roles for the central nervous (p. 500) system (CNS) and the peripheral nervous system (PNS). Sensory information that is perceived by the PNS is encoded and transmitted to the CNS, where this information is decoded, processed with the help of representation maps, and then encoded into motor command output that is transmitted back again to the level of the PNS, where the commands are decoded into specific muscle or muscle synergy actions (Matheson, 2002; d’Avella et al., 2003). In invertebrates, especially arthropods (but interestingly not in the neuromuscular system of the octopus arm described in “The Unique Organization of the Octopus Nervous System”), this decoding processes can take place even at the level of the neuromuscular junction, due to the polyneural innervation by both excitatory and sometimes also inhibitory motor neurons, each with different short-term plasticity properties (Bullock & Horridge, 1965; Atwood & Karunanithi, 2002). These properties allow fine tuning of the motor commands from the CNS and determine the final motor responses in the limbs.
In the classical vertebrate model, one condition for planning motor commands is based on somatotopic organization of the body in the CNS, that is, the representation of the sensory and motor systems of the body in the central nervous system, in body part coordinates. As we explained in the introduction, the octopus has an active body with eight long and highly flexible arms. Controlling these hyperredundant limbs, which have a virtually unlimited number of DOFs, would be extremely challenging for any biological system. The octopus seems to have overcome this problem through the coevolution of its unique body plan together with the special organization of its higher motor control centers—the octopus higher motor centers do not follow the classical somatotopic representation principles that vertebrates do (Zullo et al., 2009).
Extensive microstimulation studies of the octopus CNS by Zullo et al. (2009) showed that discrete electrical stimulation within the higher motor centers, the basal lobes (see illustration in the chapter on Learning and Memory), can initiate reproducible stereotypic actions belonging to the animal’s behavioral repertoire, but it was not possible to elicit some of the natural movements of the octopus. For example, microstimulation never elicited the stereotypic fetching movement (see “Peripheral Control of Arm-to-Arm Interactions”) that the animal commonly uses to bring objects to its mouth. This could be well explained by the intrinsic nature of this movement, requiring the sensory response from the grasped object along the arm to initiate the central command for coordinating the peripheral motor program that reshapes the arm into a quasi-articulated structure used to bring the object to the mouth (Sumbre et al., 2006).
Four types of main complex motor behaviors were elicited by stimulating the basal lobes—arm extension, crawling, jetting, and inking—and these appeared to be built up of sets of basic components specifically recruited for each action (Zullo et al., 2009). These patterns of motor behavior were categorized as discrete action components or complex (like the four described earlier); each complex component is composed of several of the discrete components. The response latency for initiation of each of these single discrete components, which were elicited first by low-intensity stimulation, was shorter than the latency for recruiting each of the complex ones, which appeared at higher intensity stimulation. The recruited components were active for the entire duration of stimulation and no recruitment of additional components was observed during the stimulation, if the stimulation intensity was not increased. Both the discrete and complex action components had no central somatotopic organization and, instead, were distributed over wide regions of the higher motor areas. Very significant with this respect, no stimulation site eliciting movements of only a single arm or body part was found, even though it was possible to demonstrate a certain degree of lateralization of the movements elicited. These findings fit previous morphological data suggesting a lack of somatotopic organization of afferent and efferent neurons that were traced by labeling at the level of the arm nerve cord (Budelmann & Young, 1985; Robertson et al., 1993). In these studies only a certain degree of lateralization of the fibers directed toward and from the higher motor centers was revealed.
Therefore, the stimulation experiments by Zullo et al. (2009) described earlier suggest that motor programs, rather than body parts, are represented in the higher motor centers of the octopus. These motor programs may be mechanistically represented in the form of “overlapping circuits”. Such intermingled and distributed neural networks would point to a unique organization, wherein single cells or groups of cells can be recruited into different pathways to generate a variety of behaviors according to the animal’s ethological requirements. These overlapping circuits are consistent with motor programs representation in the form of motor primitives as described by Flash and Hochner (2005) and that does not need representation of actions in (p. 501) body part coordinates. They may be compared to what Graziano (2016) described as “ethological action maps.” Although Graziano’s definition is based on the coexistence of action maps together with a standard, blurred body map along the motor cortex, Zullo et al. (2009) showed that the responses elicited from the octopus’s higher motor centers are similar in nature to responses obtained by stimulating vertebrate multisensory integrative areas (Cooke et al., 2003). These areas in vertebrates tend to be morphologically distinct from the primary motor area.
The understanding that the higher motor centers of the octopus function as higher order integrative areas is emerging also from an investigation of the central sensory representation (Zullo, Sumbre, & Hochner, unpublished data). In this study, recordings from the same areas that elicited motor responses (basal lobes) showed activity in response to sensory stimulation of various body parts. Thus, like the motor responses, the sensory inputs do not appear to be somatotopically organized. Moreover, the recordings at these sites show multimodal sensory responses (e.g., visual, tactile). These results suggest that cross-modal sensory integration is possibly achieved in the octopus higher motor centers.
How Sensory Information From the Arms Is Processed in the Brain—Behavioral Studies
Behavioral learning and memory experiments can be used to assess how octopuses perceive sensory information from their arms. Octopuses use their arms differently in various behavioral tasks, they can have preferred arms (Byrne et al., 2006), and they can use both visual and tactile information to control their arms. Touching the arms evokes robust electrical activity in the central brain (e.g., basal lobes; Zullo, Sumbre, & Hochner, unpublished data). This means that the suggestion of Wells (1978) that “octopuses are not aware of their arms” is most likely not true. However, it may be not entirely false; the recent results described later suggest that the octopus uses the information gathered by its arms in a special way that remarkably fits the lack of a central somatotopic sensory representation of arm identities or arm spatial coordinates.
Previous studies by Wells (1978) showed that octopuses can learn by touch, using their arms as chemical or tactile sensors, and that this learning is generalized to all arms. Indeed, it has not been proven to be possible to train octopuses to learn different tasks with different arms (or perhaps this is only because no one has been stubborn enough). This suggests that octopuses simplify chemo-tactile learning by generalizing the learned task and memorizing it in the CNS in the context of all arms, in the same way octopuses generalized visual learning to the two eyes (Muntz, 1961). A recent set of experiments was designed to clarify how tactile, chemical, and visual information are processed and integrated into higher cognitive abilities such as operant tasks, solving mazes, and arm skill leaning. These showed that octopuses do not have clear skill learning at the level of the arm; they cannot learn how to use an arm to solve an operant task faster (Richter et al., 2015, 2016). They learn relatively slowly to adopt the more successful strategy (or to abandon the less successful one). For example, octopuses can learn to solve a three-choice maze requiring it to use a single arm to reach into a visually cued goal compartment (Gutnick et al., 2011), but as the octopus learns the task, it takes longer for the arm to reach the goal compartment, presumably because the octopus learns to employ a slower but more successful strategy of watching its arm tip searching among the compartment entrances before deciding to simply push the bend of the arm tip into the visually cued goal compartment. Although this visually guided searching strategy seems both a simple and logical explanation, it is still possible that the octopus learns to interpret and use some proprioceptive information arriving from its arm tip to direct the arm tip to the visually cued compartment (Gutnick et al., 2011).
Control of Goal-Directed Arm Movements
In goal directed arm movements the octopus needs to precisely move an end point of its arm from point to point. The typical visually guided reaching movement (see Figure 21.4), in which the octopus extends one or several arms toward a target, exemplifies the strategies used by the octopus to simplify control in a motor system with a potentially infinite number of DOFs. The octopus simplifies the control of this goal-directed movement by propagating a stereotypical bend from the base of the arm to its tip. This strategy radically simplifies the motor control because the stereotypical propagation of the bend is controlled by a local motor program that is embedded intrinsically in the neuromuscular system of the arm and requires minimal computational intervention of the central control, because the bend always propagates in the same way in all reaching movements except for one single controllable parameter (a single DOF), which is the velocity of the propagation. This control strategy thus collapses the practically infinite (p. 502) number of DOFs that need to be controlled during the reaching movement to just three DOFs: two for the direction of the base of the arm in space, and the third for scaling the velocity of the bend propagation along the arm (Gutfreund et al., 1996). Recent results reveal that the reaching movement involves in addition to the bend propagation toward the arm tip also different levels of elongation of the segment between the base of the arm and the bend (Hanassy et al., 2015), and it is still unknown if the bend propagation and arm elongation are controlled by the same motor program or by a separate one. For example, it is suggested that the level of elongation is correlated with the distance of the eyes of the octopus from the target, thus suggesting that arm elongation might be controlled independently from the bend propagation (Hanassy et al., 2015).
Electromyography (EMG) recordings during reaching in freely behaving animals were used to test the relationship between muscle activity and kinematic parameters. The recordings suggested that reaching is controlled by a feedforward (ballistic) motor program as the level of muscle activity (detected by the EMG) predicted global parameters, like peak velocity, even though the level of activity was measured at the initial stages of the extension before reaching the peak velocity (Gutfreund et al., 1998).
Movements with the natural kinematic characteristics could be elicited in an amputated arm by tactile stimulation of the skin or by electrical stimulation of the base of the arm nerve cord (Sumbre et al., 2001). This shows that the circuitry for generating the movement is embedded in the neuromuscular system of the arm itself. If the detailed movement programs are embedded in the peripheral nervous system of the arm, somatotopic arm representation in the higher motor centers may no longer be necessary. Rather, the higher control centers may represent more complex and multilimb behaviors, as indeed was found by Zullo et al. (2009). Interestingly, in reaching or extension movements of multiple arms, either synchronously or consecutively, the velocity profiles of all the arms are similar. This suggests that the higher motor center generates only one motor command to all arms if they are activated in the same behavioral context. The conclusion from the kinematic analysis described earlier, found by Gutfreund et al. (1996), has been supported by a later study of Zullo et al. (2009), showing that stimulation of a specific site in the higher motor centers (the basal lobes) triggers arm extension in several arms simultaneously, with similar kinematic parameters.
Kinematic and dynamic (muscle action) analysis of the fetching behavior (Fig. 21.4) provides an even more striking demonstration of how the arm’s flexibility is exploited to simplify not only motor programs like the reaching but also more complex computational processes (Sumbre et al., 2006). In the fetching movement, the octopus grasps an object with a few arbitrary suckers after they hit the target during a reaching movement. Then, the arm is reshaped into a dynamic quasi-articulated structure (Fig. 21.4) by dividing the part of the arm between the base of the arm and the object into three segments—proximal, medial, and distal (L1, L2, L3, respectively, in Fig. 21.4). The distal segment grasps the object and serves as a “hand,” while the proximal and the medial segments, which are of similar length, reshape to resemble our upper arm and forearm. Then, like in the fetching movement of humans, the octopus brings the object precisely to the mouth mainly by rotating the medial joint (elbow) in a stereotypical movement that has only three controllable DOFs (Sumbre et al., 2005).
The octopus usually uses this stereotypical movement for fetching objects following a successful reaching to a target. But it also has the ability to fetch objects by conveying them along the suckers (like when collecting stones or simply by pulling the object by shortening and curving the arm (Richter et al., 2015). (See Video 21.2.) The last two forms of fetching are associated with finding the target during arm searching or arm-waving behaviors, especially when vision or direct path between the body and target is unavailable like when vision is obscured or the arm movement is physically constrained (Richter et al., 2015).
How does the octopus compute the shape of this stiffened structure to fit the arbitrary grasping position along the arm? By correlating muscle activity with kinematic features of the octopus fetching movement, Sumbre et al. (2006) revealed the mechanism for this calculation and reported that EMG recordings at various locations along the arm, with two electrodes, showed that grasping the object (in these experiments it was a piece of fish) elicits two waves of muscle activation that propagate one toward the other; one propagates proximally starting from the target grasping area of the arm, and the other propagates distally starting from the base of the arm. It was proposed that the medial joint is formed where the two waves meet. This simple mechanism explains how the articulated structure (p. 503) can be dynamically computed at the level of the arm for each fetching movement (Sumbre et al., 2006). This strategy very elegantly explains how reshaping the arm into an articulated structure can occur without central representation because the computation is all done in the periphery.
The fetching movement is thus a striking demonstration of biologic “morphological computation,” a notion tightly associated with embodied organization and robotics (Pfeifer et al., 2007).
Peripheral Control of Arm-to-Arm Interactions
The long and flexible arms of the octopus raise a major difficulty in the control of the unavoidable interferences between the arms and between arms and other body parts. Preventing appendages interactions in articulated animals can rely on central computations, which can be based on central representation maps. In some articulated animals the configuration of the body prevents such interferences simply because body parts physically cannot touch each other (like fish fins, for example). In other cases this is taken care of by the motor control system where central pattern generators (CPGs) that produce sequences of alternating stereotypical stepping movements of the appendages prevent them from being in the same place at the same time. The anatomical organization and flexibility of the octopus prevents (p. 504) the usage of such strategies for controlling the interactions between body parts (see introduction and “The Unique Organization of the Octopus Nervous System”). The potential problem that arises from the possible interference between octopus arms is intensified by the instinctive tendency of the suckers to attach by vacuum to anything they contact (Wells, 1962; Rowell, 1966; Kier & Smith, 1990, Grasso, 2008). The neuromuscular system of each of these sucking rings, which are an integral part of octopus object manipulation system and are densely aligned on the ventral side of the arm, can generate a huge negative pressure to firmly adhere the arm to a substrate by vacuum (Kier & Smith, 1990; Grasso, 2008). Their tendency to hold on to any substrate could pose a significant problem for interarm coordination if not appropriately regulated.
The evolved solution is based on a local mechanism that helps controlling arms and body interactions at the level of the peripheral nervous system of the arms. Nesher et al. (2014) showed that freshly amputated Octopus vulgaris arms have a strong tendency to use their suckers to grab and hold any object they touch, except other amputated arms (or other body parts of the octopus covered with skin). (See Video 21.3.) This avoidance mechanism is mediated by the skin itself as the suckers of amputated arms did grab the flesh of skinned arms and avoided grabbing skin that was stretched over plastic Petri dishes. In further experiments they found that amputated arms also avoided, almost entirely, grasping plastic Petri dishes coated with gel soaked in skin crude extracted with hexane. The measured grabbing forces toward these plates were approximately 10-fold weaker than those applied to Petri dishes coated with gel containing only hexane and 20-fold weaker than those applied to dishes coated with gel containing hexane extract of fish skin. Based on these findings, it was suggested that molecules in the skin trigger a local inhibition mechanism that prevents the suckers from activating their reflexive attachment mechanism, thereby preventing octopus arms from attaching to each other or to themselves in a reflexive manner.
It was further shown by Nesher et al. (2014) that behaving octopuses can alter and even reverse this mechanism. In contrast to the deterministic behavior of amputated arms, the behavior of live animals toward amputated arms was not unequivocal as sometimes octopuses grabbed amputated arms even by the skin and treated them as pray. In other cases, octopuses refrained from grabbing amputated arms and after initial physical contact they showed a very odd and unusual response that is entirely uncharacteristic of any feeding behavior; they would touch the amputated arm but instead of attaching their suckers to it, they would only “pet” it. When the arm of the behaving animal would incidentally touch the exposed flesh at the amputation site, it would grab the amputated arm only there and bring it to its mouth and then hold it only at the amputation site (where the flesh is exposed) using only its beak and avoid grabbing it with its arms as it does in normal cases of food treatment (See Video 21.4). This shows how peripheral control mechanisms can contribute to the behavior of the animals as a whole.
According to these results, it was concluded that the peripheral self-recognition mechanism that constrains interactions between arms at the periphery is indeed a simple solution to a potentially very complicated control problem in the hyperredundant flexible body of the octopus. These findings support the embodied evolution principle as they show that distributing part of the control labor to the periphery helps the central control system in coping with the infinite number of arms-controlled DOFs. This mechanism is unique because it contributes to the emerged behavior without any central processing.
Locomotion With Flexible Appendages
Octopuses move in a mysterious way. Being flexible, the movements that they make are often difficult to specify and correspondingly difficult to investigate. The literature does not contain a description of octopod walking comparable with descriptions of the six-legged, tripod gait of insects, or the stereotyped locomotor patterns of snails.
(Wells, 1978, p. 246)
Wells’s description emphasizes the essence of the mysterious appearance of octopus locomotion—the lack of obvious organization, which is in great contrast to the most salient characteristic in the locomotion of all other animals. Recent findings suggest that octopus locomotion, such as the crawling described next, is controlled in a probabilistic manner—a completely different strategy than the deterministic ones governed by CPGs that are found in all skeletal animals and also in many soft-bodied animals like the snails mentioned by Wells.
Another basic feature that may also contribute to the “mysterious” impression of octopus locomotion (p. 505) was recently discovered. During all locomotory maneuvers, octopuses hold their head in a horizontal orientation relative to the world (i.e., the axis between the eyes is perpendicular to gravity). This strategy most likely simplifies the control of interactions of the flexible arms with the environment as it reduces the complexity of the interactions from three dimensions to virtually only two dimensions (Levy & Hochner, 2015). On the other hand, because in the octopus the orientation of the head is determined by the arms, its constant orientation, in turn, constrains and simplifies the control of the arms that interact with the surroundings during various forms of locomotion. Likely these unique postural features, together with the morphology, have evolved to overcome, in an embodied way, the difficulty involved in the locomotion with soft appendages.
Octopus Crawling as an Example for Locomotion with Soft Appendages
In contrast to goal-directed arm movements, like reaching and fetching (see “Peripheral Control of Arm-to-Arm Interactions”), locomotion requires coordinating the arms. This must naturally be centrally controlled and, therefore, the octopus must maintain a delicate balance between the independency of the arms to enable goal-directed movements and the rich autonomous functionality of the arms, and efficient central arm coordination in locomotion. The crawling behavior of the octopus demonstrates an amazing solution for such organization.
Most animals are bilaterally symmetrical, presumably to enable sensory-guided locomotion in a specific direction relative to the body direction (Llinas, 1987; Buchsbaum et al., 2013). The octopus is bilaterally symmetric like most moving animals, but its arms, however, which are virtually identical, are arranged in a symmetrical circle around its mouth, in addition to being arranged in a bilateral symmetry of four arms on each side of the body. Its two eyes are laterally located on the sides of the head, which is connected directly to the body without a neck (constraining the facing and body orientation to be the same), and each eye captures half of the visual field around the head. This means that the octopus sees every point around it monocularly.
The arms are used in crawling to push the body by elongating and thus each arm has a predefined pushing direction, which is the opposite direction to its position around the body. Arms that work together apply equal forces, constraining the direction of locomotion to be the weightless vector addition of the pushing directions of the active arms (See Video 21.5). Since the arms are “all around the body,” virtually any desired crawling direction can be achieved simply by choosing the correct set of arms to use for pushing. Thus, rotating the body to change crawling direction is unnecessary. This strategy for controlling the crawling direction, based on one-dimension arm pushing, is essentially simpler to control relative to other alternatives, such as bending the arms, which would require more elaborated arm maneuvers and more complex computation. Shortening-elongating maneuvers require a relatively simple pattern of muscle activation (see “The Unique Organization of the Octopus Nervous System” and also Kier, 1988; Feinstein et al., 2011), while other alternatives would require the controlling of much larger number of DOFs. Coordinating the arms in the strategy used by the octopus is simple enough to be computationally practical, as it requires only few DOFs; the octopus needs only to choose which arms to recruit (maximum eight DOFs) and the momentary locomotion velocity (one DOF), which uniformly determines the velocity of all the pushing arms.
All other studied types of natural locomotion in other animals are governed by CPGs; these are neural networks that produce rhythmical coordinated outputs. Surprisingly, octopus crawling was found to lack this rather basic universal characteristic as diverse attempts to reveal organized crawling patterns using techniques like fast Fourier transform (FFT) that easily revealed repeated patterns in the locomotion of other animals, failed (Fig. 21.5). The lack of a CPG in the control of any locomotion may initially seem impossible, but after considering the unique morphology of the octopus with the very high hyperredundancy, one can propose that a control strategy based on probabilistic control (i.e., of the recruited appendages) is more efficient than a determinist, CPG-driven one.
It is important to note that this moment-to-moment activation of the arms is not associated only with octopus crawling. Such probabilistic strategies for controlling arm coordination exist in various forms of octopus locomotion (like walking and climbing; Levy & Hochner, 2015). This is important, because it can be claimed that CPGs are not required in octopus crawling because the body rests on the substrate and therefore arm coordination is only necessary for moving but not for maintaining stability (as in articulated animals where the legs (p. 506) (p. 507) must also be coordinated for keeping the body stable during locomotion above the ground). These additional findings suggest that probabilistic control strategies are most suitable for controlling highly dynamic soft embodiments.
While occasional observations suggest that octopuses crawl in all directions relative to their body (facing) direction without obvious preferences, deeper statistical analyses do show directional preferences (Fig. 21.6). Crawling to the right hemiplain is statistically symmetrical to crawling to the left hemiplain, but the crawling direction to each side is skewly distributed around a mode value of about ±45° relative to the facing direction and with a very large variance, consistent with the crawling in all directions. This finding can be considered an extension to the previously established statement that “all bilateral animals have preferred direction of locomotion” because locomotion in any animal must be distributed around some direction value with positive variance in reference to the external world; the difference is that in other animals the variance is very small, even negligible.
Why does the octopus tend to prefer crawling in 45 degrees relative to the facing direction? The simplest explanation is that with the octopus eyes placed on the sides of the head, crawling forward or backward would place a target in the probably poorly defined border area of the visual fields between the two eyes. This can be a disadvantage to an animal with monocular vision. In addition, although all the arms are morphologically the same, the front arms are less obscured by the mantle and are used more often for visually guided goal-directed movements like reaching. At the same time, the rear arms serve more for standing and anchoring the body to the ground, likely because the mantle is on the rear side of the body and standing on the rear arms is more stable as it lowers the center of gravity (as shown in Fig. 21.1). And indeed, as shown by Levy et al. (2015), the statistical distribution of crawling direction matches both the notion of the poorer visual accuracy at the border between the visual fields and the preference of using the hind arms for pushing during crawling versus the preference of using the front arms to negotiate the environment. The broad distribution around 45 degrees further supports the fact that this preferred body orientation in crawling is determined in a probabilistic fashion.
Conclusion: The Uniqueness in Octopus Motor Control
Here we presented results gathered during the search of biologically inspired ideas for better designing of soft robots and flexible manipulators (Walker et al., 2005; Calisti et al., 2011, 2012; Pfeifer et al., 2014). We believe that the collective appreciation of the results demonstrates how the octopus manages to cope with the complexity involved in representing information of hyperredundant body and appendages. The results suggest that this was achieved by evolving novelties that collectively form the octopus as a functional and morphological soft embodiment. These novelties have been revealed at all levels, from the special neuromuscular system and up to the higher motor control centers.
As a summary, we present a concise list of these novelties and show how they may help solving the complexity involved in controlling the behavior of soft-bodied animals.
1. The orientation of the head, which is determined by the arms that interact with the environment, is in fixed reference to the world. This (p. 508) simplifies the control by (a) reducing the control of the arms from three dimensions to virtually two dimensions and by (b) reducing the number of controlled DOFs of the arms during locomotion.
2. Higher motor centers in the brain are not somatotopically organized. Instead, complex motor programs (motor actions) are represented. This simplifies the control by overcoming the difficulty in representing hyperredundancy in body parts coordinates.
3. Goal-directed arm movements are stereotypical (e.g., reaching, fetching, and arm pushing in crawling) with only very few controlled variables (DOFs). This simplifies the control by liberating the central brain from most of the control burden and leaving it with only a few DOFs to control.
4. The details of motor programs (motor actions) are stored at the level of the neuromuscular system of the arm. This simplifies the control by (a) liberating the central brain from the need to control specific muscle groups, eliminating the need for central arm representation; and by (b), enabling some of the motor program memory trace to be stored in the physical dimension of the arm.
5. The flexibility of the arm enables computations at the level of the peripheral neuromuscular system (e.g., fetching). This simplifies the control by enabling computations without central representation.
6. A peripheral self-recognition mechanism helps controlling arm positions in space. This simplifies the control by relieving the controller from the need to have a spatial representation of the arm positions.
7. The radial symmetry of arms organization around the head and the use of a simple arm pushing strategy for thrust, for example, in crawling. This simplifies the control by (a) enabling locomotion in any direction (e.g., crawling), (b) enabling a probabilistic (stochastic) control strategy of only few DOFs, and (c) overcoming the difficulty in using CPGs for the control of morphologically unconstrained appendages.
8. Neuromuscular system that is organized in very small motor units, which are composed of small muscle cells with similar properties, both in the transverse and longitudinal muscle groups. This simplifies the control by (a) constraining part of the arm’s functionality in the morphology, (b) enabling more direct transformation of motor neuronal input into muscle actions and by that (c), facilitating employment of feedforward type of neural commands (e.g., in reaching).
In summary, the octopus is an outstanding example to show how embodied organization, a concept borrowed from robotics (Pfeifer et al., 2007), is a useful tool for understanding the coevolution of animal morphology together with the control system to enable the emergence of an adaptive dynamic interactions of the animals with their specific ecological niche.
Our research is supported by the European Commission EP7 projects OCTOPUS and STIFF-FLOP. We thank Prof. Jenny Kien for editorial assistance and suggestions.
Anderson, R. C., & Mather, J. A. (2010). It’s all in the cues: Octopuses (Enteroctopus dofleini) learn to open jars. Ferrantia, 59, 8–13.Find this resource:
Atwood, H. L., & Karunanithi, S. (2002). Diversification of synaptic strength: Presynaptic elements. Nature Reviews Neuroscience, 3, 497–516.Find this resource:
Brooks, R. A. (1991). Intelligence without representation. Artificial Intelligence, 47, 139.Find this resource:
Buchsbaum, R., Buchsbaum, M., Pearse, J., & Pearse, V. (2013). Animals without backbones: An introduction to the invertebrates. Chicago, IL: University of Chicago Press.Find this resource:
Budelmann, B. U., & Young, J. Z. (1985). Central pathways of the nerves of the arms and mantle of octopus. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences, 310, 109–122.Find this resource:
Bullock, T. H., & Horridge, G. A. (1965). Structure and function in the nervous systems of invertebrates. San Francisco, CA: W. H. Freeman.Find this resource:
Byrne, R. A., Kuba, M. J., Meisel, D. V., Griebel, U., & Mather, J. A. (2006). Does Octopus vulgaris have preferred arms? Journal of Comparative Psychology, 120, 198–204.Find this resource:
Calisti, M., Arienti, A., Renda, F., Levy, G., Hochner, B., Mazzolai, B., Dario, P., & Laschi, C. (2012). Design and development of a soft robot with crawling and grasping capabilities. In Robotics and automation (ICRA), 2012 IEEE International Conference (pp. 4950–4955). IEEE.Find this resource:
Calisti, M., Giorelli, M., Levy, G., Mazzolai, B., Hochner, B., Laschi, C., & Dario, P. (2011). An octopus-bioinspired solution to movement and manipulation for soft robots. Bioinspiration and Biomimetics, 6, 036002.Find this resource:
Chiao, C-C., Chubb, C., & Hanlon, R. T. (2015). A review of visual perception mechanisms that regulate rapid adaptive camouflage in cuttlefish. Journal of Comparative Physiology A, 201, 933–945.Find this resource:
Cooke, D. F., Taylor, C. S., Moore, T., & Graziano, M. S. (2003). Complex movements evoked by microstimulation of the ventral intraparietal area. Proceedings of the National Academy of Sciences, 100, 6163–6168.Find this resource:
d’Avella, A., Saltiel, P., & Bizzi, E. (2003). Combinations of muscle synergies in the construction of a natural motor behavior. Nature Neuroscience, 6, 300.Find this resource:
(p. 509) Feinstein, N., Nesher, N., & Hochner, B. (2011). Functional morphology of the neuromuscular system of the Octopus vulgaris arm. Vie Et Milieu-Life and Environment, 61(4), 219–229.Find this resource:
Finn, J. K., Tregenza, T., & Norman, M. D. (2009). Defensive tool use in a coconut-carrying octopus. Current Biology, 19(23), R1069–R1070.Find this resource:
Fiorito, G., & Gherardi, F. (1999). Prey-handling behaviour of Octopus vulgaris (Mollusca, Cephalopoda) on Bivalve preys. Behavioural Processes, 46, 75–88.Find this resource:
Fiorito, G., von Planta, C., & Scotto, P. (1990). Problem solving ability of Octopus vulgaris Lamarck (Mollusca, Cephalopoda). Behavioral and Neural Biology, 53, 217–230.Find this resource:
Flash, T., & Hochner, B. (2005). Motor primitives in vertebrates and invertebrates. Current Opinion in Neurobiology, 15, 660–666.Find this resource:
Graham, D. (1972). Behavioral analysis of temporal organization of walking movements in 1st instar and adult stick insect (Carausius morosus). Journal of Comparative Physiology, 81, 23–52.Find this resource:
Grasso, F. W. (2008). Octopus sucker-arm coordination in grasping and manipulation. American Malacological Bulletin, 24, 13–23.Find this resource:
Graziano, M. S. A. (2016). Ethological action maps: A paradigm shift for the motor cortex. Trends in Cognitive Sciences, 20, 121–132.Find this resource:
Gutfreund, Y,. Flash, T., Fiorito, G., & Hochner, B. (1998). Patterns of arm muscle activation involved in octopus reaching movements. Journal of Neuroscience, 18, 5976–5987.Find this resource:
Gutfreund, Y., Flash, T., Yarom, Y., Fiorito, G., Segev, I., & Hochner, B. (1996). Organization of octopus arm movements: A model system for studying the control of flexible arms. Journal of Neuroscience, 16, 7297–7307.Find this resource:
Gutfreund, Y., Matzner, H., Flash, T., & Hochner, B. (2006). Patterns of motor activity in the isolated nerve cord of the octopus arm. Biology Bulletin, 211, 212–222.Find this resource:
Gutnick, T., Byrne, R. A., Hochner, B., & Kuba, M. (2011). Octopus vulgaris uses visual information to determine the location of its arm. Current Biology, 21, 460–462.Find this resource:
Hanassy, S., Botvinnik, A., Flash, T., & Hochner, B. (2015). Stereotypical reaching movements of the octopus involve both bend propagation and arm elongation. Bioinspiration & Biomimetics, 10, 035001.Find this resource:
Hanlon, R. T., & Messenger, J. B. (1996). Cephalopod behavior. Cambridge, UK: Cambridge University Press.Find this resource:
Hochner, B. (2012). An embodied view of Octopus neurobiology. Current Biology, 22, R887–R892.Find this resource:
Hochner, B. (2013). How nervous systems evolve in relation to their embodiment: What we can learn from octopuses and other molluscs. Brain, Behavior and Evolution, 82, 19–30.Find this resource:
Huffard, C. L. (2007). Ethogram of Abdopus aculeatus (d’Orbigny, 1834) (Cephalopoda: Octopodidae): Can behavioural characters inform octopodid taxomony and systematics? Journal of Molluscan Studies, 73, 185–193.Find this resource:
Huffard, C. L., Boneka, F., & Full, R. J. (2005). Underwater bipedal locomotion by octopuses in disguise. Science, 307, 1927.Find this resource:
Kandel, E., Schwartz, J., Jessell, T., Siegelbaum, S., & Hudspeth, A. J. (2012). Principles of neural science (5th ed.). New York, NY: McGraw-Hill Education.Find this resource:
Kier, W. M. (1988). The arrangement and function of molluscan muscle. In The Mollusca, vol. 11 (Wilbure, K. M., ed). San Diego, CA: Academic Press.Find this resource:
Kier, W. M., & Smith, K. K. (1985). Tongues, tentacles and trunks—the biomechanics of movement in muscular-hydrostats. Zoological Journal of the Linnean Society-London, 83, 307–324.Find this resource:
Kier, W. M., & Smith, A. M. (1990). The morphology and mechanics of octopus suckers. Biological Bulletin, 178, 126–136.Find this resource:
Kier, W. M., & Stella, M. P. (2007). The arrangement and function of octopus arm musculature and connective tissue. Journal of Morphology, 268, 831–843.Find this resource:
Kim, S., Laschi, C., & Trimmer, B. (2013). Soft robotics: A bioinspired evolution in robotics. Trends in Biotechnology, 31, 287–294.Find this resource:
Levy, G., & Hochner, B. (2015). Octopuses hold their head in a fixed horizontal orientation relative to the world likely to simplify the control of locomotion with flexible arms. Chicago, IL: Society for Neuroscience.Find this resource:
Levy, G., Flash, T., & Hochner, B. (2015). Arm coordination in octopus crawling involves unique motor control strategies. Current Biology, 25, 1195–1200.Find this resource:
Llinas, R. (1987). Mindness as a functional state of the brain. Oxford, UK: Blackwell.Find this resource:
Mather, J. A. (2008). Cephalopod consciousness: Behavioural evidence. Conscious Cognition, 17, 37–48.Find this resource:
Matheson, T. (2002). Metathoracic neurons integrating intersegmental sensory information in the locust. The Journal of Comparative Neurology, 444, 95–114.Find this resource:
Matzner, H., Gutfreund, Y., & Hochner, B. (2000). Neuromuscular system of the flexible arm of the octopus: Physiological characterization. Journal of Neurophysiology, 83, 1315–1328.Find this resource:
Mendes, C. S., Bartos, I., Akay, T., Marka, S., & Mann, R. S. (2013). Quantification of gait parameters in freely walking wild type and sensory deprived Drosophila melanogaster. eLife, 2, e00231.Find this resource:
Muntz, W. (1961). The function of the vertical lobe system of octopus in interocular transfer. Journal of Comparative and Physiological Psychology, 54, 186.Find this resource:
Nesher, N., Levy, G., Grasso, F. W., & Hochner, B. (2014). Self-recognition mechanism between skin and suckers prevents octopus arms from interfering with each other. Current Biology, 24, 1271–1275.Find this resource:
Norman, M. D., Finn, J., & Tregenza, T. (2001). Dynamic mimicry in an Indo-Malayan octopus. Proceedings of The Royal Society of London Series B-Biological Sciences, 268, 1755–1758.Find this resource:
Packard, A. (1972). Cephalopods and fish: The limits of convergence. Biology Review, 47, 241–307.Find this resource:
Penfield, W., & Rasmussen, T. (1950). The cerebral cortex of man: A clinical study of localization of function. Macmillan.Find this resource:
Pfeifer, R., Iida, F., & Lungarella, M. (2014). Cognition from the bottom up: on biological inspiration, body morphology, and soft materials. Trends in Cognitive Sciences, 18, 404–413.Find this resource:
Pfeifer, R., Lungarella, M., & Iida, F. (2007). Self-organization, embodiment, and biologically inspired robotics. Science, 318, 1088–1093.Find this resource:
Richter, J. N., Hochner, B., & Kuba, M. J. (2015). Octopus arm movements under constrained conditions: Adaptation, modification and plasticity of motor primitives. The Journal of Experimental Biology, 218, 1069–1076.Find this resource:
Richter, J. N., Hochner, B., & Kuba, M. J. (2016). Pull or push? Octopuses solve a puzzle problem. PloS one, 11 (3), e0152048.Find this resource:
(p. 510) Robertson, J. D., Schwartz, O. M., & Lee, P. (1993). Carbocyanine dye labeling reveals a new motor nucleus in octopus brain. Journal of Comparative Neurology, 328, 485–500.Find this resource:
Rokni, D., & Hochner, B. (2002). Ionic currents underlying fast action potentials in the obliquely striated muscle cells of the octopus arm. Journal of Neurophysiology, 88, 3386–3397.Find this resource:
Rowell, C. H. F. (1966). Activity of interneurones in arm of octopus in response to tactile stimulation. Journal of Experimental Biology, 44, 589–605.Find this resource:
Sumbre, G., Fiorito, G., Flash, T., & Hochner, B. (2005). Neurobiology: Motor control of flexible octopus arms. Nature, 433, 595–596.Find this resource:
Sumbre, G., Fiorito, G., Flash, T., & Hochner, B. (2006). Octopuses use a human-like strategy to control precise point-to-point arm movements. Current Biology, 16, 767–772.Find this resource:
Sumbre, G., Gutfreund, Y., Fiorito, G., Flash, T., & Hochner, B. (2001). Control of octopus arm extension by a peripheral motor program. Science, 293, 1845–1848.Find this resource:
Walker, I. D., Dawson, D. M., Flash, T., Grasso, F. W., Hanlon, R. T., Hochner, B., . . . Zhang, Q. M. (2005). Continuum robot arms inspired by cephalopods. Enhanced and Synthetic Vision, 5804, 303–314.Find this resource:
Wells, M. J. (1962). Brain and behaviour in cephalopods. Stanford, CA: Stanford University Press.Find this resource:
Wells, M. J. (1978). Octopus: Physiology and behaviour of an advanced invertebrate. London: Chapman and Hall.Find this resource:
Yekutieli, Y., Flash, T., & Hochner, B. (2009). Biomechanics: Hydroskeletal. In L. R. Squire (Ed.), Encyclopedia of neuroscience (p. 189). Oxford, UK: Academic Press.Find this resource:
Young, J. Z. (1963). The number and sizes of nerve cells in Octopus. Proceedings of the Zoological Society of London, 140, 229–254.Find this resource:
Young, J. Z. (1965). The diameters of the fibres of the peripheral nerves of OctopusProceedings of the Royal Society of London Series B, 162, 47–79.Find this resource:
Young, J. Z. (1971). The anatomy of the nervous system of Octopus vulgaris. Oxford, UK: Clarendon Press.Find this resource:
Zullo, L., & Hochner, B. (2011). A new perspective on the organization of an invertebrate brain. Communicative and Integrative Biology, 4, 26–29.Find this resource:
Zullo, L., Sumbre, G,. Agnisola, C., Flash, T., & Hochner, B. (2009). Nonsomatotopic organization of the higher motor centers in Octopus. Current Biology, 19, 1632–1636.Find this resource: