Journal of Neurophysiology

Dual Sensory-Motor Function for a Molluskan Statocyst Network

R. Levi, P. Varona, Y. I. Arshavsky, M. I. Rabinovich, A. I. Selverston


In mollusks, statocyst receptor cells (SRCs) interact with each other forming a neural network; their activity is determined by both the animal's orientation in the gravitational field and multimodal inputs. These two facts suggest that the function of the statocysts is not limited to sensing the animal's orientation. We studied the role of the statocysts in the organization of search motion during hunting behavior in the marine mollusk, Clione limacina. When hunting, Clione swims along a complex trajectory including numerous twists and turns confined within a definite space. Search-like behavior could be evoked pharmacologically by physostigmine; application of physostigmine to the isolated CNS produced “fictive search behavior” monitored by recordings from wing and tail nerves. Both in behavioral and in vitro experiments, we found that the statocysts are necessary for search behavior. The motor program typical of searching could not be produced after removing the statocysts. Simultaneous recordings from single SRCs and motor nerves showed that there was a correlation between the SRCs activity and search episodes. This correlation occurred even though the preparation was fixed and, therefore the sensory stimulus was constant. The excitation of individual SRCs could in some cases precede the beginning of search episodes. A biologically based model showed that, theoretically, the hunting search motor program could be generated by the statocyst receptor network due to its intrinsic dynamics. The results presented support for the idea that the statocysts are actively involved in the production of the motor program underlying search movements during hunting behavior.


Sensory organs provide the CNS with information about both the environment and activity of the motor apparatus so that the resultant behavior is adapted to changing conditions. However, since many sensory organs are complex neural networks with their own dynamics and have efferent inputs from the CNS, one can suggest that they may contribute to the production of output motor patterns in addition to their usual transduction function (Janse et al. 1988; Meyer and Bullock 1977). Here we support this suggestion by studying the neural mechanisms of motor control during hunting behavior in the marine mollusk Clione limacina (Gastropoda, Opisthobranchia, Pteropoda).

Clione (Fig. 1A) is a planktonic mollusk that swims by rhythmic movements of a pair of wings (Arshavsky et al. 1985; Satterlie et al. 1985). The main effector organ determining swimming direction is the tail that presumably acts as a rudder. During swimming in the sea, Clione is usually oriented vertically, with its head up (Arshavsky et al. 1985; Deliagina et al. 1998, 1999; Satterlie et al. 1985). The vertical orientation is actively maintained under the control of signals from the gravity sensory organs, the statocysts (Deliagina et al. 1998, 1999; Panchin et al. 1995a). A deviation from the vertical orientation evokes a compensatory gravitational reflex: Clione bends its tail up, which results in a restoration of an initial position. After bilateral extirpation of the statocysts, Clione is not able to maintain a definite body orientation (Panchin et al. 1995a).

fig. 1.

A and B: photos of Clione during routine swimming (A) and during hunting (B). W, wings; T, tail; Tntl, 3 pairs of protracted tentacles (cephaloconi). B: wings are blurred because the frequency of their oscillations during hunting is about 3 times higher than during routine swimming. C: neural circuitry involved in the control of Clione spatial orientation and search behavior during hunting. ChR, chemoreceptors; TR, tactile receptors; SRC, statocyst receptor cells; ChunIN, cerebral hunting interneurons; CPINs, cerebro-pedal interneurons affecting the central pattern generator for swimming (CPG), wing motoneurons (WMN), and tail motoneurons (TMN). Arrows between neurons show functional connections obtained in experiments with paired recordings (Panchin et al. 1995a,b).

Clione is a predator that feeds exclusively on a related pteropod mollusk, Limacina helicina (Arshavsky et al. 1989; Hermans and Satterlie 1992; Lalli 1970; Norekian and Satterlie 1993). With an undeveloped visual system, Clione uses relatively nondirectional information from its chemoreceptors to locate prey. The presence and especially a contact with Limacina appear to stimulate chemoreceptors and tactile receptors, which trigger a complex hunting behavior. This very characteristic mode of behavior was first described over 100 yr ago (Boas 1886). At its inception, three pairs of tentacles (cephaloconi) are extended to capture the prey (Fig. 1B), the frequency of wing oscillations increases by approximately three times, and the tail strongly bends. As hunting progresses, the direction of the tail bending continuously changes and Clione loops along a complex trajectory covering the surrounding space in an attempt to locate and capture the prey. This aspect of hunting behavior was designated search behavior. One of the most typical features of hunting behavior is a reversal of the reflex to head stimulation. Normally, stimulation of the head evokes a defense reaction: wing oscillations are inhibited and Clione sinks avoiding the irritation (Arshavsky et al. 1985; Satterlie et al. 1985). Instead, during hunting, the same stimulus evokes a further protraction of the tentacles and a grasping of the stimulating object (Arshavsky et al. 1993).

Hunting-like behavior of Clione can be evoked experimentally in the absence of prey by injection of the acetylcholinesterase inhibitor, physostigmine, into the hemocoel (Arshavsky et al. 1993). Similar to the hunting behavior triggered by a prey, physostigmine initiates tentacle protraction, an acceleration of wing beats, and the reversal of the reaction to head stimulation. Application of physostigmine to a preparation of the isolated CNS evokes bouts of “fictive hunting behavior,” i.e., the excitation of motor systems (cerebral protractor tentacle motoneurons, the central pattern generator for swimming, and buccal protractor neurons) involved in hunting behavior. In addition, physostigmine produces a reversal of the reaction to head nerve stimulation imitating the response to head stimulation in the intact animal. After physostigmine application, head nerve stimulation evokes strong excitation of both protractor tentacle motoneurons and the swimming central pattern generator instead of their inhibition as observed before physostigmine application. Therefore using physostigmine gave us the possibility for an electrophysiological analysis of the neural mechanisms underlying search behavior.

Neural structures involved in the control of Clione's spatial orientation and search behavior are shown in Fig. 1C. They include a pair of statocysts and several groups of neurons located within the cerebral and pedal ganglia (Deliagina et al. 1999). As in other gastropods (Alkon 1975; Detwiler and Alkon 1973; Gallin and Wiederhold 1977; Janse 1982, 1983), the statocysts in Clione are paired spherical organs, 0.15 mm in diameter, located on the dorsal surface of the pedal ganglia. Each statocyst contains a stone-like structure, the statolith that moves inside the sphere under the influence of gravity. The statocyst internal wall is lined with 9-11 statocyst receptor cells (SRCs) (Tsirulis 1974). The SRCs are flat (5 μm thick) mechanoreceptors responding to the pressure exerted by the statolith. Their axons run into the cerebral ganglia. At least some of the SRCs in Clione form nonreciprocal inhibitory connections (Panchin et al. 1995a), such as in another gastropod mollusk, Hermissenda (Detwiler and Alkon 1973). In addition, some SRCs are electrically coupled. Thus SRCs form a small neural network. The SRCs affect several groups of cerebro-pedal interneurons controlling output motor systems (the central pattern generator for swimming, wing, and tail motoneurons) that are mainly located in the pedal ganglia (Fig. 1C) (Arshavsky et al. 1985; Panchin et al. 1995a,b; Satterlie and Spencer 1985).

The activity of SRCs in mollusks is influenced not only by the animal's orientation in the gravity field but also by efferent inputs (Janse et al. 1988; Salanki and Jahan-Parwar 1985; Wolff 1975). These inputs include different modalities and thus control the activity of SRCs depending on external and internal influences. Among neurons affecting the SRCs in Clione, there is a pair of identified interneurons designated “cerebral hunting interneurons” (ChunINs) (Arshavsky et al. 1993; Panchin et al. 1995b) or “cerebral interneurons initiating prey capture” (Norekian and Satterlie 1995). The CHunINs affect also the activity of different motor systems involved in the control of hunting behavior, including the central pattern generator for swimming and tentacle motoneurons. The CHunINs are excited by physostigmine application to the isolated nervous system (Arshavsky et al. 1993) and by tactile stimulation of the lips and cephaloconi in a semi-intact preparation (Norekian and Satterlie 1995). It is reasonable to suggest that, in an intact Clione, the CHunINs are activated in the presence of Limacina by signals from chemoreceptors and tactile receptors (Fig. 1C).

During search motion, signals from SRCs are not used for stabilization of a definite orientation in the gravitational field as during regular swimming. Here we raise the question of whether or not the statocysts participate in the organization of search movements. We have studied the role of the statocysts in the control of search behavior in in vivo and in vitro experiments. It was found that the motor program typical of search movements could not be evoked after surgically removing the statocysts. Simultaneous recordings from SRCs and wing and tail nerves showed that each episode of “fictive searching” was accompanied by a bout of receptor activity even though the position of the preparation in the gravity field was constant. A model of a statocyst receptor network showed that, theoretically, the SRCs are not only able to respond to position changes in the gravitational field but also can generate a complex activity under influence of an excitatory input. The evidence presented here supports the idea that the statocysts are directly involved in the production of search behavior.


Behavioral experiments

The experimental procedures were carried out at the Ocean Sciences Center of Memorial University, St. John's, Newfoundland. Clione were collected off the coast and maintained in tanks with running seawater until the experiments. Animals of 3-5 cm in length were used. The experiments were done in a 70 × 70 × 70 cm glass tank. The temperature of water was about 10°C. The animal's image and its reflection in the mirror, arranged at about 45°, were recorded by a digital video camera (320 × 240 resolution, 30-Hz frame rate). The motion was tracked by a commercial software program (Shake 3, Apple Computers, Cupertino, CA). A direct linear transformation method was used to extract the three-dimensional coordinates from the two-dimensional coordinates and reconstruct the trajectory (Wood and Marshall 1986). Speed of motion was calculated from the reconstructed trajectories.

To estimate and compare quantitatively the complexity of the trajectories during different behaviors, we calculated their fractal dimension (D) as defined by Mandelbrot (1967). The fractal dimension provides a measure of the complexity of the trajectories (Nonnenmacher et al. 1993). The length of each trajectory is measured with different rulers of increasing sizes. For simple trajectories, such as a straight line, circle, or recurrent curve (i.e., a helix), there is weak dependence of the overall length on the ruler size. However, the overall length of complex trajectories will differ when measured with a long ruler or a short one because a long ruler cannot follow all the curvatures of the trajectory. Representative Clione swimming trajectory sections of 300 frames in duration were selected for each trajectory and used for the analysis. To calculate the fractal dimension, we used the Richardson plot: log of the size of each ruler (step length) versus log of the length of the trajectory measured with each step length (Nonnenmacher et al. 1993). The fractal dimension D is 1 - S, where S is the slope of the Richardson plot. The simplest trajectory yields D ≈ 1. The more the trajectory deviates from a simple curve, the larger is D. The slope of the Richardson plot was measured at the region with the highest negative slope. The results are shown as mean ± SD, and N is the number of trajectories analyzed. We used the nonparametric Mann-Whitney test to determine statistical significance at the 0.05 level.

To induce hunting, Limacina held in a pipette was presented to Clione and then removed once hunting started. To induce hunting-like behavior pharmacologically (Arshavsky et al. 1993), animals were immersed in 10-5 M physostigmine (Sigma) diluted in filtered seawater. They were kept in physostigmine solution until a tentacle protraction appeared and put into the tank for the video recording. The effect of immersion was similar to the effect of physostigmine injection into the hemocoel. The statocysts were removed through a pair of small incisions made on the dorsal skin surface, above the pedal ganglia.

Electrophysiological experiments

Preparations for electrophysiological experiments were made with animals in ice-cold seawater to prevent excitation of nociceptive afferent fibers. The preparation, including cerebral, pedal, and abdominal ganglia with the tail and wing nerves, was pinned to a Sylgardlined petri dish as previously described (Arshavsky et al. 1985; Panchin et al. 1995b). Extracellular recordings from nerves were made by using glass suction electrodes or stainless steel electrodes. To monitor tail activity, we recorded from pedal nerves 2 and 3, according to nomenclature used by Deliagina et al. (1999). Intracellular recordings were made using glass electrodes (10 MΩ) filled with 3 M KCl. The signals were acquired with a Digidata board (Axon Instruments) and stored for later analysis with Orbital Spike ( and custom-made spike sorting programs. To produce the instantaneous firing rate plots, spikes were detected, and spike timing was convolved with a Gaussian function (the spike density function in Orbital Spike manual). For the normalized plots, each value was divided by the maximum value of this plot. Fictive hunting behavior was induced by application of physostigmine. The seawater covering the isolated nervous system was replaced by physostigmine solution at a concentration of 10-5 M (Arshavsky et al. 1993).


A computational model was designed as described by Varona et al. (2002). The model consisted of six neuron-like elements [model statocyst receptor cells (MSRCs)] implemented in Lotka-Volterra type equations. Each MSRC sent and received two inhibitory signals to the rest of the network in an asymmetrical way (see Fig. 8A). All MSRCs received an excitatory input from a neuron (H in Fig. 8A) simulating CHunIN activity. Such a Lotka-Volterra type dynamical system demonstrates nonsteady state irregular dynamics. In particular, none of the MSRCs was able to fire permanently and suppress the other elements, i.e., to become a “winner” in such a competition. Rather, the “winner” changed sequentially. Consequently, this mode of dynamics was termed “winnerless competition” (Rabinovich et al. 2001). The Lotka-Volterra formalism allowed us to prove theoretically the stability of the sequential winnerless switching (Afraimovich et al. 2004). The irregular sequential switching can be obtained for a wide range of architectures when the main connections are nonreciprocal and inhibitory.

fig. 8.

Model for the statocyst network dynamics. A: architecture of the model network: color circles represent 6 MSRCs, and small black circles represent inhibitory connections; thicker traces mean stronger connections. CHunINs are represented by H, and excitatory connections to the MSRCs are depicted with small white circles. B: firing rates of the model receptor neurons (MSRCs) induced by activating the model CHunIN (H). Activities of the 1st MSRCs (units 1-3 in A) are shown in the top panel; bottom panel corresponds to the activity of units 4-6; the colors used correspond to the colors of MSRCs shown in A. Arrows show an example of the coordinated activation among units even when the neurons are silent for a long period. Firing rates and time scale are arbitrary.


Behavioral experiments

Figure 2 shows three-dimensional trajectories of Clione movements during different behaviors. Regular motor behavior is shown in Fig. 2A. Clione swam with its head up due to rhythmic wing movements. Periodically, wing oscillations stopped, and Clione sank passively. In 5-10 s, wing oscillations restarted, and Clione rose up along a more or less vertical trajectory. The average speed of swimming was 1.4 ± 0.3 cm/s (n = 5). Figure 2B shows a trajectory of swimming in statocystectomized Clione. The mollusk did not keep a vertical orientation and usually swam along a helix-like trajectory determined by a slight bending of the tail. The tail bending changed rarely and thus the direction of swimming changed rarely as well. Eventually, the animals sank to the bottom of the tank.

fig. 2.

Swimming trajectories during different behaviors. A: 3-dimensional trajectory of routine swimming. Here and in the following figures, different colors are used to emphasize the 3-dimensional perception of the trajectories and change according to the x axis; t is the duration of the trajectory. B: swimming trajectory of Clione with the statocysts surgically removed. C: trajectory of swimming during hunting behavior evoked by the contact with Limacina. D: trajectory of swimming after immersion of Clione in physostigmine solution (10-5 M). Blue arrows indicate the beginning of the trajectory.

Motor activity of intact Clione changed radically during hunting behavior triggered by a contact with Limacina (Fig. 2 C). The velocity of swimming increased several folds to 5.7 ± 0.5 cm/s (n = 4). Clione swam along a complex trajectory consisting of numerous twists and turns in different planes. The mollusk seemed to be scanning the surrounding space in search of prey. Successive turns of the trajectory were determined by changes in tail bending. These episodes of search behavior lasted from a few to hundreds of seconds.

Previously, it has been shown that hunting-like behavior of Clione can be evoked in the absence of prey by injection of the acetylcholinesterase inhibitor, physostigmine, into the hemocoel (see methods). Figure 2D shows an episode of pharmacologically induced search behavior. As during natural hunting, each episode lasted from a few to hundreds of seconds. The Clione swam with high speed of 4 ± 1 cm/s (n = 3) along a trajectory similar to that observed during natural hunting behavior (cf. Fig. 2, C and D). This similarity allowed us to use pharmacologically evoked search behavior as an experimental model of searching during actual hunting behavior.

The complexity of trajectories, as determined by their fractal dimension, during regular swimming was D = 1.85 ± 0.05 (n = 3). The complexity of trajectories in statocystectomized mollusks was only slightly higher (D = 2.2 ± 0.5; n = 3), and the difference was not statistically significant (P > 0.05). In contrast, the complexity of trajectories during search behavior was much higher than during regular swimming. It was equal to 3.9 ± 0.3 (n = 5) during hunting search triggered by a contact with the prey and 5.0 ± 0.8 (n = 9) during pharmacologically induced search behavior. The difference between D during regular swimming and during both natural and pharmacologically induced hunting behaviors was statistically significant (P < 0.05). The search trajectories were irregular, with frequent changes in the planes of looping. However, there was an underlying organization of motions since the trajectories consisted of characteristic loops confined within a definite space (see movies at

Fictive search activity in the isolated nervous system

To investigate the neural mechanisms that organize the search behavior, we performed electrophysiological experiments where we studied “fictive searching”. Because the tail plays a crucial role in determining the movement direction, we recorded the wing nerve activity together with the activity in tail nerves before and after physostigmine application. Simultaneous recordings from the wing nerve and three tail nerves were performed in 10 experiments and demonstrated similar results. One example is shown in Fig. 3, A-C. In this experiment, the wing nerve was recorded together with the right tail nerves, n2 and n3, and left nerve n3. Before physostigmine application, there was no coordination in activities of the wing and three tail nerves (Fig. 3A). During the period of recording (120 s), there was only one episode of swimming activity that was accompanied by an excitation in the left tail nerve n3. The pattern of nerve activities changed radically after physostigmine application (Fig. 3B). During the same period of recording, there were seven bursts of a coordinated activity in four nerves including an activation of the swimming generator (Fig. 3D) and complex patterns of excitation of the tail motoneurons. Each burst of coordinated activity in the wing and tail nerves represents an episode of fictive searching. The duration of the episodes in different experiments ranged from 7 to about 100 s, with intervals between episodes of 10-100 s. In the experiment shown in Fig. 3, the average duration of episodes and the average duration of intervals were equal to 12 ± 5 s (n = 19) and 43 ± 28 s, respectively. One can suggest that these patterns of motoneuron activity underlie complex movements of Clione during hunting.

fig. 3.

Fictive search behavior evoked by physostigmine (10-5 M) application to a preparation of the isolated nervous system. A: simultaneous extracellular recordings from 4 pedal nerves: the wing nerve, tail nerves 2 (n2-R) and 3 (n3-R) from the right side and nerve 3 from the left side (n3-L) before physostigmine application. B: same after physostigmine application. C: expansion of a “search” episode marked in B. Moments of clear changes in the activity of tail motor units recorded from different nerves are shown by dotted lines (phases 1, 2, and 3). D: another experiment; activity of wing motoneurons during a “fictive hunting” episode. Simultaneous extracellular recording from the wing nerve and intracellular recording from the large 2A wing motoneuron. This motoneuron is active during the ventral phase of the wing flexion in the swimming cycle (Arshavsky et al. 1985).

To analyze the motor pattern more closely, we detected individual units from the tail nerves and compared their activity during different search episodes. Figure 4 shows the activity of four tail motor units recorded from three different nerves at 3 of 19 successive hunting episodes. It can be seen that the activity of the motoneurons is irregular. All episodes shown in Fig. 4 are different in the timing of the activity in each individual unit. At the same time, the activation sequence of four different units is the same at least in the beginning of the episodes when they are most active. This indicates some order to the organization of tail motor activity, and consequently, that the tail movements are also organized during search movements. This may be necessary for the coordination of the motion within a restricted space as shown in Fig. 2, C and D.

fig. 4.

Instantaneous firing rate of 4 different units from tail nerves (n2-R, n3-R, and n2-L) during 3 successive episodes of fictive search behavior. We considered the beginning of the activity in the wing nerve as the beginning of the episode.

Correlated activity of SRCs and motor units

In fixed preparations, the statocysts are expected to show very little variation in activity because their position in the gravitational field does not change. Indeed, intracellular recordings from the SRCs showed that they were tonically active or quiescent (data not shown). Earlier it was found that the activity of SRCs changes during bouts of fictive hunting behavior (Arshavsky et al. 1993). Here we studied the activity of SRCs during fictive search behavior in more detail. SRCs were intracellularly recorded in 12 experiments. One of the examples is shown in Fig. 5. The activity of a SRC was recorded together with the wing 1A motoneuron, supplying the dorsal aspect of the wing (Arshavsky et al. 1985), and the activity in the tail nerve. Bursts in the wing and tail nerves followed one after another with short intervals (about 10 s). One can see that episodes of fictive search behavior, monitored by the bursts of activity both in the tail nerve and in the wing motoneuron, were accompanied by discharges of the SRC. Such a correlated activity between wing motoneurons, tail motoneurons, and SRCs was observed in all experiments. Particularly, in the experiment shown in Fig. 5, the increased firing rate in the tail motoneuron always followed the termination of the burst in the SRC.

fig. 5.

Simultaneous intracellular recording from wing motoneuron 1A, a SRC, and an extracellular recording from a tail nerve (n2-R) during fictive search behavior. Arrow indicates a burst of activity in the tail nerve following the termination of activity in the SRC.

To examine the timing of the SRC and motor activity during each search episode, we aligned the bursts in the wing nerve for several consecutive search episodes. Figure 6 shows four different episodes of fictive search behavior produced by physostigmine application. The pattern of SRC activity was irregular and changed from one episode to another. In two cases (Fig. 6, i and iii), the excitation of the SRC preceded the beginning of the hunting episode. In one case, the SRC was excited simultaneously with the beginning of an episode (Fig. 6ii), and in one case, the SRC was slightly inhibited (Fig. 6iv).

fig. 6.

Phase of statocyst receptor activity during fictive hunting behavior. Simultaneous intracellular recording from a statocyst receptor (SRC) and extracellular recording from the wing nerve. Beginning of each fictive search episode was aligned to compare the timing of activity in the SRC. Onset of the activity in the wing nerve is regarded as the beginning of the search episodes (marked by a dashed line). Bottom traces represent the instantaneous firing rate (IFR) of the SRC activity. Four episodes (i-iv) were chosen to show a high variability in receptor activity during different episodes.

Search behavior cannot occur without statocysts

Previously, we showed that the statocyst receptors do not only function as part of a gravimetric organ but can also be active independent of gravitational signals. To address the question of whether or not the statocysts are necessary for the production of searching, we conducted a series of behavioral experiments on statocystectomized animals. We studied only pharmacologically induced hunting behavior because, as it was shown earlier, natural hunting behavior is very sensitive and any intervention disrupts Clione's reaction to the prey. On the other hand, even strong interventions did not influence the effect of physostigmine (Arshavsky et al. 1993). As in experiments on intact mollusks, statocystectomized animals were exposed to a physostigmine solution until tentacle protraction appeared and then were put back into the tank. The speed of motion for statocystectomized animals exposed to physostigmine was 3.9 ± 0.5 (n = 9) cm/s. In this case, the animals could not control the direction of swimming and sank to the bottom of the tank as the statocystectomized animals not exposed to physostigmine. However, for the drug-exposed animals, the sinking motion was slower since the protracted tentacles added drag acting like a parachute. Our analysis showed that statocystectomized Clione never demonstrated the complex pattern of motion typical of search behavior after immersion in physostigmine (cf. Figs. 2D and 7A). The complexity of the trajectories in statocystectomized animals exposed (D = 2.0 ± 0.02; n = 9) and not exposed (D = 2.2 ± 0.5; n = 3) to physostigmine did not differ significantly (P > 0.05). In contrast, the complexity of the trajectories during physostigmine-evoked search behavior in intact animals was significantly higher (D = 5.0 ± 0.8; n = 9) than in statocystectomized animals exposed to physostigmine (P < 0.05).

fig. 7.

Decisive role of the statocysts in the control of search behavior. A: 3-dimensional swimming trajectory of statocystectomized animals after immersion in physostigmine. B and C: extracellular recordings from wing and tail nerve (n3-R) in a preparation isolated from Clione statocystectomized 2 days before the experiment. B: before physostigmine application. C: after physostigmine application. Note that the bursts of activity shown in Fig. 3B are absent in the preparation without the statocysts. Similar results were obtained in 3 other experiments.

The behavioral results were supported by in vitro experiments. Figure 7, B and C, shows an experiment performed on a preparation of the nervous system isolated from the statocystectomized animal. Unlike preparations with intact statocysts, physostigmine application failed to induce episodes of activity in wing and tail nerves typical of fictive searching (cf. Figs. 7C, 3B, and 5).

A model study of statocyst activity during hunting

Taking all the above into account, a possible explanation for the experimental results is that the basic spatio-temporal pattern underlying search behavior is partially generated within the statocyst network. The possibility for such a mechanism was studied using a model network of MSRCs (Fig. 8A, see methods) that is able to display a dual behavior. When there was no excitatory input to the MSRCs from the hunting neuron (H), only a single MSRC had a high level of activity. This activity mimicked the SRC response to the pressure from the statolith (data not shown). However, when the excitatory input from the model hunting neuron was turned on, the whole network was activated. In this case the activity of each MSRC depended not on the effect of the statolith pressure to a single MSRC, but on the intrinsic network dynamics. The model showed that a winnerless competition (see methods), induced by the excitatory input, generated an irregular switching among MSRCs. The characteristic switching dynamics evoked by turning on the excitatory input are shown in Fig. 8B. The model demonstrates two interesting features of the outputs produced by the MSRC network: 1) the activities of all MSRCs are nonsteady state and irregular, and 2) although the timing of the activity (when it is present) is irregular, the sequence of the switching among the MSRC activity is the same in each episode (see arrows in Fig. 8B). As we mentioned in methods, the irregular switching of activity can be obtained for a wide variety of architectures and connection strengths, provided a minimum number of nonreciprocal connections is present (Afraimovich et al. 2004; Varona et al. 2002). This activity pattern persists even when the variable that represents the stimulus from the statolith has a complex spatio-temporal pattern that mimics the statolith motion during hunting. In fact, any excitatory forcing represented by this variable (transient or permanent) to the statocyst receptors will not destroy the switching activity displayed by the network as long as the excitation from the model hunting neuron is stronger than the excitation from the statolith. The model thus shows that a network of nonreciprocally inhibiting neurons can generate a complex switching activity. These signals could be involved in the organization of the search motor program by affecting the activity of downstream cerebral and pedal neurons. If a winnerless competition among the SRCs is used for the organization of the motor program, a variable correlation between the SRCs and the motoneurons activity should be expected. In this framework, the model can help to understand the recordings shown in Fig. 6. Given that during a hunting episode the activity of the individual SRCs is set by its particular competition, excitation phases and levels will change between different episodes.


The main function of the statocysts is to inform the CNS about the animal's orientation in the gravitational field. A change in the animal's spatial orientation causes corrective motor responses directed to the restoration of the initial orientation. However, some features in the organization of the statocysts in mollusks suggest that the statocysts can perform more complex functions by being involved in the organization of complex motor behavior. First, SRCs interact with each other and, thus form neural networks (Detwiler and Alkon 1973; Panchin et al. 1995a). Second, activity of SRCs is determined not only by the animal orientation in the gravity field, but also by signals from the CNS (Janse et al. 1988; Salanki and Jahan-Parwar 1985; Wolff 1975). Here we studied the role of the statocysts in the organization of search movements during hunting behavior in the carnivorous mollusk, Clione. In the case of Clione, its gravimetric organ receives excitatory inputs from the CHunINs that play an important role in the integration of the activity of neural networks involved in the control of hunting behavior (Arshavsky et al. 1993). We have shown several lines of evidence that support the idea that signals from the statocysts are involved in the organization of the hunting search motor program in a more significant way than expected from a sensory system.

Clione search behavior is directed toward locating and capturing prey. The motor strategy during searching is likely to be determined by the difficulty in detecting an odor source in the water. As has been shown in our behavioral experiments, after triggering the hunting, the Clione search movements were not directed by the prey since they continued even when the prey was taken out of the water (see methods). In addition, a similar pattern of movements was observed during pharmacologically induced hunting-like behavior when a directional cue was completely absent. Thus both contact with the prey and the acetylcholinesterase inhibitor, physostigmine, triggered a stereotyped motor program similar to a fixed action pattern. The animal search movements during hunting, evoked either pharmacologically or by a prey, were confined within a definite space. In particular, when hunting was triggered by a contact with a prey, the trajectory of twisting and looping occurred within the area where the prey had been sensed. In an animal with an undeveloped visual system, such a motor strategy should increase the chance of locating and capturing the prey. A similar strategy is used by the freshwater snail Planorbis corneus. When searching food, Planorbis loops on a two-dimensional surface (Deliagina and Orlovsky 1990).

Since during hunting the orientation of Clione is continuously changing, it is evident that the signals from the statocyst receptors are not used for the stabilization of the orientation. To examine whether the statocysts play any role in the organization of search behavior, we studied the effect of physostigmine in statocystectomized animals. It was found that physostigmine could not evoke the motor program typical of searching after statocyst extirpation. This result was obtained both in in vivo and in vitro experiments. The inability of physostigmine to evoke search behavior can hardly be explained by a nonspecific traumatic effect of the operation since the statocysts, which are connected with the pedal ganglia by a connective tissue only, can be removed without touching the CNS. Statocystectomized Clione demonstrated normal activity of the swimming CPG monitored by wing movements, and the operation did not affect tentacle protraction evoked by physostigmine. Earlier it was shown that physostigmine evoked hunting-like behavior in a semi-intact preparation consisting of the CNS connected with the head and tail, which is a more severe operation (Arshavsky et al. 1993). Thus the results obtained in experiments with statocyst removal suggest that the statocysts are necessary for the production of the search program.

For studying the possible mechanism of statocyst participation in generating the motor search program, we recorded the activity of single SRCs together with the activity in motor nerves and motoneurons. The motor output typical of search behavior could be evoked in in vitro preparations by applying physostigmine. The basic pattern of motor output appeared to be generated independently of changes in the gravitational field as sensed by gravimetric organ, since the orientation of the preparation in the gravitational field was fixed. The recordings from the SRCs and motor nerves showed that their activity during fictive search behavior was correlated. In some cases an excitation of the SRCs preceded the beginning of fictive hunting episodes (Fig. 6, i and iii). This suggests that the SRC activity pattern is not likely to be determined by signals from motor centers involved in search movements (an efferent copy) but has another source. In fact, if the search motor pattern originated from the statocysts network, such a correlation should be expected.

Based on the results obtained in this work, we suggest the following hypothesis on the role of the statocysts in generating the motor program underlying search behavior. A contact with the prey or injection of physostigmine leads to excitation of the CHunINs (Arshavsky et al. 1993; Norekian and Satterlie 1995), which send an excitatory input to the SRCs. The network of the SRCs generates a complex output due to the intrinsic network dynamics (winnerless competition). This output is directed to the cerebral ganglia to participate in the forming of the search motor program. The theoretical validity for this hypothesis was demonstrated in the model study. It was found that a simple network formed by neuron-like elements with nonreciprocal inhibitory connections generated a complex switching activity in response to the excitatory input (H in Fig. 8A), mimicking the effect of the CHunINs. The excitation from the hunting neuron in the model was used only to trigger the activation of the receptor network dynamics. The model CHunIN activity did not require any temporal structure to produce the switching dynamics. In addition, the model showed that such a network can produce an ordered pattern of activation in otherwise irregular activity. This is desirable feature to organize a complex but coordinated motion. The known anatomical connections suggest that the statocyst output can be reflected, at least partly, in the activity of tail motoneurons. Therefore the model results are in correspondence with results obtained electrophysiological experiments. Indeed, recordings from the tail nerves showed an irregular activation of motor units during fictive hunting. However, at the same time, clear organization was apparent since the sequence of motor unit activation was preserved.

Other sensory systems affect molluskan statocysts (Alkon et al. 1978; Janse et al. 1988; Tsubata et al. 2003), and it has been suggested that the integration, even at the level of the gravimetric organ, is needed for the control of behavior. A compelling example has been reported in the pulmonate snail, Lymnaea stagnalis (Janse et al. 1988). Lymnaea migrates to the surface of water to breath and then dives again. Signals from oxygen-sensitive receptors are directed to the cerebral ganglia that, in turn, affect the activity of the SRCs. This modified activity of the SRCs determines downward-directed motions of mollusks in O2-rich conditions and upward motions in O2-poor conditions. This result also shows that the statocysts can participate in forming motor programs.

Although the statocysts are necessary for the generation of the search motor program in Clione, this does not mean that they are sufficient to produce the program. Our working hypothesis is that the motor dynamics strongly depend on signals from the statocysts. The SRC network activity propagates to the cerebral ganglia and to motoneurons located in the pedal ganglia. It is likely that motor centers in the cerebral and pedal ganglia make a contribution in the formation of the final motor program. The particular contribution of the statocysts and central mechanisms to the generation of the search program is therefore a subject of future studies.

Independent of the specific role played by the statocysts in Clione search behavior, the results obtained in this work, as well as the results obtained by Janse et al. (1988) on the pulmonate snail Lymnaea, show that the statocysts can perform dual functions depending on behavioral context. During regular swimming of Clione, the statocysts perform a purely sensory function and gravimetric reflexes are used for maintaining a vertical spatial orientation. In contrast, during hunting the statocysts participate in generating a motor search program. Therefore in both cases, the statocysts' output is used for organizing motor behavior, but the signals themselves are modified to fit the changing behavioral context.


We thank F. Cuthbert and the field services of the Ocean Science center, Memorial University of Newfoundland, for their administrative and logistic help. We also thank Drs. H. Chiel, G. Gillette, and V. Brezina for reading an earlier version of the manuscript and very helpful comments.


This work was supported by National Institute of Neurological Disorders and Stroke Grant 7RO1-NS-38022 and National Science Foundation Grant EIA-0130708.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract