Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Decision Making and Behavioral Choice during Predator Avoidance

Decision Making and Behavioral Choice during Predator Avoidance REVIEW ARTICLE published: 28 August 2012 doi: 10.3389/fnins.2012.00125 Decision making and behavioral choice during predator avoidance 1,2 2 Jens Herberholz * and Gregory D. Marquart Department of Psychology, University of Maryland, College Park, MD, USA Neuroscience and Cognitive Science Program, University of Maryland, College Park, MD, USA Edited by: One of the most important decisions animals have to make is how to respond to an attack Björn Brembs, Freie Universität from a potential predator. The response must be prompt and appropriate to ensure survival. Berlin, Germany Invertebrates have been important models in studying the underlying neurobiology of the Reviewed by: escape response due to their accessible nervous systems and easily quantifiable behavioral Cynthia M. Harley, University of output. Moreover, invertebrates provide opportunities for investigating these processes at a Minnesota, USA Jochen Smolka, Lund University, level of analysis not available in most other organisms. Recently, there has been a renewed Sweden focus in understanding how value-based calculations are made on the level of the nervous *Correspondence: system, i.e., when decisions are made under conflicting circumstances, and the most desir- Jens Herberholz, Department of able choice must be selected by weighing the costs and benefits for each behavioral choice. Psychology, University of Maryland, This article reviews samples from the current literature on anti-predator decision making in College Park, MD 20742, USA. e-mail: jherberh@umd.edu invertebrates, from single neurons to complex behaviors. Recent progress in understanding the mechanisms underlying value-based behavioral decisions is also discussed. Keywords: predation, escape, decision making, behavioral choice, neural circuits INTRODUCTION has been measured and described in a number of invertebrate Successful avoidance of a predatory attack is essential for sur- species. This has opened up exciting new avenues for gaining a vival and future reproductive success. Failure to detect a preda- better understanding of complex “neuroeconomic” processes at a tor before an attack initiation, failure to fight off an attack, or level of analysis not feasible in vertebrates. failure to respond to an attack with an immediate escape, can The first section of this review summarizes some of the fore- be deadly. Many aspects of nervous system function must be most examples of anti-predator behavior and underlying neural optimized to control anti-predator behavior, including careful circuitry found in four different arthropods. Both the specializa- sensory assessment of threat stimuli, which sometimes involves tions and shared features of these nervous systems that allow these multimodal integration, rapid transmission of this information animals to escape immediate predatory threats are discussed. The within neural structures, and finally, fast and accurate motor second part focuses on economic decisions made by invertebrates activation. Importantly, predator avoidance is often produced in situations where the risk of predation must be carefully weighed under conflicting circumstances. Many daily activities that are against other vitally important needs. Finally, we suggest some essential for survival, such as feeding, mate search, or habitat important future directions for the further identification of neural selection, can increase visibility and thus vulnerability to preda- mechanisms underlying behavioral decisions. tion. Animals trying to satisfy important needs while avoiding predation face a trade-off, e.g., between eating and the risk of MECHANISMS OF PREDATOR AVOIDANCE being eaten. Thus, the selection of the most desirable behav- While predators can provide direct cues such as visual or ior requires careful calculation of costs and benefits associated mechanosensory signals that alert prey to the presence of a preda- with different behavioral options. For example, foraging ani- tor, indirect cues, such as odors, also allow the assessment of a mals must accurately measure predation risk and weigh this risk potential predatory threat. However, indirect cues are frequently against current nutritional state. Such cost-benefit analyses are more ambiguous and seldom provide information on the degree or made by the nervous system through the integration of exter- immediacy of the danger posed. And indirect cues that signal the nal sensory signals with current internal states, and these deci- presence of a predator (although no predator is currently present) sions ideally lead to behavioral choices that optimize an animal’s can divert attention from other vital activities or suppress these fitness. activities altogether. Different risk assessment behaviors, appre- Invertebrates are superbly suited to measure both the behav- hension, and vigilance, are responses to indirect predator cues ior and neural mechanisms underlying predator avoidance. In commonly described in vertebrate animals (Kavaliers and Cho- many invertebrates, an accessible nervous system with described leris, 2001). Although they are likely to exist in invertebrates, these neural escape circuits controls discrete escape behaviors. Thus, “anticipatory” predator avoidance behaviors are much less studied the link between neural machinery and behavioral expression is in invertebrates where the evolution of extremely fast and power- often identifiable and quantifiable. More recently, economic deci- ful escape reactions in response to immediate attack has arguably sion making, i.e., costs-benefit calculations under predatory risk, reduced the necessity for extensive predator scanning and risk www.frontiersin.org August 2012 | Volume 6 | Article 125 | 1 Herberholz and Marquart Decisions underlying escape assessment. Additionally, while numerous behaviors in an ani- circuits in invertebrates are frequently divided into two broad mal’s repertoire contribute to predator avoidance, most are subtle categories: those that contain “command” or “command-like” ele- and difficult to subject to neurobiological analysis. For instance, ments and those that do not (Kupfermann and Weiss, 1978, 2001; an animal’s decision when and where to forage is greatly shaped by Edwards et al., 1999; Eaton et al., 2001). In command systems, the the risk of predation (Lima and Dill, 1990). How an animal calcu- activity of the command neuron is thought to be necessary and suf- lates this predatory risk and weighs it against concurrent internal ficient for the production of a behavior. Often a single spike in this and external demands is certainly an interesting question; however, neuron is sufficient for the readout of an entire escape program. the time-scale and context of such a decision make it difficult to While highly adaptive, these rapid behaviors are highly stereo- subject to detailed electrophysiological or neuroanatomical analy- typed, showing little variability. In contrast, the escape behaviors sis. Instead, what has overwhelmingly sufficed for the study of produced by systems ostensibly lacking a command element typ- predator avoidance in neuroscience has been the analysis of much ically display a greater degree of complexity and flexibility and more discrete escape or startle behaviors. Because escape behav- are frequently made up of a sequence of independently variable iors are so critical, they must interface with and frequently override components. This flexibility affords the animal a greater degree the performance of any ongoing or planned behaviors. And while of control over the precise timing, direction, and structure of the other behaviors may have a greater evolutionary importance over escape behavior. Traditionally, however, this is assumed to come the long term, seldom are they as time-sensitive and unforgiving at an additional computational cost that adds to the latency of the as escape. Thus, it is unsurprising that the circuits tasked with the action (Bullock, 1984). Alternatively, variability may be added to sensory acquisition, computation, and action upon salient preda- behavioral decisions by sequential neural processing. For exam- tory cues are frequently the largest, most robust, and most highly ple, in the medicinal leech decision neurons can be active during stereotyped neural systems in an organism. competing behaviors (e.g., swimming and body shortening), and If a predator is around, it is critical to identify and react to stimulation of one decision neuron can produce two different predatory cues at an appropriate time and in an effective man- behavioral outputs, swimming and crawling. Hypothesized to be ner. Consequently, escape behaviors must be fast, accurate, and organized in a hierarchical order, the first neuron in the chain robust in order to be effective countermeasures against the often would drive general behavioral action, the next one would com- rapid predatory behaviors they combat. It is believed that the time- mand selection from a pool of discrete motor patterns, and the sensitive nature of these behaviors necessitates a small number of next one would initiate the most desirable behavioral choice (Esch large elements in order to both maximize conduction velocity and and Kristan, 2002). minimize synaptic delay. Thus, escape circuits commonly have “giant fibers (GFs),” frequently the largest axons in an animal’s GIANT-NEURON MEDIATED ESCAPE nerve cord, which can be readily identified by their size, location, Crayfish or morphology. These characteristics allow for rapid identifica- Crayfish are equipped with powerful escape reactions mediated by tion and often make these neurons accessible to a wide range of rapidly responding neural circuits (reviewed in Wine and Krasne, cell biological and electrophysiological studies. 1982; Krasne and Wine, 1984; Edwards et al., 1999). These circuits Because of their simplicity and clear function, these circuits control at least three distinct motor programs that propel the ani- have been excellent models for the study of the neural basis of mals in different directions, but always away from real or assumed behavior. Recent work, however, has uncovered a surprising degree threats. Circuits and their associated tail-flips can be divided into of flexibility not previously recognized in these “simple,”“reflexive” two major categories, giant and non-giant. Two circuits, the lateral systems. High-speed video recordings have exposed a previously giant (LG) and medial giant (MG) system contain giant interneu- unappreciated level of complexity to arthropod escape behav- rons as key “command” components, are made for speed, and iors that has made researchers question the structure and even require strong and phasic input for their activation. In contrast, a identity of the underlying circuits that were originally assumed poorly elucidated non-giant system is believed to control slower, to be responsible for escape (Hammond and O’Shea, 2007a,b; but more variable escape tail-flips (Edwards et al., 1999). These Card and Dickinson, 2008a,b; Fotowat et al., 2009). Addition- escape circuits have been the focus of 65 years of intensive research ally, wireless-recording techniques have been adapted to small since they were first described by Wiersma (1947, 1952) in his invertebrate models allowing, for the first time, the correlation pioneering work. of neural activity from multiple identified neurons with the time- The LG interneurons, two large fibers consisting of a series course of escape behavior in unrestrained preparations (Fotowat of gap junction-linked neurons that project from tail to head, et al., 2011; Harrison et al., 2011). And while neural-behavioral are activated by tactile and strong hydrodynamic stimulation of correlations are not uncommon, escape behavior in invertebrates sensory hairs and proprioceptors located on the abdomen. The provides possibly one of the few opportunities to simultaneously LG interneurons also receive excitatory inputs from rostral sen- record from all the critical elements in a neural circuit and relate sory organs, but these inputs alone are insufficient to fire the LG. it to what is now appreciated as an increasingly complex, but still If these inputs sum with strong caudal inputs, however, a sin- tractable, behavior. This provides quite possibly one of the best gle LG action potential (in one of the two fibers) is sufficient to current opportunities for the comprehensive analysis of the neural produce an escape motion that thrusts the animal upward and underpinnings of decision making surrounding a behavior. away from the point of caudal stimulation (Liu and Herberholz, While there is likely a broad spectrum of complexity in the cir- 2010). The motor program is activated within milliseconds after cuits embedded in even the most simple nervous system, escape stimulation and speed and accuracy is guaranteed through several Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 2 Herberholz and Marquart Decisions underlying escape structural and functional specializations within the circuit (Her- berholz et al., 2002). Once activated, the LG interneurons drive giant motor neurons via rectifying electrical synapses, which acti- vate fast flexor muscles in the last two thoracic and first three abdominal segments causing a bending of the abdomen around the thoracic-abdominal joint and thus the stereotyped “jack-knife” motion that propels the animal upward (Wine and Krasne, 1972). Latency is minimal, with 5–15 ms between stimulation and start of the behavioral response, and varies according to both internal (e.g., animal size: Edwards et al., 1994) and external conditions (e.g., water temperature: Heitler and Edwards, 1998). This short latency is accomplished by the high transmission velocity due to the diameter of the GFs and by electrical coupling among most circuit components (Figure 5A). The MG interneurons, a pair of large fibers projecting from head to tail, are activated by strong, phasic visual or tactile inputs directed to the front of the animal. The MG interneurons receive their excitatory inputs in the brain where both neurons are elec- trically coupled to each other. One action potential in one of the MGs is sufficient to drive the fast and stereotyped backward escape response. The MG interneurons connect electrically to giant motor neurons, which activate fast flexor muscles in all abdominal seg- ments, causing the bending of the entire abdomen and propelling the animal backward away from the point of stimulation. MG tail- flips in response to tactile stimulation are as fast as LG-mediated tail-flips and happen within a few milliseconds (Wine and Krasne, 1972). Visually activated MG tail-flips are slower, but are still produced as quickly as 50 ms after detection of a visual danger stimulus (Liden and Herberholz, 2008; Liden et al., 2010). Non-giant-mediated tail-flips are controlled by a circuit that lacks giant interneurons. These tail-flips are elicited by a variety FIGURE 1 | Escape success and latencies measured in juvenile crayfish of different stimuli, typically more gradual and less forceful in attacked by dragonfly nymphs. (A) Attacks evoking tail-flips mediated by presentation than those activating giant-mediated tail-flips. They the medial giant (MG) or lateral giant (LG) interneurons are equally effective are produced with longer latencies, usually up to 10-fold slower to prevent capture whereas attacks eliciting non-giant (Non-G) tail-flips are than giant-mediated tail-flips, and considered, in a way, “volun- much less effective. (B) Unsuccessful MG and Non-G, but not LG tary” because the animal “chooses” to activate certain patterns responses are frequently followed by a series of Non-G tail-flips (left bars), which substantially increase the overall rate of escape (right bars). (C) of fast flexor muscle groups. Thus, the timing and direction of Escape latencies for crayfish attacked by predators (solid bars) or stimulated non-giant tail-flips can be modulated, resulting in a much more with a handheld probe (striped bars) are similar for giant mediated (MG and variable behavior compared to the giant-mediated tail-flips (Wine LG) tail-flips, but significantly shorter for predator evoked Non-G tail-flips. and Krasne, 1982; Wine, 1984). Non-giant tail-flips are also used Modified from Herberholz et al. (2004). during “swimming,” where a series of tail flexions and extensions propels the animal backward through the water. Although our understanding of the neural underpinnings of of all cases and escaping, after being captured, using a series of tail-flip escape, especially tail-flips produced by the LG circuit, is non-giant tail-flips in more than 75% of the remaining cases extensive and essentially unmatched by that of other experimen- (Figure 1B). Interestingly, latencies for non-giant tail-flips that tal models, our knowledge of escape circuit activation in response were produced as initial response to the predator strike were much to real predatory danger is virtually non-existent. Using dragon- shorter than latencies of non-giant tail-flips elicited by tactile stim- fly nymphs as natural predators, Herberholz et al. (2004) showed ulation with a handheld probe (Figure 1C). This suggests that that all three escape circuits of juvenile crayfish were activated in crayfish prepared the non-giant escape before the strike was deliv- response to attacks (Figure 1A). Initial escape responses to preda- ered, possibly integrating visual and hydrodynamic cues from the tory strikes were primarily mediated by giant tail-flips; frontal approaching predator in anticipation of the attack. The study also attacks evoked MG tail-flips whereas attacks directed to the rear revealed that crayfish relied entirely on their fast and powerful of the crayfish elicited LG tail-flips. While few attacks elicited tail-flip escape behaviors; crayfish showed no signs of predator non-giant tail-flips initially, overall escape performance improved recognition, vigilance, or avoidance behaviors in any of the trials substantially when non-giant tail-flips were produced following (Herberholz et al., 2004). Thus, the decision to escape, at least from capture. Overall, crayfish were successful at evading dragonfly this specific predator, is based on the activation of fixed action pat- nymphs, avoiding the predator’s strike with giant tail-flips in 50% terns elicited by predatory stimuli. The decision to escape is made www.frontiersin.org August 2012 | Volume 6 | Article 125 | 3 Herberholz and Marquart Decisions underlying escape at individual decision-making neurons; if the predatory signal is that rather than a simple escape jump, the escape behavior in sufficient to activate them, escape will inevitably follow. wild-type fruit flies involves a complex sequence of events con- sisting of at least four distinct subcomponents: an initial freeze Drosophila followed by postural adjustments, wing-elevation, and finally an There are a number of similarities between the GF system in escape jump coordinated with the initial down stroke of flight ini- Drosophila and the MG system in crayfish. Like the MG system, the tiation (Figure 2C). These behaviors do not appear to merely be GF system contains GFs originating in the brain that project down a fixed action pattern as new information continues to be inte- contralaterally to primary motor neurons that control the tho- grated into and affect subsequent components of the behaviors racic musculature responsible for the fruit fly’s escape behaviors even after sequence initiation (Hammond and O’Shea, 2007b). (reviewed in Wyman et al., 1984; Allen et al., 2006). In these giant These preflight behaviors were found to influence both the trajec- fibers, a single spike is normally sufficient for the activation of an tory as well as initial flight stability of the escape behavior (Card escape jump followed by flight initiation. Despite the motor por- and Dickinson, 2008b). tion of both the MG and GF being well described, comparatively This newly appreciated complexity of the response suggests little is known about the visual and mechanosensory pathways that that this escape behavior is either not in fact mediated by the GF feed into the giant fiber systems of either animal (Figures 5A,B). system or that additional unidentified pathways must be involved While the escape behaviors produced by these circuits are that are responsible for the preflight sequence that proceeds the extremely fast due to high conductance velocities and the minimal escape jump (Card and Dickinson, 2008b). Toward this end, evi- synaptic delay from a preponderance of electrical synapses, this dence for a previously unknown escape circuit was recorded by speed has generally been thought to come at the expense of flexi- Fotowat et al. (2009). In the absence of GF activation, the activ- bility (Bullock, 1984). Thus, giant-mediated escape behaviors are ity of this novel circuit correlated with the production of escape traditionally characterized as highly stereotyped with little vari- behavior in response to looming stimuli. While this pathway is ance in timing or direction; and whatever variance the result of yet to be anatomically identified, its activity shares features similar stochastic properties of the system and not the consequence of to well-described circuits responsive to looming stimuli in both neural computation (Bullock, 1984). vertebrates and invertebrates (e.g., pigeon: Sun and Frost, 1998; Although Drosophila has been a preeminent genetic model locust: Rind and Simmons, 1992; crab: Oliva et al., 2007; bull- since the start of the twentieth century, its diminutive size lim- frog: Nakagawa and Hongjian, 2010). All of this strongly suggests ited its use in electrophysiology until the 1970s (Bellen et al., that the GF system is not necessary for the production of escape 2010). And while the GF system was identified in 1948 (Power, behavior in the fruit fly, but that the GF system, possibly akin to 1948), it was not electrophysiologically characterized and linked the escape circuits in the crayfish, may be one of many present in to the production of escape behavior until the early 1980s (Wyman Drosophila. et al., 1984). This escape behavior was initially characterized as an Being that sudden changes in luminance (light-off ) are the only abbreviated form of “voluntary” flight initiation (Trimarchi and stimulus to reliably produce GF-mediated escape behavior, and Schneiderman, 1995a). While voluntary flight initiation is pre- then only in white-eyed fruit fly mutants, what role, if any, that ceded by a series of postural adjustments that prepare the fly for the GF system plays in actual escape behavior of wild-type fruit stable, directional flight, escape flight lacks these preflight pos- flies is now unclear. Although stimuli that reliably recruit the GF tural leg, and wing movements. Instead, escape initiation consists system in wild-type flies are unknown, it seems unlikely that the almost exclusively in the extension of the fruit fly’s mesothoracic GF system is simply the vestige of a lost escape circuit. While the legs that propels the insect off of the substrate, which is only then newly identified looming sensitive pathway might be tuned to a followed by the unfolding and initiation of wing movements (Card selective set of stimulus features, the GF system could still serve as and Dickinson, 2008a). a robust, broadly tuned escape circuit capable of producing rapid As the GF system was the only identified Drosophila escape escape behavior when more selective systems fail (Fotowat et al., circuit, it was assumed to mediate the escape behavior elicited 2009). by all visual, chemical, and mechanosensory stimuli that elicit an escape jump (McKenna et al., 1989). However, a num- VISUAL INTERNEURON MEDIATED ESCAPE ber of observations have accumulated that conflicted with this Locust canonical interpretation. For instance, in the housefly GF activ- While locusts produce avoidance behavior in response to a variety ity was shown not to be necessary for the production of of noxious stimuli (Riede, 1993; Friedel, 1999), the best studied of an escape jump in response to looming stimuli (Holmqvist, these are escape jumps in response to looming stimuli (reviewed 1994). Additionally, Trimarchi and Schneiderman (1995b) pro- in Pearson and O’Shea, 1984; Burrows, 1996; Figure 3). Like the vided evidence for an olfactory-induced flight initiation rem- escape behavior of fruit flies, the locust escape jump is a com- iniscent of the fruit flies’ escape behavior that was also not plex behavior composed of a sequence of distinct components, mediated by the GFs. More recently, the simplicity of the which allow the animal to direct this jump (Santer et al., 2005b). observed escape behavior was reassessed through high-speed In preparation for a jump, tilting postural movements mediated video analysis (Hammond and O’Shea, 2007a,b; Card and Dick- by the pro- and mesothoracic legs rotate the long axis of the locust inson, 2008a,b). This work illustrated that these “simple” escape toward the direction of the eventual jump (Hassenstein and Hus- behaviors were far more complex and nuanced than originally tert, 1999; Santer et al., 2005b; Figure 3A). The actual jump is assumed (Figures 2A,B). Card and Dickinson (2008a) showed produced through the cocking of the hindlegs, storage of energy Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 4 Herberholz and Marquart Decisions underlying escape FIGURE 2 | Escape flight planning and execution in Drosophila. dots mark head and abdomen points. (B) Probability that body parts of the fly (A) High-speed video sequence shows a typical escape to a looming frontal (black, T1 and T3 legs; red, T2 legs; blue, wings; gray, body) were moving prior stimulus with a prism allowing for simultaneous observation of ventral and to takeoff (green line). (C) As stimulus intensity increases, independent motor side profiles. Time stamps are milliseconds elapsed since stimulus onset. Red programs are activated eliciting discrete escape subbehaviors prior to takeoff. dots mark the initial contact point of the second leg tarsi with substrate. White Adapted with permission from Card and Dickinson (2008b). by the co-contraction of tibia flexor and extensor muscles, and sufficient energy in the animal’s hindlegs, co-contraction must finally the release of this energy, triggered by flexor inhibition begin as soon as possible in order to allow for a timely escape. (Burrows and Morris, 2001). Given the time required to store In contrast, the adjustment of pro- and mesothoracic limbs can www.frontiersin.org August 2012 | Volume 6 | Article 125 | 5 Herberholz and Marquart Decisions underlying escape FIGURE 3 | Escape jump and DCMD activity in locusts in response to extracellularly in the nerve cord from one locust (red traces). Raster plots show looming stimuli. (A) Four high-speed video frames from a locust producing an DCMD spikes recorded in 10 repetitions of the stimulus. Black and blue traces escape jump with time to collision listed in milliseconds. The position of the show average DCMD firing rate and its standard deviation, respectively. (D) femur-tibia joint is marked in red to calculate pixel movements of the joint. Timing of joint movements, DCMD peak and takeoff obtained from seven (B) Muscle recordings from the same trial. Stimulus angular size is shown on locusts. The DCMD peak occurred after the IJM and before the FJM and takeoff top with joint movements and flexor and extensor recordings below. (IJM, initial for all l /|v | values (l /|v |D ratio of stimulus radius (l ) to the velocity (v ) of the joint movement; FJM, final joint movement.) (C) DCMD activity measured stimulus). Adapted with permission from Fotowat and Gabbiani (2007). continue throughout co-contraction, allowing for alterations of LGMDs, the DCMDs produce action potentials in response to escape trajectory up until the escape jump is triggered (Santer looming stimuli, with their firing rate increasing as the looming et al., 2005b). On the other hand, if the hindlegs were used to object gets closer. Thus, the DCMDs were originally thought to control direction, it is thought that the decision of where to jump play a major role in jump production, sometimes compared to would have to be made over 100 ms before the jump is produced. the giant fibers in crayfish and fruit flies that control their fast Not only are locusts able to direct these jumps up to 50˚ to either escape maneuvers (Burrows, 1996). However, locusts prepare for side of their long axis, but their escape circuitry allows them to con- jumps by co-contracting flexor and extensor tibiae muscles for trol the timing, distance, and elevation of these jumps (Santer et al., s100 ms before the jump is released by relaxation of the flexor 2005b; Simmons et al., 2010). Similar to Drosophila, this complex muscles. Thus, the jump is not simply triggered by suprathresh- sequence of events does not appear to be a fixed action pattern old excitation of the DCMDs, because withdrawal of excitation that once initiated must be taken to completion as the locust can and inhibition are needed during the preparatory phase of the relax this co-contraction and release the stored up energy without jump (Figure 3B). Nevertheless, the DCMDs seem to participate the production of an escape jump (Heitler and Burrows, 1977). in all phases of the jump. Fotowat and Gabbiani (2007) compared Motor areas controlling these escape jumps are innervated by electrophysiological recordings with high-speed video recordings a pair of large interneurons, the descending contralateral move- and found that the rising phase of the firing rate of the DCMDs ment detectors (DCMDs) which receive excitatory inputs from coincided with the preparatory phase of the jump, whereas the lobula giant movement detector (LGMD) neurons that are respon- peak firing rate coincided with the co-activation period of flexor sive to looming stimuli. With a one-to-one relationship with the and extensor muscles, and decay of firing rate to less than 10% Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 6 Herberholz and Marquart Decisions underlying escape coincided with takeoff. This suggests that different stages of jump Three of these LG classes receive proprioceptive inputs from the production could be controlled by distinct phases in the firing legs, and thus could potentially integrate some contextual informa- pattern of the DCMDs (Figures 3C,D). Hindleg flexion in prepa- tion during predator escape (Berón de Astrada and Tomsic, 2002). ration for the jump, however, is not dependent on DCMD activity. Oliva et al. (2007) tested the escape behavior of grapsid crabs on When the connective containing the DCMD neuron was severed, a freely rotating styrofoam ball and recorded escape movements hindleg flexion still occurred, and it could also be evoked with (i.e., running) while looming stimuli were presented. They also visual stimuli that did not cause high firing activity in the DCMDs. recorded intracellularly from the LG neurons in restrained crabs This showed that while the activity of the DCMDs may contribute and compared these recordings with the behavioral data. Escape to hindleg flexion, it was not necessary for it and, thus, other runs were initiated soon after the LG neuron increased its fir- descending pathways would seem to be involved (Santer et al., ing rate, and after maximum stimulus expansion, the LG neurons 2008). Using a telemetry system to record DCMD and motor neu- stopped firing, coinciding with run deceleration in freely behaving ron activity in freely behaving locusts, it was found that the number animals. Moreover, the spike frequency of the LG neurons reflected of recorded DCMD spikes predicted motor neuron activity and the timing and speed of the escape response (Figures 4A,B). Inter- jump occurrence, and the time of peak firing rate predicted time estingly, the activity of the LG neurons is strongly affected by of takeoff (Fotowat et al., 2011). Although this underlined the role season with responses weaker in winter when predation risk is of the DCMDs as neurons exhibiting discrete firing responses to typically low and the animals are less active (Sztarker and Tomsic, looming stimuli, which in turn affected discrete stages of escape 2008). motor output, jump production remained intact, and occurred The relation between LG neuron activity and escape behavior at the same time as in control animals following DCMD abla- was also nicely demonstrated in experiments that tested short- tion. Thus, another neuron for jump production must exist, and term and long-term visual memory in crabs. Tomsic et al. (2003) this may be the descending ipsilateral movement detector neu- showed that LG neurons changed their responses to a visual threat ron (DIMD), which responds to looming targets, similarly to the (displacement of a black screen above the animal) in correspon- DCMD (Fotowat et al., 2011). Additionally, another descending dence with the behavioral changes observed in unrestrained ani- interneuron that responds to looming stimuli has recently been mals. Modification of LG neuron activity occurred during learning described. Thus visually mediated escape behavior in locusts is and persisted, after spaced training, for 24 h. However, while the likely controlled by at least three different descending neurons memory of freely behaving crabs reflects a strong stimulus-context (Gray et al., 2010). How these neurons interact to produce the association, LG neurons generalize the learned stimulus into new escape behavior remains to be determined (Figure 5C). spatial locations. Thus, despite being able to clearly distinguish the Locusts also produce an avoidance behavior during flight. learned stimulus from other similar stimuli (i.e., stimulus mem- When looming stimuli are presented, flying locusts produce a ory), the LG neurons do not appear to be involved in processing gliding dive similar to the dives used by other insects to evade contextual visual information (i.e., where the stimulus was learned; aerial predators. After DCMD neurons are activated by a loom- Sztarker and Tomsic, 2011). In summary, the LG neurons are ing stimulus, they produce short-latency excitatory postsynaptic sensory neurons located in the eyestalk, and their neural activity potentials (EPSPs) in a motor neuron that raises the wing into the patterns closely match escape behavior produced in unrestrained gliding posture. Stimuli that evoked high-frequency firing in the crabs (Medan et al., 2007). Their exact role in producing the DCMDs also reliably elicited the gliding response, and the behavior behavior, however, is unknown. To answer this question, detailed was less frequently observed when high-frequency DCMD spikes investigation of the descending pathways that connect the LG neu- were absent (Santer et al., 2005a). However, similar to the escape rons to the motor centers that control escape runs will be required jump, DCMD activity was not always sufficient to evoke gliding. (Figure 5D). Most likely, its high-frequency activity must be precisely timed with wingbeat phase because glides can only be produced during VALUE-BASED DECISION MAKING wing elevation. In addition, other neurons that are implicated in Adaptive behavioral decisions are essential for the survival and jump production (e.g., the DIMDs) may also be involved in escape reproductive success of most animals, including humans. Animals gliding in flying locusts (Santer et al., 2006). can typically choose from several behavioral alternatives, which need to be evaluated before the most desirable option is selected. To Crabs determine what behavior is most desirable at any given point, the The role of identified neurons in visually mediated escape behav- nervous system must integrate external conditions (e.g., predation ior has also recently been studied in grapsid crabs (reviewed risk) with current internal drives (e.g., hunger state), thus trading in Hemmi and Tomsic, 2012). The firing rate of these motion- off the costs and benefits of different alternatives before deciding sensitive lobula giant (LG) neurons in response to looming stim- which one to choose. For example, a hungry animal is more likely uli corresponds with the intensity of the crab’s escape behavior. to choose a behavioral option that involves risks because the value Four distinct classes of these neurons have been anatomically placed on foraging is greater than the value placed on other alter- and physiologically described. All four classes show wide-field natives such as hiding. If the benefit of finding a meal outweighs tangential arborization in the lobula, somata located beneath, the estimated cost of being attacked by a predator, the decision is and axons that project toward the midbrain; however, they to forage. If the value placed on foraging is low because the ani- are uniquely identifiable due to differences in morphology and mal is satiated, other behavioral options become more valuable response preferences (Medan et al., 2007). and behavioral output will shift toward less risky activities. The www.frontiersin.org August 2012 | Volume 6 | Article 125 | 7 Herberholz and Marquart Decisions underlying escape FIGURE 4 | Response of a crab’s LG neuron to looming stimuli and size of the looming object is shown in bottom trace. (B) Mean spike correlation with escape run. (A) Intracellular trace from one LG rate from a single LG neuron (top) and mean escape running speed neuron in response to a looming stimulus. Raster plot shows (bottom). Arrowheads mark the start of stimulus expansion and long responses from one neuron to nine repetitions of the stimulus. arrows mark increase in spike rate above resting level. Adapted with Histogram shows mean spike rate obtained from all nine trials. Angular permission from Oliva et al. (2007). literature on value-based decision making, especially with a focus review some recent experiments on value-based decision mak- on prey behavior in predator-prey interactions, is extensive and ing in response to predatory threat, and provide two examples covers a wide range of organisms (e.g., Ydenberg and Dill, 1986; where economic decisions can be linked to identifiable neural Lima and Dill, 1990). circuitry. The relatively new field of “neuroeconomics” is concerned with the neural underpinnings of value-based decision making CRAYFISH in humans and other non-human primates (Schall, 2001; Rangel When juvenile crayfish are exposed to fast-moving shadows while et al., 2008) and there is now fast growing interest in under- foraging in an artificial stream environment, they respond by standing the neural mechanisms that govern cost-benefit calcu- choosing one of two behavioral actions: they either freeze in lations. An increasing number of studies performed in humans place and remain motionless for several seconds before resum- and other primates are combining non-invasive techniques such ing foraging or they produce a tail-flip mediated by the MG as functional magnetic resonance imaging or cortical recordings neuron that propels the animal backward and away from the with discrimination tasks or cognitive experiments (Glimcher and approaching shadow and the expected food source (Liden and Rustichini, 2004; Huettel et al., 2005; Sugrue et al., 2005). The Herberholz, 2008; Figure 6A). Thus, crayfish respond to visual complexity of the mammalian brain, however, presents many threat signals that simulate the imminent attack of a preda- challenges. It is difficult to directly correlate neuronal activity tor with defensive behaviors that are discrete and incompatible. and behavioral expression and to obtain detailed information When Liden and Herberholz (2008) exposed groups of juvenile on neural circuit organization, cellular mechanisms, and the crayfish to different shadow velocities, they found that the frequen- interplay between sensory and motor systems. Decision-making cies of the two behavioral responses were dependent on shadow circuitry has been studied quite extensively in various inverte- speed. Slower moving shadows evoked more tail-flips than freez- brates, but descriptions of neural mechanisms underlying value- ing, but as shadow speed increased the frequency of tail-flips based (economic) behavioral decisions are rare (Kristan and decreased and crayfish primarily produced freezing behavior. The Gillette, 2007; Kristan, 2008). This is surprising because behav- study also showed that different individuals choose different anti- ioral experiments have shown that invertebrates make decisions predator strategies when exposed to one type of shadow. Some that are not always simple and reflexive, but are often the prod- animals decided to freeze in response to the danger signal while uct of careful cost-benefit calculations (Ydenberg and Dill, 1986; others decided to tail-flip. This suggests that different crayfish Lima and Dill, 1990; Chittka et al., 2009). Thus, invertebrates have different thresholds for each behavioral action, but what are ideally suited to study the neural mechanisms underlying underlies this difference remains to be determined. Because all value-based decision making. In the following section, we will tested animals were of identical size and shared the same social Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 8 Herberholz and Marquart Decisions underlying escape FIGURE 5 | Circuitry for arthropod escape behavior. Neural circuits sensory interneurons, projection (ascending or descending) or command underlying escape behaviors for crayfish (A), Drosophila (B), locust (C), and neurons, premotor neurons, and motor neurons with associated sensory crab (D) are illustrated. Circuits are divided into five levels: sensory neurons, (Continued) www.frontiersin.org August 2012 | Volume 6 | Article 125 | 9 Herberholz and Marquart Decisions underlying escape FIGURE 5 | Continued result of at least two circuits; a giant fiber (GF) system mediating jumps stimuli on the left and motor output on the right. Solid circles and lines lacking preparatory leg and wing movements and a yet to be identified represent identified neurons and connections while dashed circles and lines escape circuit that produces escape jumps with preparatory preflight limb represent neurons and connections yet to be identified. Stacked circles and wing adjustments. (PSI, peripherally synapsing interneuron, DLMns, represent a population of neurons. Lines end in four ways: with a dorsal lateral motor neurons, TTMn, tergotrochanteral muscle neuron.) perpendicular line, a concave cup, a circle, or dashes. Perpendicular lines (C) Locusts possess at least two escape circuits as well, one responsive to represent electrical synapses. Concave cups represent electrical synapses. looming stimuli and another responsive to auditory and mechanosensory Circles represent inhibitory synapses. Dashes indicate an unknown synapse stimuli. While numerous neurons that are believed to play a role in these type. Generic abbreviations: MSns, mechanosensory neurons; MSis, behaviors have been identified, both circuits remain incomplete. [LGMD, mechanosensory interneurons; VSns, visual sensory neurons; VSis, visual lobula giant movement detector neuron; LGMD2, lobula giant movement sensory interneurons; OSns, olfactory sensory neurons; OSis, olfactory detector neuron 2, DCMD, descending contralateral movement detector sensory neurons; ASns, auditory sensory neurons; ASis, auditory sensory neuron; DIMD, descending ipsilateral movement detector neuron; LDCMD, interneurons. (A) Crayfish tail-flips are controlled by one of three circuits, the late descending contralateral movement detector neuron, C, C (“cocking”) lateral giant (LG), medial giant (MG), and non-giant escape circuit. While the neuron, M, M-neuron, FETi, fast extensor tibia motor neuron, FLTis, flexor LG system is almost fully elucidated and the abdominal motor outputs of the tibia motor neurons, 714, neuron 714.] (D) In crabs, a class of visual MG are also well described, very little beyond the fast flexor motor neurons interneurons, the lobula giants (LGs), have been identified that are thought to (FFMns) are known to play a part in non-giant tail-flips. SG, segmental giant play a role in the crab’s escape behavior; however, no other elements in this neuron, MoG, motor giant neuron. (B) Drosophila escape jumps are the escape circuit have been elucidated. experiences and feeding history, other intrinsic factors must be SEA SLUG responsible. The marine snail has been a fruitful model for studying the Recently, Liden et al. (2010) used the same experimental design neural mechanisms underlying decision making and behavioral to show that crayfish base their escape decisions on the values choice. Using a “competing behaviors” paradigm, early work sug- of each behavioral option. They measured escape latencies for gested that different incompatible behaviors were organized in shadow-induced MG-mediated tail-flips by comparing photodi- a hierarchical model, each controlled by command-like neu- ode signals with bath electrode recordings that non-invasively rons that produced one behavior while inhibiting others. For captured neural and muscular activity produced during tail-flips example, when the sea slug was feeding, avoidance withdrawal (Figure 6B). They found that very fast approaching shadows in response to a tactile stimulus was suppressed (Kovac and become inescapable because they collided with the animal before a Davis, 1977). This suppression is caused by identified interneu- tail-flip could be generated. Moreover, tail-flips are costly because rons that are part of the motor circuit that generates feed- they move the animal away from the expected food source. Thus, ing. Thus, feeding behavior takes precedence over withdrawal, the observed suppression of tail-flipping in favor of freezing in while escape swimming dominates most other behaviors, includ- animals facing inescapable shadows, where the value of a tail- ing feeding (Jing and Gillette, 1995). The A1 neurons, a bilat- flip would be low, reflects the output product of an “economic” eral pair of interneurons located in the cerebropleural gan- decision-making process. Although tail-flipping is considered a glion of the snail, are necessary elements of the escape swim- less risky strategy when experiencing a predator attack, crayfish ming behavior, and their activity also inhibits feeding behav- also defaulted to freezing behavior when the expected reward ior. became more valuable. When food odor concentration in the arti- Recent work, however, has shown that sea slugs base their ficial stream was increased 10-fold, shadows that evoked mostly decisions on cost-benefit computations (Gillette et al., 2000; tail-flips under standard conditions now generated mainly freez- Figure 7). When presented with food stimuli, feeding behav- ing behavior. Interestingly, if high food value was paired with a ior or avoidance behavior can be activated, depending on the strong predator signal (a slow moving shadow) that reliably evoked concentration of the food stimulus and the current behavioral tail-flips under regular conditions, the behavioral shift toward state of the animal. At low concentrations and in satiated ani- freezing was less pronounced. Thus, a strong predator signal was mals, food stimuli typically evoked avoidance behavior. When the able to override the exaggerated food incentive (Figure 6C). This threshold for feeding was exceeded, avoidance behavior was sup- illustrates that crayfish calculate the costs and benefits of differ- pressed, and in hungry snails, even nociceptive stimuli elicited ent behavioral options and they carefully weigh predation risk feeding behavior (Figure 7A). This suggests both appetitive and against expected reward, eventually selecting the most valuable noxious stimuli provide inputs to neural networks underlying behavioral choice (Liden et al., 2010). Because these observed tail- feeding and avoidance behavior, but the final behavioral deci- flips are always generated by activation of MG neurons and the sion is determined by hunger state. Thus, in partially or fully MG circuit is accessible for neurophysiological and neurochemical satiated animals, the value placed on feeding behavior is low experiments, the neural workings underlying value-based deci- while it is high for avoidance behavior that protects the animal sion making in crayfish can now be investigated on the cellular from predators. Using a simple cost-benefit analysis, the animal level. This establishes the crayfish as an important new model weighs nutritional needs against predator risk and selects the for studying the neuroeconomics underlying predator avoidance. most desirable choice (Gillette et al., 2000; Figure 7B). Impor- However, to understand the decision-making process on the net- tantly, feeding and avoidance can be observed as fictive motor work level, identification of interneurons that form the descending patterns in isolated central nervous systems of the snail and some visual pathway for freezing behavior will be required. of the neurons controlling these behaviors have been individually Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 10 Herberholz and Marquart Decisions underlying escape FIGURE 6 | Escape choices and neural activation in crayfish exposed to collided with the animal and produced the peak response in PD no. 2. The first approaching shadows. (A) Experimental diagram and four video frames small deflection (arrow) in the BE trace is due to MG neuron activation, while illustrating a crayfish foraging (first two panels) and then tail-flipping (last two the large phasic potential and the smaller more erratic potentials that follow panels) in response to a fast approaching shadow with time in seconds. (B) are due to muscular activity during tail-flips. (C) Left: when exposed to a Left: example recordings from photodiodes positioned on the tank walls (PD medium speed shadow (2 m/s), crayfish produce fewer tail-flips (black bars) no. 1 and PD no. 2) when a shadow passes by, and from bath electrodes (BE) and more freezing (gray bars) when food odor concentration flowing through located inside the tank that capture field potentials generated during a tail-flip. the tank is high. Right: when exposed to slower (1 m/s) shadows, the effect of Right: Traces from PD no. 2 and BE at higher temporal resolution. In this food odor concentration on behavioral choice is less pronounced. (A) Modified example, animal initiated a tail-flip response (arrow) 4 ms before the shadow from Liden and Herberholz (2008). (B,C) Modified from Liden et al. (2010). identified (Jing and Gillette, 2003). Moreover, in isolated central electrically stimulated, avoidance turns were converted to orient- nervous systems, spontaneous feeding network activity reflects ing turns (Hirayama and Gillette, 2012). Thus, the neurophysi- feeding thresholds of the nervous system donors (for proboscis ological and neurochemical mechanisms underlying cost-benefit extension and biting); while orienting turns were more frequent calculations can now be investigated in the isolated nervous sys- in low-feeding threshold donors, avoidance turns dominated tem of this animal. This is expected to substantially contribute in high-feeding threshold donors. When a “command” neuron to our cellular understanding of value-based decision-making in the feeding network of a high-feeding threshold donor was processes. www.frontiersin.org August 2012 | Volume 6 | Article 125 | 11 Herberholz and Marquart Decisions underlying escape of behavioral events and the corresponding underlying neural mechanisms (Harley et al., 2009; Harley and Ritzmann, 2010). Based on the high-speed video analysis of the behavior of fruit flies and locusts, a reexamination of the “simple” escape behav- ior of other arthropods is warranted. Perhaps an analysis at a temporal resolution comparable to that of the speed of produc- tion of these behaviors will uncover a degree of flexibility and control not previously appreciated in these animals as well. For example, while the escape tail-flip and freezing behavior of the crayfish in response to visual stimuli have been assumed to be two distinct behaviors, which has been supported by video analy- sis at 250 fps (Liden et al., 2010), possibly higher speed analysis will show that these distinct decisions are in fact part of a single escape sequence. Such an observation could provide direction in the search for the neural circuit(s) responsible for freezing, the identification of which would provide a unique opportunity to explore decision making between two circuits underlying known behavioral alternatives. While this new appreciation for the complexity of arthropod escape behavior has reinvigorated work on giant fibers and escape behavior, it raises two significant issues. First, if the giant fiber systems previously assumed to underlie observed escape behav- iors are not in fact necessary or sufficient for the production of these behaviors, what circuits are? While Fotowat et al. (2009) have made initial progress toward characterizing the activity of part of an additional putative escape circuit, the neurons will have to be anatomically identified and the circuit fleshed out in future work. Second, if the giant fibers are not involved in escape behav- iors produced under existing experimental contexts, what contexts elicit their recruitment? It would be exceedingly wasteful for the largest axons in the fruit fly’s nerve cord to go unused. There must be some combination of internal states and external stimulus con- ditions that lead to GF-mediated escape response and work should be directed toward identifying these constraints. It is likely other arthropod models will have a similar redun- dancy in escape circuitry as has been described in the crayfish. Thus, a comprehensive understanding of decision making during predator avoidance will have to wait until all pathways and not just parts of some are fully characterized (Figure 5). While the identification of all escape circuits in any one arthropod is non- trivial, that parts of both command and non-command systems have been successfully identified in various arthropods is evidence FIGURE 7 | Effects of internal state on behavioral choice in a sea slug. of the feasibility of such a research program. For example, the LG (A) Four video frames showing feeding behavior in Pleurobranchaea neurons in grapsid crab are fully characterized and individually californica. Betaine application induces an orienting turn (panel 2) followed by proboscis extension and biting (panel 3). Chemosensory structures identifiable cells that can be accessed for intracellular recordings (panel 4): rhinophore (Rh), oral veil (OV), tentacle (Tn), and proboscis (Prob). in live animals. The activity of these neurons is highly correlated (B) Partial satiation raised the threshold for proboscis extension and biting with behavioral output, which suggests that they play a major role (i.e., feeding), and increased the frequency of withdrawal and turns (i.e., in mediating escape decisions. However, relevant analysis of the avoidance) in response to betaine. Modified from Gillette et al. (2000). complete escape circuit is still missing and descending pathways that orchestrate motor actions need to be identified. As such, future work should focus on completing the picture CONCLUSION AND FUTURE DIRECTIONS Recent work in the arthropods discussed suggests that the escape of currently known circuits, where often substantial sensory or motor elements remain poorly characterized, as well as identify- behavior of all may be more complex and varied than has generally been assumed. Quantitative ethograms that divide complex escape ing unknown but hinted at command or non-command circuits. This hunt for currently uncharacterized circuits might be aided by maneuvers into a sequence of simpler events can help identify variability within each system. Moreover, combining ethograms the possible similarity to and knowledge of already characterized systems found in related species (Figure 5). For instance, the with measures of neural structure or neural activity can elu- cidate the link between discrete motor actions within a series poorly studied non-giant tail-flip circuit in crayfish might share Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 12 Herberholz and Marquart Decisions underlying escape characteristics with that of the DCMD circuit in locusts and direct measures of nervous system function in natural settings is knowledge of the structure and function of the DCMD circuit highly desirable. could aid in the identification and characterization of this escape Finally, the neuromodulation of escape behavior by system. monoamines such as octopamine, serotonin and dopamine is Due to the assumption that giant fiber systems were a singular worth further exploration. Although a number of the escape cir- system responsible for the production of all escape behaviors, there cuits discussed have been shown to be responsive to the application is currently much confusion as to what discrete escape behavior is or removal of monoamines (Glanzman and Krasne, 1983; Busta- subserved by what specific circuit. Since it now appears that there mante and Krasne, 1991; Stern et al., 1995; Pflüger et al., 2004; are likely many circuits that produce a range of escape behav- Harvey et al., 2008; Rind et al., 2008), little is known about the iors, the spectrum of these behaviors and the stimulus conditions context in which these monoamines affect the performance of that lead to their display will need to be carefully cataloged and behavioral decisions. Since most invertebrate aminergic effects behavioral assays developed that can differentiate them. However, are mediated by metabotropic receptors that can have a grad- without the ability to simultaneously record both escape behaviors ual but pronounced impact on behavior, monoamines are an and neural activity, it will be difficult to ascribe a discrete escape attractive candidate for how a nervous system may be biased sequence or subcomponent of escape behavior to a particular cir- toward the production of one behavior over another (Crisp and cuit or set of neurons. For this, the use of telemetry that allows for Mesce, 2006; Mesce and Pierce-Shimomura, 2010). Through these in vivo recordings in freely behaving animals (Fotowat et al., 2011; monoamines, escape behaviors might modulate or be modulated Harrison et al., 2011) will have to be expanded to other inverte- by competing behaviors. Monoamines (e.g., dopamine and sero- brates. While it will be some time before these techniques can be tonin) have been targeted for roles in decision making and the adapted to all models, some should be able to benefit immediately. encoding of punishment and reward (Daw et al., 2002). Thus, the Arguably, these techniques might have the most to offer in models study of monoamines in the context of the evolutionarily criti- like the crayfish where large parts of a number of well-described cal task of predator avoidance provides an excellent opportunity escape circuits have long been worked out (Figure 5A). In such to explore the postulated neurochemical currency of neuroeco- a model, not only can the function of identified neurons be cor- nomic decision making. Unfortunately, little work on value-based related to the performance of distinct components of a complex decision making has been undertaken with invertebrates despite behavioral sequence, but also how an animal chooses between a the description of numerous value-based decisions that are likely range of escape behaviors might be elucidated. Recordings with to involve identified circuits including those mediating escape implanted electrodes or bath electrodes, which non-invasively or avoidance behavior. Research in this field is currently lim- record neural and muscular field potentials in freely behaving ited to a few invertebrate species, namely the previously discussed animals, have begun to reveal some of the basic neural patterns sea slug and crayfish, where basic neural mechanisms underly- underlying escape decisions in crayfish (Herberholz et al., 2001, ing cost-benefit computations have been partially uncovered. It is 2004; Liden and Herberholz, 2008; Liden et al., 2010). surprising that researchers interested in neuroeconomics have not There is a notable lack of neuroethological studies focused on taken greater advantage of these highly tractable models, as they escape mechanisms produced under natural conditions. While are likely to contribute much to this new field, as they have con- staged encounters with natural predators in the laboratory pro- tributed to neuroscience in general (Clarac and Pearlstein, 2007). vide some insight into the interplay between neural function and Possibly we have just begun to realize that invertebrate models are ecologically relevant escape behavior, these studies are sparse. Field ideally suited to answer some of the most challenging questions studies on the other hand are often focused on ecology and behav- faced today by neuroscience research. ior and not designed to investigate neural processes. Occasionally, data sets obtained separately in the field and laboratory allow for a ACKNOWLEDGMENT Part of the work reviewed in this article was supported by a comparative view and for correlating firing patterns of individual neurons and natural escape behavior (e.g., Hemmi and Tomsic, grant from the National Science Foundation to Jens Herberholz (IOS-0919845). 2012); however, the development of new technologies that permit REFERENCES neurons in a crab (Decapoda: in the locust. J. Exp. Biol. 204, Chittka, L., Skorupski, P., and Raine, N. Allen, M. J., Godenschwege, T. A., Brachyura). J. Comp. Physiol. A 188, 3471–3481. E. (2009). Speed-accuracy tradeoffs Tanouye, M. A., and Phelan, P. 539–551. Bustamante, J., and Krasne, F. B. (1991). in animal decision making. Trends (2006). Making an escape: develop- Bullock, T. H. (1984). “Compara- Effects of octopamine on transmis- Ecol. Evol. (Amst.) 24, 400–407. ment and function of the Drosophila tive neuroethology of startle, rapid sion at the first synapse of the cray- Clarac, F., and Pearlstein, E. (2007). giant fibre system. Semin. Cell Dev. escape, and giant fiber-mediated fish lateral giant escape reaction Invertebrate preparations and their Biol. 17, 31–41. responses,” in Neural Mechanisms pathway. J. Comp. Physiol. A 169, contribution to neurobiology in the Bellen, H. J., Tong, C., and Tsuda, of Startle Behavior, ed. R. C. 369–377. second half of the 20th century. H. (2010). 100 years of Drosophila Eaton (New York: Plenum Press), Card, G., and Dickinson, M. H. (2008a). Brain Res. Rev. 54, 113–161. research and its impact on verte- 1–13. Performance trade-offs in the flight Crisp, K. M., and Mesce, K. A. brate neuroscience: a history lesson Burrows, M. (1996). The Neurobiology initiation of Drosophila. J. Exp. Biol. (2006). Beyond the central pat- for the future. Nat. Rev. Neurosci. 11, of an Insect Brain. Oxford: Oxford 211, 341–353. tern generator: amine modulation 514–522. University Press. Card, G., and Dickinson, M. H. (2008b). of decision-making neural pathways Berón de Astrada, M., and Tomsic, D. Burrows, M., and Morris, G. (2001). Visually mediated motor planning in descending from the brain of the (2002). Physiology and morphol- The kinematics and neural control the escape response of Drosophila. medicinal leech. J. Exp. Biol. 209, ogy of visual movement detector of high-speed kicking movements Curr. Biol. 18, 1300–1307. 1746–1756. www.frontiersin.org August 2012 | Volume 6 | Article 125 | 13 Herberholz and Marquart Decisions underlying escape Daw, N. D., Kakade, S., and Dayan, Hammond, S., and O’Shea, M. (2007a). Hirayama, K., and Gillette, R. (2012). and prospectus. Can. J. Zool. 68, P. (2002). Opponent interactions Escape flight initiation in the fly. J. A neuronal network switch for 619–640. between serotonin and dopamine. Comp. Physiol. A 193, 471–476. approach/avoidance toggled by Liu, Y. C., and Herberholz, J. (2010). Neural Netw. 15, 603–616. Hammond, S., and O’Shea, M. (2007b). appetitive state. Curr. Biol. 22, Sensory activation and receptive Eaton, R. C., Lee, R. K., and Foreman, Ontogeny of flight initiation in the 118–123. field organization of the lateral giant M. B. (2001). The Mauthner cell fly Drosophila melanogaster : impli- Holmqvist, M. H. (1994). A visually escape neurons in crayfish. J. Neuro- and other identified neurons of the cations for the giant fibre sys- elicited escape response in the fly physiol. 104, 675–684. brainstem escape network of fish. tem. J. Comp. Physiol. A Neuroethol. that does not use the giant fiber path- McKenna, M., Monte, P., Helfand, S. L., Prog. Neurobiol. 63, 467–485. Sens. Neural. Behav. Physiol. 193, way. Vis. Neurosci. 11, 1149–1161. Woodard, C., and Carlson, J. (1989). Edwards, D. H., and Heitler, W. J., 1125–1137. Huettel, S. A., Song, A. W., and A simple chemosensory response and Krasne, F. B. (1999). Fifty years Harley, C. M., English, B. A., and Ritz- McCarthy, G. (2005). Decisions in Drosophila and the isolation of of a command neuron: the neu- mann, R. E. (2009). Characterization under uncertainty: probabilistic acj mutants in which it is affected. robiology of escape behavior in of obstacle negotiation behaviors in context influences activation of Proc. Natl. Acad. Sci. U.S.A. 86, the crayfish. Trends Neurosci. 22, the cockroach, Blaberus discoidalis. J. prefrontal and parietal cortices. J. 8118–8122. 153–161. Exp. Biol. 212, 1463–1476. Neurosci. 25, 3304–3311. Medan, V., Oliva, D., and Tomsic, D. Edwards, D. H., Yeh, S. R., Barnett, Harley, C. M., and Ritzmann, R. E. Jing, J., and Gillette, R. (1995). Neuronal (2007). Characterization of lobula L. D., and Nagappan, P. R. (1994). (2010). Electrolytic lesions within elements that mediate escape swim- giant neurons responsive to visual Changes in synaptic integration dur- central complex neuropils of the ming and suppress feeding behav- stimuli that elicit escape behaviors ing the growth of the lateral giant cockroach brain affect negotiation of ior in the predatory sea slug Pleu- in the crab Chasmagnathus. J. Neu- neuron of crayfish. J. Neurophysiol. barriers. J. Exp. Biol. 213, 2851–2864. robranchaea. J. Neurophysiol. 74, rophysiol. 98, 2414–2428. 72, 899–908. Harrison, R. R., Fotowat, H., Chan, R., 1900–1910. Mesce, K. A., and Pierce-Shimomura, Esch, T., and Kristan, W. B. (2002). Kier, R. J., Olberg, R., Leonardo, Jing, J., and Gillette, R. (2003). Direc- J. T. (2010). Shared strategies for Decision-making in the leech ner- A., and Gabbiani, F. (2011). Wire- tional avoidance turns encoded by behavioral switching: understanding vous system. Intgr. Comp. Biol. 42, less neural/EMG telemetry systems single interneurons and sustained by how locomotor patterns are turned 716–724. for small freely moving animals. multifunctional serotonergic cells J. on and off. Front. Behav. Neurosci. Fotowat, H., Fayyazuddin, A., Bellen, IEEE Trans. Biomed. Circuits Syst. 5, Neurosci. 23, 3039–3051. 4:49. doi:10.3389/fnbeh.2010.00049 J. H., and Gabbiani, F. (2009). A 103–111. Kavaliers, M., and Choleris, E. (2001). Nakagawa, H., and Hongjian, K. (2010). novel neuronal pathway for visu- Harvey, J., Brunger, H., Middleton, C. Antipredator responses and defen- Collision-sensitive neurons in the ally guided escape in Drosophila A., Hill, J. A., Sevdali, M., Sweeney, sive behavior: ecological and etho- optic tectum of the bullfrog, Rana melanogaster. J. Neurophysiol. 102, S. T., Sparrow, J. C., and Elliot, C. logical approaches for the neuro- catesbeiana. J. Neurophysiol. 104, 875–885. J. H. (2008). Neuromuscular control sciences. Neurosci. Biobehav. Rev. 25, 2487–2499. Fotowat, H., and Gabbiani, F. (2007). of a single twitch muscle in wild type 577–586. Oliva, D., Medan, V., and Tomsic, D. Relationship between the phases of and mutant Drosophila, measured Kovac, M. P., and Davis, W. J. (1977). (2007). Escape behavior and neu- sensory and motor activity dur- with an ergometer. Invert. Neurosci. Behavioral choice: neural mecha- ronal responses to looming stimuli ing a looming-evoked multistage 8, 63–70. nisms in Pleurobranchaea. Science in the crab Chasmagnathus granu- escape behavior. J. Neurosci. 27, Hassenstein, B., and Hustert, R. (1999). 198, 632–634. latus (Decapoda: Grapsidae). J. Exp. 10047–10059. Hiding responses of locusts to Krasne, F. B., and Wine, J. J. (1984). Biol. 210, 865–880. Fotowat, H., Harrison, R. R., and Gab- approaching objects. J. Exp. Biol. “The production of crayfish tailflip Pearson, K. G., and O’Shea, M. (1984). biani, F. (2011). Multiplexing of 202, 1701–1710. escape responses,” in Neural Mecha- “Escape behavior of the locust. The motor information in the discharge Heitler, W. J., and Burrows, M. nisms of Startle Behavior, ed. R. C. jump and its initiation by visual of a collision detecting neuron dur- (1977). The locust jump. I. The Eaton (New York: Plenum Press), stimuli,” in Neural Mechanisms of ing escape behaviors. Neuron 69, motor programme. J. Exp. Biol. 66, 179–211. Startle Behavior, ed. R. C. Eaton 147–158. 203–219. Kristan, W. B. (2008). Neuronal (New York: Plenum Press), 163–178. Friedel, T. (1999). The vibrational star- Heitler, W. J., and Edwards, D. H. (1998). decision-making circuits. Curr. Biol. Pflüger, H. J., Duch, C., and Hei- tle response of the desert locust Effect of temperature on a voltage- 18, 928–932. del, E. (2004). Neuromodulatory Schistocerca gregaria. J. Exp. Biol. sensitive electrical synapse in cray- Kristan, W. B., and Gillette, R. (2007). octopaminergic neurons and their 202, 2151–2159. fish. J. Exp. Biol. 201, 503–513. “Behavioral choice,” in Invertebrate functions during insect motor Gillette, R., Huang, R. C., Hatcher, Hemmi, J. M., and Tomsic, D. (2012). Neurobiology, eds G. North and R. J. behaviour. The Ernst Florey mem- N., and Moroz, L. L. (2000). Cost- The neuroethology of escape in Greenspan (New York: Cold Spring ory lecture. Acta Biol. Hung. 55, benefit analysis potential in feed- crabs: from sensory ecology to neu- Harbor Laboratory Press), 533–553. 3–12. ing behavior of a predatory snail rons and back. Curr. Opin. Neuro- Kupfermann, I., and Weiss, K. R. Power, M. E. (1948). The thoracico- by integration of hunger, taste, and biol. 22, 194–200. (1978). The command neuron con- abdominal nervous system of pain. Proc. Natl. Acad. Sci. U.S.A. 97, Herberholz, J., Antonsen, B. L., and cept. Behav. Brain Sci. 1, 3–10. an adult insect, Drosophila 3585–3590. Edwards, D. H. (2002). A lateral Kupfermann, I., and Weiss, K. R. (2001). melanogaster. J. Comp. Neurol. Glanzman, D. L., and Krasne, F. B. excitatory network in the escape Motor program selection in simple 88, 347–409. (1983). Serotonin and octopamine circuit of crayfish. J. Neurosci. 22, model systems. Curr. Opin. Neuro- Rangel, A., Camerer, C., and Montague, have opposite modulatory effects on 9078–9085. biol. 11, 673–677. P. R. (2008). A framework for study- the crayfish’s lateral giant escape Herberholz, J., Issa, F. A., and Edwards, Liden, W. H., and Herberholz, J. (2008). ing the neurobiology of value-based reaction. J. Neurosci. 3, 2263–2269. D. H. (2001). Patterns of neural cir- Behavioral and neural responses of decision making. Nat. Rev. Neurosci. Glimcher, P. W., and Rustichini, A. cuit activation and behavior during juvenile crayfish to moving shadows. 9, 545–556. (2004). Neuroeconomics: the con- dominance hierarchy formation in J. Exp. Biol. 211, 1355–1361. Riede, K. (1993). Prepulse inhibi- silience of brain and decision. Sci- freely behaving crayfish. J. Neurosci. Liden, W. H., Phillips, M. L., and Her- tion of the startle reaction in the ence 306, 447–452. 21, 2759–2767. berholz, J. (2010). Neural control of locust Locusta migratoria (Insecta: Gray, J. R., Blincow, E., and Robertson, Herberholz, J., Sen, M. M., and Edwards, behavioural choice in juvenile cray- Orthoptera: Acridoidea). J. Comp. R. M. (2010). A pair of motion- D. H. (2004). Escape behavior and fish. Proc. Biol. Sci. 277, 3493–3500. Physiol. A 172, 351–358. sensitive neurons in the locust escape circuit activation in juvenile Lima, S. L., and Dill, L. M. (1990). Rind, F. C., Santer, R. D., and Wright, encode approaches of a looming crayfish during prey-predator inter- Behavioral decisions made under G. A. (2008). Arousal facilitates object. J. Comp. Physiol. A 927–938. actions. J. Exp. Biol. 207, 1855–1863. the risk of predation: a review collision avoidance mediated by a Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 14 Herberholz and Marquart Decisions underlying escape looming sensitive visual neuron in preparation: the neuroethology of short- and long-term visual mem- Neural Mechanisms of Startle Behav- a flying locust. J. Neurophysiol. 100, visual startle in locusts. J. Insect ory in an arthropod. J. Neurosci. 23, ior, ed. R. C. Eaton (New York: 670–680. Physiol. 56, 876–883. 8539–8546. Plenum Press), 133–161. Rind, F. C., and Simmons, P. (1992). Stern, M., Thompson, K., and Zhou, Trimarchi, J. R., and Schneiderman, Ydenberg, R. C., and Dill, L. M. Orthopteran DCMD neuron: a P. (1995). Octopaminergic neurons A. M. (1995a). Initiation of flight (1986). The economics of fleeing reevaluation of responses to mov- in the locust brain: morphologi- in the unrestrained fly, Drosophila from predators. Adv. Study Behav. 16, ing objects. I. Selective responses to cal, biochemical and electrophysi- melanogaster. J. Zool. 235, 211–222. 229–249. approaching objects. J. Neurophysiol. ological characterisation of poten- Trimarchi, J. R., and Schneiderman, A. 68, 1654–1666. tial modulators of the visual sys- M. (1995b). Different neural path- Conflict of Interest Statement: The Santer, R. D., Rind, F. C., Stafford, R., tem. J. Comp. Physiol. A 177, ways coordinate Drosophila flight authors declare that the research was and Simmons, P. J. (2006). Role of an 611–625. initiations evoked by visual and conducted in the absence of any com- identified looming-sensitive neuron Sugrue, L. P., Corrado, G. S., and New- olfactory stimuli. J. Exp. Biol. 198, mercial or financial relationships that in triggering a flying locust’s escape. some, W. T. (2005). Choosing the 1099–1104. could be construed as a potential con- J. Neurophysiol. 95, 3391–3400. greater of two goods: neural curren- Wiersma, C. A. G. (1947). Giant nerve flict of interest. Santer, R. D., Simmons, P. J., and Rind, cies for valuation and decision mak- fiber system of the crayfish; a contri- F. C. (2005a). Gliding behaviour ing. Nat. Rev. Neurosci. 6, 363–375. bution to comparative physiology of Received: 07 March 2012; paper pend- elicited by lateral looming stimuli Sun, H., and Frost, B. (1998). Computa- synapse. J. Neurophysiol. 10, 23–38. ing published: 23 April 2012; accepted: 08 in flying locusts. J. Comp. Physiol. A tion of different optical variables of Wiersma, C. A. G. (1952). Repetitive dis- August 2012; published online: 28 August 191, 61–73. looming objects in pigeon nucleus charges of motor fibers caused by a 2012. Santer, R. D., Yamawaki, Y., Rind, F. C., rotundus neurons. Nat. Neurosci. 1, single impulse in giant fibers of the Citation: Herberholz J and Marquart and Simmons, P. J. (2005b). Motor 296–303. crayfish. J. Cell. Physiol. 40, 399–419. GD (2012) Decision making and behav- activity and trajectory control dur- Sztarker, J., and Tomsic, D. (2008). Wine, J. J. (1984). The structural basis ioral choice during predator avoid- ing escape jumping in the locust Neuronal correlates of the visually of an innate behavioural pattern. J. ance. Front. Neurosci. 6:125. doi: Locusta migratoria. J. Comp. Physiol. elicited escape response of the crab Exp. Biol. 112, 283–319. 10.3389/fnins.2012.00125 A 191, 965–975. Chasmagnathus upon seasonal vari- Wine, J. J., and Krasne, F. B. (1972). The This article was submitted to Frontiers Santer, R. D., Yamawaki, Y., Rind, F. C., ations, stimuli changes and percep- organization of escape behaviour in in Decision Neuroscience, a specialty of and Simmons, P. J. (2008). Prepar- tual alterations. J. Comp. Physiol. A the crayfish. J. Exp. Biol. 56, 1–18. Frontiers in Neuroscience. ing for escape: an examination of the 194, 587–596. Wine, J. J., and Krasne, F. B. (1982). Copyright © 2012 Herberholz and Mar- role of the DCMD neuron in locust Sztarker, J., and Tomsic, D. (2011). Brain “The cellular organization of cray- quart. This is an open-access article dis- escape jumps. J. Comp. Physiol. A modularity in arthropods: individ- fish escape behavior,” in The Biol- tributed under the terms of the Creative 194, 69–77. ual neurons that support “what” but ogy of Crustacea, eds D. C. Sande- Commons Attribution License, which Schall, J. D. (2001). Neural basis of not “where” memories. J. Neurosci. man and H. L. Atwood (New York: permits use, distribution and reproduc- deciding, choosing and acting. Nat. 31, 8175–8180. Academic Press), 241–292. tion in other forums, provided the original Rev. Neurosci. 2, 32–42. Tomsic, D., Berón de Astrada, M., Wyman, R. J., Thomas, J. B., Salkoff, authors and source are credited and sub- Simmons, P. J., Rind, F. C., and Santer, R. and Sztarker, J. (2003). Identifica- L., and King, D. G. (1984). “The ject to any copyright notices concerning D. (2010). Escapes with and without tion of individual neurons reflecting Drosophila giant fiber system,” in any third-party graphics etc. www.frontiersin.org August 2012 | Volume 6 | Article 125 | 15 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Frontiers in Neuroscience Unpaywall

Decision Making and Behavioral Choice during Predator Avoidance

Frontiers in NeuroscienceJan 1, 2012

Decision Making and Behavioral Choice during Predator Avoidance

Abstract

REVIEW ARTICLE published: 28 August 2012 doi: 10.3389/fnins.2012.00125 Decision making and behavioral choice during predator avoidance 1,2 2 Jens Herberholz * and Gregory D. Marquart Department of Psychology, University of Maryland, College Park, MD, USA Neuroscience and Cognitive Science Program, University of Maryland, College Park, MD, USA Edited by: One of the most important decisions animals have to make is how to respond to an attack Björn Brembs, Freie Universität from a potential predator. The response must be prompt and appropriate to ensure survival. Berlin, Germany Invertebrates have been important models in studying the underlying neurobiology of the Reviewed by: escape response due to their accessible nervous systems and easily quantifiable behavioral Cynthia M. Harley, University of output. Moreover, invertebrates provide opportunities for investigating these processes at a Minnesota, USA Jochen Smolka, Lund University, level of analysis not available in most other organisms. Recently, there has been a renewed Sweden focus in understanding how value-based calculations are made on the level of the nervous *Correspondence: system, i.e., when decisions are made under conflicting circumstances, and the most desir- Jens Herberholz, Department of able choice must be selected by weighing the costs and benefits for each behavioral choice. Psychology, University

Loading next page...
 
/lp/unpaywall/decision-making-and-behavioral-choice-during-predator-avoidance-gP39kS6szx

References (177)

Publisher
Unpaywall
ISSN
1662-453X
DOI
10.3389/fnins.2012.00125
Publisher site
See Article on Publisher Site

Abstract

REVIEW ARTICLE published: 28 August 2012 doi: 10.3389/fnins.2012.00125 Decision making and behavioral choice during predator avoidance 1,2 2 Jens Herberholz * and Gregory D. Marquart Department of Psychology, University of Maryland, College Park, MD, USA Neuroscience and Cognitive Science Program, University of Maryland, College Park, MD, USA Edited by: One of the most important decisions animals have to make is how to respond to an attack Björn Brembs, Freie Universität from a potential predator. The response must be prompt and appropriate to ensure survival. Berlin, Germany Invertebrates have been important models in studying the underlying neurobiology of the Reviewed by: escape response due to their accessible nervous systems and easily quantifiable behavioral Cynthia M. Harley, University of output. Moreover, invertebrates provide opportunities for investigating these processes at a Minnesota, USA Jochen Smolka, Lund University, level of analysis not available in most other organisms. Recently, there has been a renewed Sweden focus in understanding how value-based calculations are made on the level of the nervous *Correspondence: system, i.e., when decisions are made under conflicting circumstances, and the most desir- Jens Herberholz, Department of able choice must be selected by weighing the costs and benefits for each behavioral choice. Psychology, University of Maryland, This article reviews samples from the current literature on anti-predator decision making in College Park, MD 20742, USA. e-mail: jherberh@umd.edu invertebrates, from single neurons to complex behaviors. Recent progress in understanding the mechanisms underlying value-based behavioral decisions is also discussed. Keywords: predation, escape, decision making, behavioral choice, neural circuits INTRODUCTION has been measured and described in a number of invertebrate Successful avoidance of a predatory attack is essential for sur- species. This has opened up exciting new avenues for gaining a vival and future reproductive success. Failure to detect a preda- better understanding of complex “neuroeconomic” processes at a tor before an attack initiation, failure to fight off an attack, or level of analysis not feasible in vertebrates. failure to respond to an attack with an immediate escape, can The first section of this review summarizes some of the fore- be deadly. Many aspects of nervous system function must be most examples of anti-predator behavior and underlying neural optimized to control anti-predator behavior, including careful circuitry found in four different arthropods. Both the specializa- sensory assessment of threat stimuli, which sometimes involves tions and shared features of these nervous systems that allow these multimodal integration, rapid transmission of this information animals to escape immediate predatory threats are discussed. The within neural structures, and finally, fast and accurate motor second part focuses on economic decisions made by invertebrates activation. Importantly, predator avoidance is often produced in situations where the risk of predation must be carefully weighed under conflicting circumstances. Many daily activities that are against other vitally important needs. Finally, we suggest some essential for survival, such as feeding, mate search, or habitat important future directions for the further identification of neural selection, can increase visibility and thus vulnerability to preda- mechanisms underlying behavioral decisions. tion. Animals trying to satisfy important needs while avoiding predation face a trade-off, e.g., between eating and the risk of MECHANISMS OF PREDATOR AVOIDANCE being eaten. Thus, the selection of the most desirable behav- While predators can provide direct cues such as visual or ior requires careful calculation of costs and benefits associated mechanosensory signals that alert prey to the presence of a preda- with different behavioral options. For example, foraging ani- tor, indirect cues, such as odors, also allow the assessment of a mals must accurately measure predation risk and weigh this risk potential predatory threat. However, indirect cues are frequently against current nutritional state. Such cost-benefit analyses are more ambiguous and seldom provide information on the degree or made by the nervous system through the integration of exter- immediacy of the danger posed. And indirect cues that signal the nal sensory signals with current internal states, and these deci- presence of a predator (although no predator is currently present) sions ideally lead to behavioral choices that optimize an animal’s can divert attention from other vital activities or suppress these fitness. activities altogether. Different risk assessment behaviors, appre- Invertebrates are superbly suited to measure both the behav- hension, and vigilance, are responses to indirect predator cues ior and neural mechanisms underlying predator avoidance. In commonly described in vertebrate animals (Kavaliers and Cho- many invertebrates, an accessible nervous system with described leris, 2001). Although they are likely to exist in invertebrates, these neural escape circuits controls discrete escape behaviors. Thus, “anticipatory” predator avoidance behaviors are much less studied the link between neural machinery and behavioral expression is in invertebrates where the evolution of extremely fast and power- often identifiable and quantifiable. More recently, economic deci- ful escape reactions in response to immediate attack has arguably sion making, i.e., costs-benefit calculations under predatory risk, reduced the necessity for extensive predator scanning and risk www.frontiersin.org August 2012 | Volume 6 | Article 125 | 1 Herberholz and Marquart Decisions underlying escape assessment. Additionally, while numerous behaviors in an ani- circuits in invertebrates are frequently divided into two broad mal’s repertoire contribute to predator avoidance, most are subtle categories: those that contain “command” or “command-like” ele- and difficult to subject to neurobiological analysis. For instance, ments and those that do not (Kupfermann and Weiss, 1978, 2001; an animal’s decision when and where to forage is greatly shaped by Edwards et al., 1999; Eaton et al., 2001). In command systems, the the risk of predation (Lima and Dill, 1990). How an animal calcu- activity of the command neuron is thought to be necessary and suf- lates this predatory risk and weighs it against concurrent internal ficient for the production of a behavior. Often a single spike in this and external demands is certainly an interesting question; however, neuron is sufficient for the readout of an entire escape program. the time-scale and context of such a decision make it difficult to While highly adaptive, these rapid behaviors are highly stereo- subject to detailed electrophysiological or neuroanatomical analy- typed, showing little variability. In contrast, the escape behaviors sis. Instead, what has overwhelmingly sufficed for the study of produced by systems ostensibly lacking a command element typ- predator avoidance in neuroscience has been the analysis of much ically display a greater degree of complexity and flexibility and more discrete escape or startle behaviors. Because escape behav- are frequently made up of a sequence of independently variable iors are so critical, they must interface with and frequently override components. This flexibility affords the animal a greater degree the performance of any ongoing or planned behaviors. And while of control over the precise timing, direction, and structure of the other behaviors may have a greater evolutionary importance over escape behavior. Traditionally, however, this is assumed to come the long term, seldom are they as time-sensitive and unforgiving at an additional computational cost that adds to the latency of the as escape. Thus, it is unsurprising that the circuits tasked with the action (Bullock, 1984). Alternatively, variability may be added to sensory acquisition, computation, and action upon salient preda- behavioral decisions by sequential neural processing. For exam- tory cues are frequently the largest, most robust, and most highly ple, in the medicinal leech decision neurons can be active during stereotyped neural systems in an organism. competing behaviors (e.g., swimming and body shortening), and If a predator is around, it is critical to identify and react to stimulation of one decision neuron can produce two different predatory cues at an appropriate time and in an effective man- behavioral outputs, swimming and crawling. Hypothesized to be ner. Consequently, escape behaviors must be fast, accurate, and organized in a hierarchical order, the first neuron in the chain robust in order to be effective countermeasures against the often would drive general behavioral action, the next one would com- rapid predatory behaviors they combat. It is believed that the time- mand selection from a pool of discrete motor patterns, and the sensitive nature of these behaviors necessitates a small number of next one would initiate the most desirable behavioral choice (Esch large elements in order to both maximize conduction velocity and and Kristan, 2002). minimize synaptic delay. Thus, escape circuits commonly have “giant fibers (GFs),” frequently the largest axons in an animal’s GIANT-NEURON MEDIATED ESCAPE nerve cord, which can be readily identified by their size, location, Crayfish or morphology. These characteristics allow for rapid identifica- Crayfish are equipped with powerful escape reactions mediated by tion and often make these neurons accessible to a wide range of rapidly responding neural circuits (reviewed in Wine and Krasne, cell biological and electrophysiological studies. 1982; Krasne and Wine, 1984; Edwards et al., 1999). These circuits Because of their simplicity and clear function, these circuits control at least three distinct motor programs that propel the ani- have been excellent models for the study of the neural basis of mals in different directions, but always away from real or assumed behavior. Recent work, however, has uncovered a surprising degree threats. Circuits and their associated tail-flips can be divided into of flexibility not previously recognized in these “simple,”“reflexive” two major categories, giant and non-giant. Two circuits, the lateral systems. High-speed video recordings have exposed a previously giant (LG) and medial giant (MG) system contain giant interneu- unappreciated level of complexity to arthropod escape behav- rons as key “command” components, are made for speed, and iors that has made researchers question the structure and even require strong and phasic input for their activation. In contrast, a identity of the underlying circuits that were originally assumed poorly elucidated non-giant system is believed to control slower, to be responsible for escape (Hammond and O’Shea, 2007a,b; but more variable escape tail-flips (Edwards et al., 1999). These Card and Dickinson, 2008a,b; Fotowat et al., 2009). Addition- escape circuits have been the focus of 65 years of intensive research ally, wireless-recording techniques have been adapted to small since they were first described by Wiersma (1947, 1952) in his invertebrate models allowing, for the first time, the correlation pioneering work. of neural activity from multiple identified neurons with the time- The LG interneurons, two large fibers consisting of a series course of escape behavior in unrestrained preparations (Fotowat of gap junction-linked neurons that project from tail to head, et al., 2011; Harrison et al., 2011). And while neural-behavioral are activated by tactile and strong hydrodynamic stimulation of correlations are not uncommon, escape behavior in invertebrates sensory hairs and proprioceptors located on the abdomen. The provides possibly one of the few opportunities to simultaneously LG interneurons also receive excitatory inputs from rostral sen- record from all the critical elements in a neural circuit and relate sory organs, but these inputs alone are insufficient to fire the LG. it to what is now appreciated as an increasingly complex, but still If these inputs sum with strong caudal inputs, however, a sin- tractable, behavior. This provides quite possibly one of the best gle LG action potential (in one of the two fibers) is sufficient to current opportunities for the comprehensive analysis of the neural produce an escape motion that thrusts the animal upward and underpinnings of decision making surrounding a behavior. away from the point of caudal stimulation (Liu and Herberholz, While there is likely a broad spectrum of complexity in the cir- 2010). The motor program is activated within milliseconds after cuits embedded in even the most simple nervous system, escape stimulation and speed and accuracy is guaranteed through several Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 2 Herberholz and Marquart Decisions underlying escape structural and functional specializations within the circuit (Her- berholz et al., 2002). Once activated, the LG interneurons drive giant motor neurons via rectifying electrical synapses, which acti- vate fast flexor muscles in the last two thoracic and first three abdominal segments causing a bending of the abdomen around the thoracic-abdominal joint and thus the stereotyped “jack-knife” motion that propels the animal upward (Wine and Krasne, 1972). Latency is minimal, with 5–15 ms between stimulation and start of the behavioral response, and varies according to both internal (e.g., animal size: Edwards et al., 1994) and external conditions (e.g., water temperature: Heitler and Edwards, 1998). This short latency is accomplished by the high transmission velocity due to the diameter of the GFs and by electrical coupling among most circuit components (Figure 5A). The MG interneurons, a pair of large fibers projecting from head to tail, are activated by strong, phasic visual or tactile inputs directed to the front of the animal. The MG interneurons receive their excitatory inputs in the brain where both neurons are elec- trically coupled to each other. One action potential in one of the MGs is sufficient to drive the fast and stereotyped backward escape response. The MG interneurons connect electrically to giant motor neurons, which activate fast flexor muscles in all abdominal seg- ments, causing the bending of the entire abdomen and propelling the animal backward away from the point of stimulation. MG tail- flips in response to tactile stimulation are as fast as LG-mediated tail-flips and happen within a few milliseconds (Wine and Krasne, 1972). Visually activated MG tail-flips are slower, but are still produced as quickly as 50 ms after detection of a visual danger stimulus (Liden and Herberholz, 2008; Liden et al., 2010). Non-giant-mediated tail-flips are controlled by a circuit that lacks giant interneurons. These tail-flips are elicited by a variety FIGURE 1 | Escape success and latencies measured in juvenile crayfish of different stimuli, typically more gradual and less forceful in attacked by dragonfly nymphs. (A) Attacks evoking tail-flips mediated by presentation than those activating giant-mediated tail-flips. They the medial giant (MG) or lateral giant (LG) interneurons are equally effective are produced with longer latencies, usually up to 10-fold slower to prevent capture whereas attacks eliciting non-giant (Non-G) tail-flips are than giant-mediated tail-flips, and considered, in a way, “volun- much less effective. (B) Unsuccessful MG and Non-G, but not LG tary” because the animal “chooses” to activate certain patterns responses are frequently followed by a series of Non-G tail-flips (left bars), which substantially increase the overall rate of escape (right bars). (C) of fast flexor muscle groups. Thus, the timing and direction of Escape latencies for crayfish attacked by predators (solid bars) or stimulated non-giant tail-flips can be modulated, resulting in a much more with a handheld probe (striped bars) are similar for giant mediated (MG and variable behavior compared to the giant-mediated tail-flips (Wine LG) tail-flips, but significantly shorter for predator evoked Non-G tail-flips. and Krasne, 1982; Wine, 1984). Non-giant tail-flips are also used Modified from Herberholz et al. (2004). during “swimming,” where a series of tail flexions and extensions propels the animal backward through the water. Although our understanding of the neural underpinnings of of all cases and escaping, after being captured, using a series of tail-flip escape, especially tail-flips produced by the LG circuit, is non-giant tail-flips in more than 75% of the remaining cases extensive and essentially unmatched by that of other experimen- (Figure 1B). Interestingly, latencies for non-giant tail-flips that tal models, our knowledge of escape circuit activation in response were produced as initial response to the predator strike were much to real predatory danger is virtually non-existent. Using dragon- shorter than latencies of non-giant tail-flips elicited by tactile stim- fly nymphs as natural predators, Herberholz et al. (2004) showed ulation with a handheld probe (Figure 1C). This suggests that that all three escape circuits of juvenile crayfish were activated in crayfish prepared the non-giant escape before the strike was deliv- response to attacks (Figure 1A). Initial escape responses to preda- ered, possibly integrating visual and hydrodynamic cues from the tory strikes were primarily mediated by giant tail-flips; frontal approaching predator in anticipation of the attack. The study also attacks evoked MG tail-flips whereas attacks directed to the rear revealed that crayfish relied entirely on their fast and powerful of the crayfish elicited LG tail-flips. While few attacks elicited tail-flip escape behaviors; crayfish showed no signs of predator non-giant tail-flips initially, overall escape performance improved recognition, vigilance, or avoidance behaviors in any of the trials substantially when non-giant tail-flips were produced following (Herberholz et al., 2004). Thus, the decision to escape, at least from capture. Overall, crayfish were successful at evading dragonfly this specific predator, is based on the activation of fixed action pat- nymphs, avoiding the predator’s strike with giant tail-flips in 50% terns elicited by predatory stimuli. The decision to escape is made www.frontiersin.org August 2012 | Volume 6 | Article 125 | 3 Herberholz and Marquart Decisions underlying escape at individual decision-making neurons; if the predatory signal is that rather than a simple escape jump, the escape behavior in sufficient to activate them, escape will inevitably follow. wild-type fruit flies involves a complex sequence of events con- sisting of at least four distinct subcomponents: an initial freeze Drosophila followed by postural adjustments, wing-elevation, and finally an There are a number of similarities between the GF system in escape jump coordinated with the initial down stroke of flight ini- Drosophila and the MG system in crayfish. Like the MG system, the tiation (Figure 2C). These behaviors do not appear to merely be GF system contains GFs originating in the brain that project down a fixed action pattern as new information continues to be inte- contralaterally to primary motor neurons that control the tho- grated into and affect subsequent components of the behaviors racic musculature responsible for the fruit fly’s escape behaviors even after sequence initiation (Hammond and O’Shea, 2007b). (reviewed in Wyman et al., 1984; Allen et al., 2006). In these giant These preflight behaviors were found to influence both the trajec- fibers, a single spike is normally sufficient for the activation of an tory as well as initial flight stability of the escape behavior (Card escape jump followed by flight initiation. Despite the motor por- and Dickinson, 2008b). tion of both the MG and GF being well described, comparatively This newly appreciated complexity of the response suggests little is known about the visual and mechanosensory pathways that that this escape behavior is either not in fact mediated by the GF feed into the giant fiber systems of either animal (Figures 5A,B). system or that additional unidentified pathways must be involved While the escape behaviors produced by these circuits are that are responsible for the preflight sequence that proceeds the extremely fast due to high conductance velocities and the minimal escape jump (Card and Dickinson, 2008b). Toward this end, evi- synaptic delay from a preponderance of electrical synapses, this dence for a previously unknown escape circuit was recorded by speed has generally been thought to come at the expense of flexi- Fotowat et al. (2009). In the absence of GF activation, the activ- bility (Bullock, 1984). Thus, giant-mediated escape behaviors are ity of this novel circuit correlated with the production of escape traditionally characterized as highly stereotyped with little vari- behavior in response to looming stimuli. While this pathway is ance in timing or direction; and whatever variance the result of yet to be anatomically identified, its activity shares features similar stochastic properties of the system and not the consequence of to well-described circuits responsive to looming stimuli in both neural computation (Bullock, 1984). vertebrates and invertebrates (e.g., pigeon: Sun and Frost, 1998; Although Drosophila has been a preeminent genetic model locust: Rind and Simmons, 1992; crab: Oliva et al., 2007; bull- since the start of the twentieth century, its diminutive size lim- frog: Nakagawa and Hongjian, 2010). All of this strongly suggests ited its use in electrophysiology until the 1970s (Bellen et al., that the GF system is not necessary for the production of escape 2010). And while the GF system was identified in 1948 (Power, behavior in the fruit fly, but that the GF system, possibly akin to 1948), it was not electrophysiologically characterized and linked the escape circuits in the crayfish, may be one of many present in to the production of escape behavior until the early 1980s (Wyman Drosophila. et al., 1984). This escape behavior was initially characterized as an Being that sudden changes in luminance (light-off ) are the only abbreviated form of “voluntary” flight initiation (Trimarchi and stimulus to reliably produce GF-mediated escape behavior, and Schneiderman, 1995a). While voluntary flight initiation is pre- then only in white-eyed fruit fly mutants, what role, if any, that ceded by a series of postural adjustments that prepare the fly for the GF system plays in actual escape behavior of wild-type fruit stable, directional flight, escape flight lacks these preflight pos- flies is now unclear. Although stimuli that reliably recruit the GF tural leg, and wing movements. Instead, escape initiation consists system in wild-type flies are unknown, it seems unlikely that the almost exclusively in the extension of the fruit fly’s mesothoracic GF system is simply the vestige of a lost escape circuit. While the legs that propels the insect off of the substrate, which is only then newly identified looming sensitive pathway might be tuned to a followed by the unfolding and initiation of wing movements (Card selective set of stimulus features, the GF system could still serve as and Dickinson, 2008a). a robust, broadly tuned escape circuit capable of producing rapid As the GF system was the only identified Drosophila escape escape behavior when more selective systems fail (Fotowat et al., circuit, it was assumed to mediate the escape behavior elicited 2009). by all visual, chemical, and mechanosensory stimuli that elicit an escape jump (McKenna et al., 1989). However, a num- VISUAL INTERNEURON MEDIATED ESCAPE ber of observations have accumulated that conflicted with this Locust canonical interpretation. For instance, in the housefly GF activ- While locusts produce avoidance behavior in response to a variety ity was shown not to be necessary for the production of of noxious stimuli (Riede, 1993; Friedel, 1999), the best studied of an escape jump in response to looming stimuli (Holmqvist, these are escape jumps in response to looming stimuli (reviewed 1994). Additionally, Trimarchi and Schneiderman (1995b) pro- in Pearson and O’Shea, 1984; Burrows, 1996; Figure 3). Like the vided evidence for an olfactory-induced flight initiation rem- escape behavior of fruit flies, the locust escape jump is a com- iniscent of the fruit flies’ escape behavior that was also not plex behavior composed of a sequence of distinct components, mediated by the GFs. More recently, the simplicity of the which allow the animal to direct this jump (Santer et al., 2005b). observed escape behavior was reassessed through high-speed In preparation for a jump, tilting postural movements mediated video analysis (Hammond and O’Shea, 2007a,b; Card and Dick- by the pro- and mesothoracic legs rotate the long axis of the locust inson, 2008a,b). This work illustrated that these “simple” escape toward the direction of the eventual jump (Hassenstein and Hus- behaviors were far more complex and nuanced than originally tert, 1999; Santer et al., 2005b; Figure 3A). The actual jump is assumed (Figures 2A,B). Card and Dickinson (2008a) showed produced through the cocking of the hindlegs, storage of energy Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 4 Herberholz and Marquart Decisions underlying escape FIGURE 2 | Escape flight planning and execution in Drosophila. dots mark head and abdomen points. (B) Probability that body parts of the fly (A) High-speed video sequence shows a typical escape to a looming frontal (black, T1 and T3 legs; red, T2 legs; blue, wings; gray, body) were moving prior stimulus with a prism allowing for simultaneous observation of ventral and to takeoff (green line). (C) As stimulus intensity increases, independent motor side profiles. Time stamps are milliseconds elapsed since stimulus onset. Red programs are activated eliciting discrete escape subbehaviors prior to takeoff. dots mark the initial contact point of the second leg tarsi with substrate. White Adapted with permission from Card and Dickinson (2008b). by the co-contraction of tibia flexor and extensor muscles, and sufficient energy in the animal’s hindlegs, co-contraction must finally the release of this energy, triggered by flexor inhibition begin as soon as possible in order to allow for a timely escape. (Burrows and Morris, 2001). Given the time required to store In contrast, the adjustment of pro- and mesothoracic limbs can www.frontiersin.org August 2012 | Volume 6 | Article 125 | 5 Herberholz and Marquart Decisions underlying escape FIGURE 3 | Escape jump and DCMD activity in locusts in response to extracellularly in the nerve cord from one locust (red traces). Raster plots show looming stimuli. (A) Four high-speed video frames from a locust producing an DCMD spikes recorded in 10 repetitions of the stimulus. Black and blue traces escape jump with time to collision listed in milliseconds. The position of the show average DCMD firing rate and its standard deviation, respectively. (D) femur-tibia joint is marked in red to calculate pixel movements of the joint. Timing of joint movements, DCMD peak and takeoff obtained from seven (B) Muscle recordings from the same trial. Stimulus angular size is shown on locusts. The DCMD peak occurred after the IJM and before the FJM and takeoff top with joint movements and flexor and extensor recordings below. (IJM, initial for all l /|v | values (l /|v |D ratio of stimulus radius (l ) to the velocity (v ) of the joint movement; FJM, final joint movement.) (C) DCMD activity measured stimulus). Adapted with permission from Fotowat and Gabbiani (2007). continue throughout co-contraction, allowing for alterations of LGMDs, the DCMDs produce action potentials in response to escape trajectory up until the escape jump is triggered (Santer looming stimuli, with their firing rate increasing as the looming et al., 2005b). On the other hand, if the hindlegs were used to object gets closer. Thus, the DCMDs were originally thought to control direction, it is thought that the decision of where to jump play a major role in jump production, sometimes compared to would have to be made over 100 ms before the jump is produced. the giant fibers in crayfish and fruit flies that control their fast Not only are locusts able to direct these jumps up to 50˚ to either escape maneuvers (Burrows, 1996). However, locusts prepare for side of their long axis, but their escape circuitry allows them to con- jumps by co-contracting flexor and extensor tibiae muscles for trol the timing, distance, and elevation of these jumps (Santer et al., s100 ms before the jump is released by relaxation of the flexor 2005b; Simmons et al., 2010). Similar to Drosophila, this complex muscles. Thus, the jump is not simply triggered by suprathresh- sequence of events does not appear to be a fixed action pattern old excitation of the DCMDs, because withdrawal of excitation that once initiated must be taken to completion as the locust can and inhibition are needed during the preparatory phase of the relax this co-contraction and release the stored up energy without jump (Figure 3B). Nevertheless, the DCMDs seem to participate the production of an escape jump (Heitler and Burrows, 1977). in all phases of the jump. Fotowat and Gabbiani (2007) compared Motor areas controlling these escape jumps are innervated by electrophysiological recordings with high-speed video recordings a pair of large interneurons, the descending contralateral move- and found that the rising phase of the firing rate of the DCMDs ment detectors (DCMDs) which receive excitatory inputs from coincided with the preparatory phase of the jump, whereas the lobula giant movement detector (LGMD) neurons that are respon- peak firing rate coincided with the co-activation period of flexor sive to looming stimuli. With a one-to-one relationship with the and extensor muscles, and decay of firing rate to less than 10% Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 6 Herberholz and Marquart Decisions underlying escape coincided with takeoff. This suggests that different stages of jump Three of these LG classes receive proprioceptive inputs from the production could be controlled by distinct phases in the firing legs, and thus could potentially integrate some contextual informa- pattern of the DCMDs (Figures 3C,D). Hindleg flexion in prepa- tion during predator escape (Berón de Astrada and Tomsic, 2002). ration for the jump, however, is not dependent on DCMD activity. Oliva et al. (2007) tested the escape behavior of grapsid crabs on When the connective containing the DCMD neuron was severed, a freely rotating styrofoam ball and recorded escape movements hindleg flexion still occurred, and it could also be evoked with (i.e., running) while looming stimuli were presented. They also visual stimuli that did not cause high firing activity in the DCMDs. recorded intracellularly from the LG neurons in restrained crabs This showed that while the activity of the DCMDs may contribute and compared these recordings with the behavioral data. Escape to hindleg flexion, it was not necessary for it and, thus, other runs were initiated soon after the LG neuron increased its fir- descending pathways would seem to be involved (Santer et al., ing rate, and after maximum stimulus expansion, the LG neurons 2008). Using a telemetry system to record DCMD and motor neu- stopped firing, coinciding with run deceleration in freely behaving ron activity in freely behaving locusts, it was found that the number animals. Moreover, the spike frequency of the LG neurons reflected of recorded DCMD spikes predicted motor neuron activity and the timing and speed of the escape response (Figures 4A,B). Inter- jump occurrence, and the time of peak firing rate predicted time estingly, the activity of the LG neurons is strongly affected by of takeoff (Fotowat et al., 2011). Although this underlined the role season with responses weaker in winter when predation risk is of the DCMDs as neurons exhibiting discrete firing responses to typically low and the animals are less active (Sztarker and Tomsic, looming stimuli, which in turn affected discrete stages of escape 2008). motor output, jump production remained intact, and occurred The relation between LG neuron activity and escape behavior at the same time as in control animals following DCMD abla- was also nicely demonstrated in experiments that tested short- tion. Thus, another neuron for jump production must exist, and term and long-term visual memory in crabs. Tomsic et al. (2003) this may be the descending ipsilateral movement detector neu- showed that LG neurons changed their responses to a visual threat ron (DIMD), which responds to looming targets, similarly to the (displacement of a black screen above the animal) in correspon- DCMD (Fotowat et al., 2011). Additionally, another descending dence with the behavioral changes observed in unrestrained ani- interneuron that responds to looming stimuli has recently been mals. Modification of LG neuron activity occurred during learning described. Thus visually mediated escape behavior in locusts is and persisted, after spaced training, for 24 h. However, while the likely controlled by at least three different descending neurons memory of freely behaving crabs reflects a strong stimulus-context (Gray et al., 2010). How these neurons interact to produce the association, LG neurons generalize the learned stimulus into new escape behavior remains to be determined (Figure 5C). spatial locations. Thus, despite being able to clearly distinguish the Locusts also produce an avoidance behavior during flight. learned stimulus from other similar stimuli (i.e., stimulus mem- When looming stimuli are presented, flying locusts produce a ory), the LG neurons do not appear to be involved in processing gliding dive similar to the dives used by other insects to evade contextual visual information (i.e., where the stimulus was learned; aerial predators. After DCMD neurons are activated by a loom- Sztarker and Tomsic, 2011). In summary, the LG neurons are ing stimulus, they produce short-latency excitatory postsynaptic sensory neurons located in the eyestalk, and their neural activity potentials (EPSPs) in a motor neuron that raises the wing into the patterns closely match escape behavior produced in unrestrained gliding posture. Stimuli that evoked high-frequency firing in the crabs (Medan et al., 2007). Their exact role in producing the DCMDs also reliably elicited the gliding response, and the behavior behavior, however, is unknown. To answer this question, detailed was less frequently observed when high-frequency DCMD spikes investigation of the descending pathways that connect the LG neu- were absent (Santer et al., 2005a). However, similar to the escape rons to the motor centers that control escape runs will be required jump, DCMD activity was not always sufficient to evoke gliding. (Figure 5D). Most likely, its high-frequency activity must be precisely timed with wingbeat phase because glides can only be produced during VALUE-BASED DECISION MAKING wing elevation. In addition, other neurons that are implicated in Adaptive behavioral decisions are essential for the survival and jump production (e.g., the DIMDs) may also be involved in escape reproductive success of most animals, including humans. Animals gliding in flying locusts (Santer et al., 2006). can typically choose from several behavioral alternatives, which need to be evaluated before the most desirable option is selected. To Crabs determine what behavior is most desirable at any given point, the The role of identified neurons in visually mediated escape behav- nervous system must integrate external conditions (e.g., predation ior has also recently been studied in grapsid crabs (reviewed risk) with current internal drives (e.g., hunger state), thus trading in Hemmi and Tomsic, 2012). The firing rate of these motion- off the costs and benefits of different alternatives before deciding sensitive lobula giant (LG) neurons in response to looming stim- which one to choose. For example, a hungry animal is more likely uli corresponds with the intensity of the crab’s escape behavior. to choose a behavioral option that involves risks because the value Four distinct classes of these neurons have been anatomically placed on foraging is greater than the value placed on other alter- and physiologically described. All four classes show wide-field natives such as hiding. If the benefit of finding a meal outweighs tangential arborization in the lobula, somata located beneath, the estimated cost of being attacked by a predator, the decision is and axons that project toward the midbrain; however, they to forage. If the value placed on foraging is low because the ani- are uniquely identifiable due to differences in morphology and mal is satiated, other behavioral options become more valuable response preferences (Medan et al., 2007). and behavioral output will shift toward less risky activities. The www.frontiersin.org August 2012 | Volume 6 | Article 125 | 7 Herberholz and Marquart Decisions underlying escape FIGURE 4 | Response of a crab’s LG neuron to looming stimuli and size of the looming object is shown in bottom trace. (B) Mean spike correlation with escape run. (A) Intracellular trace from one LG rate from a single LG neuron (top) and mean escape running speed neuron in response to a looming stimulus. Raster plot shows (bottom). Arrowheads mark the start of stimulus expansion and long responses from one neuron to nine repetitions of the stimulus. arrows mark increase in spike rate above resting level. Adapted with Histogram shows mean spike rate obtained from all nine trials. Angular permission from Oliva et al. (2007). literature on value-based decision making, especially with a focus review some recent experiments on value-based decision mak- on prey behavior in predator-prey interactions, is extensive and ing in response to predatory threat, and provide two examples covers a wide range of organisms (e.g., Ydenberg and Dill, 1986; where economic decisions can be linked to identifiable neural Lima and Dill, 1990). circuitry. The relatively new field of “neuroeconomics” is concerned with the neural underpinnings of value-based decision making CRAYFISH in humans and other non-human primates (Schall, 2001; Rangel When juvenile crayfish are exposed to fast-moving shadows while et al., 2008) and there is now fast growing interest in under- foraging in an artificial stream environment, they respond by standing the neural mechanisms that govern cost-benefit calcu- choosing one of two behavioral actions: they either freeze in lations. An increasing number of studies performed in humans place and remain motionless for several seconds before resum- and other primates are combining non-invasive techniques such ing foraging or they produce a tail-flip mediated by the MG as functional magnetic resonance imaging or cortical recordings neuron that propels the animal backward and away from the with discrimination tasks or cognitive experiments (Glimcher and approaching shadow and the expected food source (Liden and Rustichini, 2004; Huettel et al., 2005; Sugrue et al., 2005). The Herberholz, 2008; Figure 6A). Thus, crayfish respond to visual complexity of the mammalian brain, however, presents many threat signals that simulate the imminent attack of a preda- challenges. It is difficult to directly correlate neuronal activity tor with defensive behaviors that are discrete and incompatible. and behavioral expression and to obtain detailed information When Liden and Herberholz (2008) exposed groups of juvenile on neural circuit organization, cellular mechanisms, and the crayfish to different shadow velocities, they found that the frequen- interplay between sensory and motor systems. Decision-making cies of the two behavioral responses were dependent on shadow circuitry has been studied quite extensively in various inverte- speed. Slower moving shadows evoked more tail-flips than freez- brates, but descriptions of neural mechanisms underlying value- ing, but as shadow speed increased the frequency of tail-flips based (economic) behavioral decisions are rare (Kristan and decreased and crayfish primarily produced freezing behavior. The Gillette, 2007; Kristan, 2008). This is surprising because behav- study also showed that different individuals choose different anti- ioral experiments have shown that invertebrates make decisions predator strategies when exposed to one type of shadow. Some that are not always simple and reflexive, but are often the prod- animals decided to freeze in response to the danger signal while uct of careful cost-benefit calculations (Ydenberg and Dill, 1986; others decided to tail-flip. This suggests that different crayfish Lima and Dill, 1990; Chittka et al., 2009). Thus, invertebrates have different thresholds for each behavioral action, but what are ideally suited to study the neural mechanisms underlying underlies this difference remains to be determined. Because all value-based decision making. In the following section, we will tested animals were of identical size and shared the same social Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 8 Herberholz and Marquart Decisions underlying escape FIGURE 5 | Circuitry for arthropod escape behavior. Neural circuits sensory interneurons, projection (ascending or descending) or command underlying escape behaviors for crayfish (A), Drosophila (B), locust (C), and neurons, premotor neurons, and motor neurons with associated sensory crab (D) are illustrated. Circuits are divided into five levels: sensory neurons, (Continued) www.frontiersin.org August 2012 | Volume 6 | Article 125 | 9 Herberholz and Marquart Decisions underlying escape FIGURE 5 | Continued result of at least two circuits; a giant fiber (GF) system mediating jumps stimuli on the left and motor output on the right. Solid circles and lines lacking preparatory leg and wing movements and a yet to be identified represent identified neurons and connections while dashed circles and lines escape circuit that produces escape jumps with preparatory preflight limb represent neurons and connections yet to be identified. Stacked circles and wing adjustments. (PSI, peripherally synapsing interneuron, DLMns, represent a population of neurons. Lines end in four ways: with a dorsal lateral motor neurons, TTMn, tergotrochanteral muscle neuron.) perpendicular line, a concave cup, a circle, or dashes. Perpendicular lines (C) Locusts possess at least two escape circuits as well, one responsive to represent electrical synapses. Concave cups represent electrical synapses. looming stimuli and another responsive to auditory and mechanosensory Circles represent inhibitory synapses. Dashes indicate an unknown synapse stimuli. While numerous neurons that are believed to play a role in these type. Generic abbreviations: MSns, mechanosensory neurons; MSis, behaviors have been identified, both circuits remain incomplete. [LGMD, mechanosensory interneurons; VSns, visual sensory neurons; VSis, visual lobula giant movement detector neuron; LGMD2, lobula giant movement sensory interneurons; OSns, olfactory sensory neurons; OSis, olfactory detector neuron 2, DCMD, descending contralateral movement detector sensory neurons; ASns, auditory sensory neurons; ASis, auditory sensory neuron; DIMD, descending ipsilateral movement detector neuron; LDCMD, interneurons. (A) Crayfish tail-flips are controlled by one of three circuits, the late descending contralateral movement detector neuron, C, C (“cocking”) lateral giant (LG), medial giant (MG), and non-giant escape circuit. While the neuron, M, M-neuron, FETi, fast extensor tibia motor neuron, FLTis, flexor LG system is almost fully elucidated and the abdominal motor outputs of the tibia motor neurons, 714, neuron 714.] (D) In crabs, a class of visual MG are also well described, very little beyond the fast flexor motor neurons interneurons, the lobula giants (LGs), have been identified that are thought to (FFMns) are known to play a part in non-giant tail-flips. SG, segmental giant play a role in the crab’s escape behavior; however, no other elements in this neuron, MoG, motor giant neuron. (B) Drosophila escape jumps are the escape circuit have been elucidated. experiences and feeding history, other intrinsic factors must be SEA SLUG responsible. The marine snail has been a fruitful model for studying the Recently, Liden et al. (2010) used the same experimental design neural mechanisms underlying decision making and behavioral to show that crayfish base their escape decisions on the values choice. Using a “competing behaviors” paradigm, early work sug- of each behavioral option. They measured escape latencies for gested that different incompatible behaviors were organized in shadow-induced MG-mediated tail-flips by comparing photodi- a hierarchical model, each controlled by command-like neu- ode signals with bath electrode recordings that non-invasively rons that produced one behavior while inhibiting others. For captured neural and muscular activity produced during tail-flips example, when the sea slug was feeding, avoidance withdrawal (Figure 6B). They found that very fast approaching shadows in response to a tactile stimulus was suppressed (Kovac and become inescapable because they collided with the animal before a Davis, 1977). This suppression is caused by identified interneu- tail-flip could be generated. Moreover, tail-flips are costly because rons that are part of the motor circuit that generates feed- they move the animal away from the expected food source. Thus, ing. Thus, feeding behavior takes precedence over withdrawal, the observed suppression of tail-flipping in favor of freezing in while escape swimming dominates most other behaviors, includ- animals facing inescapable shadows, where the value of a tail- ing feeding (Jing and Gillette, 1995). The A1 neurons, a bilat- flip would be low, reflects the output product of an “economic” eral pair of interneurons located in the cerebropleural gan- decision-making process. Although tail-flipping is considered a glion of the snail, are necessary elements of the escape swim- less risky strategy when experiencing a predator attack, crayfish ming behavior, and their activity also inhibits feeding behav- also defaulted to freezing behavior when the expected reward ior. became more valuable. When food odor concentration in the arti- Recent work, however, has shown that sea slugs base their ficial stream was increased 10-fold, shadows that evoked mostly decisions on cost-benefit computations (Gillette et al., 2000; tail-flips under standard conditions now generated mainly freez- Figure 7). When presented with food stimuli, feeding behav- ing behavior. Interestingly, if high food value was paired with a ior or avoidance behavior can be activated, depending on the strong predator signal (a slow moving shadow) that reliably evoked concentration of the food stimulus and the current behavioral tail-flips under regular conditions, the behavioral shift toward state of the animal. At low concentrations and in satiated ani- freezing was less pronounced. Thus, a strong predator signal was mals, food stimuli typically evoked avoidance behavior. When the able to override the exaggerated food incentive (Figure 6C). This threshold for feeding was exceeded, avoidance behavior was sup- illustrates that crayfish calculate the costs and benefits of differ- pressed, and in hungry snails, even nociceptive stimuli elicited ent behavioral options and they carefully weigh predation risk feeding behavior (Figure 7A). This suggests both appetitive and against expected reward, eventually selecting the most valuable noxious stimuli provide inputs to neural networks underlying behavioral choice (Liden et al., 2010). Because these observed tail- feeding and avoidance behavior, but the final behavioral deci- flips are always generated by activation of MG neurons and the sion is determined by hunger state. Thus, in partially or fully MG circuit is accessible for neurophysiological and neurochemical satiated animals, the value placed on feeding behavior is low experiments, the neural workings underlying value-based deci- while it is high for avoidance behavior that protects the animal sion making in crayfish can now be investigated on the cellular from predators. Using a simple cost-benefit analysis, the animal level. This establishes the crayfish as an important new model weighs nutritional needs against predator risk and selects the for studying the neuroeconomics underlying predator avoidance. most desirable choice (Gillette et al., 2000; Figure 7B). Impor- However, to understand the decision-making process on the net- tantly, feeding and avoidance can be observed as fictive motor work level, identification of interneurons that form the descending patterns in isolated central nervous systems of the snail and some visual pathway for freezing behavior will be required. of the neurons controlling these behaviors have been individually Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 10 Herberholz and Marquart Decisions underlying escape FIGURE 6 | Escape choices and neural activation in crayfish exposed to collided with the animal and produced the peak response in PD no. 2. The first approaching shadows. (A) Experimental diagram and four video frames small deflection (arrow) in the BE trace is due to MG neuron activation, while illustrating a crayfish foraging (first two panels) and then tail-flipping (last two the large phasic potential and the smaller more erratic potentials that follow panels) in response to a fast approaching shadow with time in seconds. (B) are due to muscular activity during tail-flips. (C) Left: when exposed to a Left: example recordings from photodiodes positioned on the tank walls (PD medium speed shadow (2 m/s), crayfish produce fewer tail-flips (black bars) no. 1 and PD no. 2) when a shadow passes by, and from bath electrodes (BE) and more freezing (gray bars) when food odor concentration flowing through located inside the tank that capture field potentials generated during a tail-flip. the tank is high. Right: when exposed to slower (1 m/s) shadows, the effect of Right: Traces from PD no. 2 and BE at higher temporal resolution. In this food odor concentration on behavioral choice is less pronounced. (A) Modified example, animal initiated a tail-flip response (arrow) 4 ms before the shadow from Liden and Herberholz (2008). (B,C) Modified from Liden et al. (2010). identified (Jing and Gillette, 2003). Moreover, in isolated central electrically stimulated, avoidance turns were converted to orient- nervous systems, spontaneous feeding network activity reflects ing turns (Hirayama and Gillette, 2012). Thus, the neurophysi- feeding thresholds of the nervous system donors (for proboscis ological and neurochemical mechanisms underlying cost-benefit extension and biting); while orienting turns were more frequent calculations can now be investigated in the isolated nervous sys- in low-feeding threshold donors, avoidance turns dominated tem of this animal. This is expected to substantially contribute in high-feeding threshold donors. When a “command” neuron to our cellular understanding of value-based decision-making in the feeding network of a high-feeding threshold donor was processes. www.frontiersin.org August 2012 | Volume 6 | Article 125 | 11 Herberholz and Marquart Decisions underlying escape of behavioral events and the corresponding underlying neural mechanisms (Harley et al., 2009; Harley and Ritzmann, 2010). Based on the high-speed video analysis of the behavior of fruit flies and locusts, a reexamination of the “simple” escape behav- ior of other arthropods is warranted. Perhaps an analysis at a temporal resolution comparable to that of the speed of produc- tion of these behaviors will uncover a degree of flexibility and control not previously appreciated in these animals as well. For example, while the escape tail-flip and freezing behavior of the crayfish in response to visual stimuli have been assumed to be two distinct behaviors, which has been supported by video analy- sis at 250 fps (Liden et al., 2010), possibly higher speed analysis will show that these distinct decisions are in fact part of a single escape sequence. Such an observation could provide direction in the search for the neural circuit(s) responsible for freezing, the identification of which would provide a unique opportunity to explore decision making between two circuits underlying known behavioral alternatives. While this new appreciation for the complexity of arthropod escape behavior has reinvigorated work on giant fibers and escape behavior, it raises two significant issues. First, if the giant fiber systems previously assumed to underlie observed escape behav- iors are not in fact necessary or sufficient for the production of these behaviors, what circuits are? While Fotowat et al. (2009) have made initial progress toward characterizing the activity of part of an additional putative escape circuit, the neurons will have to be anatomically identified and the circuit fleshed out in future work. Second, if the giant fibers are not involved in escape behav- iors produced under existing experimental contexts, what contexts elicit their recruitment? It would be exceedingly wasteful for the largest axons in the fruit fly’s nerve cord to go unused. There must be some combination of internal states and external stimulus con- ditions that lead to GF-mediated escape response and work should be directed toward identifying these constraints. It is likely other arthropod models will have a similar redun- dancy in escape circuitry as has been described in the crayfish. Thus, a comprehensive understanding of decision making during predator avoidance will have to wait until all pathways and not just parts of some are fully characterized (Figure 5). While the identification of all escape circuits in any one arthropod is non- trivial, that parts of both command and non-command systems have been successfully identified in various arthropods is evidence FIGURE 7 | Effects of internal state on behavioral choice in a sea slug. of the feasibility of such a research program. For example, the LG (A) Four video frames showing feeding behavior in Pleurobranchaea neurons in grapsid crab are fully characterized and individually californica. Betaine application induces an orienting turn (panel 2) followed by proboscis extension and biting (panel 3). Chemosensory structures identifiable cells that can be accessed for intracellular recordings (panel 4): rhinophore (Rh), oral veil (OV), tentacle (Tn), and proboscis (Prob). in live animals. The activity of these neurons is highly correlated (B) Partial satiation raised the threshold for proboscis extension and biting with behavioral output, which suggests that they play a major role (i.e., feeding), and increased the frequency of withdrawal and turns (i.e., in mediating escape decisions. However, relevant analysis of the avoidance) in response to betaine. Modified from Gillette et al. (2000). complete escape circuit is still missing and descending pathways that orchestrate motor actions need to be identified. As such, future work should focus on completing the picture CONCLUSION AND FUTURE DIRECTIONS Recent work in the arthropods discussed suggests that the escape of currently known circuits, where often substantial sensory or motor elements remain poorly characterized, as well as identify- behavior of all may be more complex and varied than has generally been assumed. Quantitative ethograms that divide complex escape ing unknown but hinted at command or non-command circuits. This hunt for currently uncharacterized circuits might be aided by maneuvers into a sequence of simpler events can help identify variability within each system. Moreover, combining ethograms the possible similarity to and knowledge of already characterized systems found in related species (Figure 5). For instance, the with measures of neural structure or neural activity can elu- cidate the link between discrete motor actions within a series poorly studied non-giant tail-flip circuit in crayfish might share Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 12 Herberholz and Marquart Decisions underlying escape characteristics with that of the DCMD circuit in locusts and direct measures of nervous system function in natural settings is knowledge of the structure and function of the DCMD circuit highly desirable. could aid in the identification and characterization of this escape Finally, the neuromodulation of escape behavior by system. monoamines such as octopamine, serotonin and dopamine is Due to the assumption that giant fiber systems were a singular worth further exploration. Although a number of the escape cir- system responsible for the production of all escape behaviors, there cuits discussed have been shown to be responsive to the application is currently much confusion as to what discrete escape behavior is or removal of monoamines (Glanzman and Krasne, 1983; Busta- subserved by what specific circuit. Since it now appears that there mante and Krasne, 1991; Stern et al., 1995; Pflüger et al., 2004; are likely many circuits that produce a range of escape behav- Harvey et al., 2008; Rind et al., 2008), little is known about the iors, the spectrum of these behaviors and the stimulus conditions context in which these monoamines affect the performance of that lead to their display will need to be carefully cataloged and behavioral decisions. Since most invertebrate aminergic effects behavioral assays developed that can differentiate them. However, are mediated by metabotropic receptors that can have a grad- without the ability to simultaneously record both escape behaviors ual but pronounced impact on behavior, monoamines are an and neural activity, it will be difficult to ascribe a discrete escape attractive candidate for how a nervous system may be biased sequence or subcomponent of escape behavior to a particular cir- toward the production of one behavior over another (Crisp and cuit or set of neurons. For this, the use of telemetry that allows for Mesce, 2006; Mesce and Pierce-Shimomura, 2010). Through these in vivo recordings in freely behaving animals (Fotowat et al., 2011; monoamines, escape behaviors might modulate or be modulated Harrison et al., 2011) will have to be expanded to other inverte- by competing behaviors. Monoamines (e.g., dopamine and sero- brates. While it will be some time before these techniques can be tonin) have been targeted for roles in decision making and the adapted to all models, some should be able to benefit immediately. encoding of punishment and reward (Daw et al., 2002). Thus, the Arguably, these techniques might have the most to offer in models study of monoamines in the context of the evolutionarily criti- like the crayfish where large parts of a number of well-described cal task of predator avoidance provides an excellent opportunity escape circuits have long been worked out (Figure 5A). In such to explore the postulated neurochemical currency of neuroeco- a model, not only can the function of identified neurons be cor- nomic decision making. Unfortunately, little work on value-based related to the performance of distinct components of a complex decision making has been undertaken with invertebrates despite behavioral sequence, but also how an animal chooses between a the description of numerous value-based decisions that are likely range of escape behaviors might be elucidated. Recordings with to involve identified circuits including those mediating escape implanted electrodes or bath electrodes, which non-invasively or avoidance behavior. Research in this field is currently lim- record neural and muscular field potentials in freely behaving ited to a few invertebrate species, namely the previously discussed animals, have begun to reveal some of the basic neural patterns sea slug and crayfish, where basic neural mechanisms underly- underlying escape decisions in crayfish (Herberholz et al., 2001, ing cost-benefit computations have been partially uncovered. It is 2004; Liden and Herberholz, 2008; Liden et al., 2010). surprising that researchers interested in neuroeconomics have not There is a notable lack of neuroethological studies focused on taken greater advantage of these highly tractable models, as they escape mechanisms produced under natural conditions. While are likely to contribute much to this new field, as they have con- staged encounters with natural predators in the laboratory pro- tributed to neuroscience in general (Clarac and Pearlstein, 2007). vide some insight into the interplay between neural function and Possibly we have just begun to realize that invertebrate models are ecologically relevant escape behavior, these studies are sparse. Field ideally suited to answer some of the most challenging questions studies on the other hand are often focused on ecology and behav- faced today by neuroscience research. ior and not designed to investigate neural processes. Occasionally, data sets obtained separately in the field and laboratory allow for a ACKNOWLEDGMENT Part of the work reviewed in this article was supported by a comparative view and for correlating firing patterns of individual neurons and natural escape behavior (e.g., Hemmi and Tomsic, grant from the National Science Foundation to Jens Herberholz (IOS-0919845). 2012); however, the development of new technologies that permit REFERENCES neurons in a crab (Decapoda: in the locust. J. Exp. Biol. 204, Chittka, L., Skorupski, P., and Raine, N. Allen, M. J., Godenschwege, T. A., Brachyura). J. Comp. Physiol. A 188, 3471–3481. E. (2009). Speed-accuracy tradeoffs Tanouye, M. A., and Phelan, P. 539–551. Bustamante, J., and Krasne, F. B. (1991). in animal decision making. Trends (2006). Making an escape: develop- Bullock, T. H. (1984). “Compara- Effects of octopamine on transmis- Ecol. Evol. (Amst.) 24, 400–407. ment and function of the Drosophila tive neuroethology of startle, rapid sion at the first synapse of the cray- Clarac, F., and Pearlstein, E. (2007). giant fibre system. Semin. Cell Dev. escape, and giant fiber-mediated fish lateral giant escape reaction Invertebrate preparations and their Biol. 17, 31–41. responses,” in Neural Mechanisms pathway. J. Comp. Physiol. A 169, contribution to neurobiology in the Bellen, H. J., Tong, C., and Tsuda, of Startle Behavior, ed. R. C. 369–377. second half of the 20th century. H. (2010). 100 years of Drosophila Eaton (New York: Plenum Press), Card, G., and Dickinson, M. H. (2008a). Brain Res. Rev. 54, 113–161. research and its impact on verte- 1–13. Performance trade-offs in the flight Crisp, K. M., and Mesce, K. A. brate neuroscience: a history lesson Burrows, M. (1996). The Neurobiology initiation of Drosophila. J. Exp. Biol. (2006). Beyond the central pat- for the future. Nat. Rev. Neurosci. 11, of an Insect Brain. Oxford: Oxford 211, 341–353. tern generator: amine modulation 514–522. University Press. Card, G., and Dickinson, M. H. (2008b). of decision-making neural pathways Berón de Astrada, M., and Tomsic, D. Burrows, M., and Morris, G. (2001). Visually mediated motor planning in descending from the brain of the (2002). Physiology and morphol- The kinematics and neural control the escape response of Drosophila. medicinal leech. J. Exp. Biol. 209, ogy of visual movement detector of high-speed kicking movements Curr. Biol. 18, 1300–1307. 1746–1756. www.frontiersin.org August 2012 | Volume 6 | Article 125 | 13 Herberholz and Marquart Decisions underlying escape Daw, N. D., Kakade, S., and Dayan, Hammond, S., and O’Shea, M. (2007a). Hirayama, K., and Gillette, R. (2012). and prospectus. Can. J. Zool. 68, P. (2002). Opponent interactions Escape flight initiation in the fly. J. A neuronal network switch for 619–640. between serotonin and dopamine. Comp. Physiol. A 193, 471–476. approach/avoidance toggled by Liu, Y. C., and Herberholz, J. (2010). Neural Netw. 15, 603–616. Hammond, S., and O’Shea, M. (2007b). appetitive state. Curr. Biol. 22, Sensory activation and receptive Eaton, R. C., Lee, R. K., and Foreman, Ontogeny of flight initiation in the 118–123. field organization of the lateral giant M. B. (2001). The Mauthner cell fly Drosophila melanogaster : impli- Holmqvist, M. H. (1994). A visually escape neurons in crayfish. J. Neuro- and other identified neurons of the cations for the giant fibre sys- elicited escape response in the fly physiol. 104, 675–684. brainstem escape network of fish. tem. J. Comp. Physiol. A Neuroethol. that does not use the giant fiber path- McKenna, M., Monte, P., Helfand, S. L., Prog. Neurobiol. 63, 467–485. Sens. Neural. Behav. Physiol. 193, way. Vis. Neurosci. 11, 1149–1161. Woodard, C., and Carlson, J. (1989). Edwards, D. H., and Heitler, W. J., 1125–1137. Huettel, S. A., Song, A. W., and A simple chemosensory response and Krasne, F. B. (1999). Fifty years Harley, C. M., English, B. A., and Ritz- McCarthy, G. (2005). Decisions in Drosophila and the isolation of of a command neuron: the neu- mann, R. E. (2009). Characterization under uncertainty: probabilistic acj mutants in which it is affected. robiology of escape behavior in of obstacle negotiation behaviors in context influences activation of Proc. Natl. Acad. Sci. U.S.A. 86, the crayfish. Trends Neurosci. 22, the cockroach, Blaberus discoidalis. J. prefrontal and parietal cortices. J. 8118–8122. 153–161. Exp. Biol. 212, 1463–1476. Neurosci. 25, 3304–3311. Medan, V., Oliva, D., and Tomsic, D. Edwards, D. H., Yeh, S. R., Barnett, Harley, C. M., and Ritzmann, R. E. Jing, J., and Gillette, R. (1995). Neuronal (2007). Characterization of lobula L. D., and Nagappan, P. R. (1994). (2010). Electrolytic lesions within elements that mediate escape swim- giant neurons responsive to visual Changes in synaptic integration dur- central complex neuropils of the ming and suppress feeding behav- stimuli that elicit escape behaviors ing the growth of the lateral giant cockroach brain affect negotiation of ior in the predatory sea slug Pleu- in the crab Chasmagnathus. J. Neu- neuron of crayfish. J. Neurophysiol. barriers. J. Exp. Biol. 213, 2851–2864. robranchaea. J. Neurophysiol. 74, rophysiol. 98, 2414–2428. 72, 899–908. Harrison, R. R., Fotowat, H., Chan, R., 1900–1910. Mesce, K. A., and Pierce-Shimomura, Esch, T., and Kristan, W. B. (2002). Kier, R. J., Olberg, R., Leonardo, Jing, J., and Gillette, R. (2003). Direc- J. T. (2010). Shared strategies for Decision-making in the leech ner- A., and Gabbiani, F. (2011). Wire- tional avoidance turns encoded by behavioral switching: understanding vous system. Intgr. Comp. Biol. 42, less neural/EMG telemetry systems single interneurons and sustained by how locomotor patterns are turned 716–724. for small freely moving animals. multifunctional serotonergic cells J. on and off. Front. Behav. Neurosci. Fotowat, H., Fayyazuddin, A., Bellen, IEEE Trans. Biomed. Circuits Syst. 5, Neurosci. 23, 3039–3051. 4:49. doi:10.3389/fnbeh.2010.00049 J. H., and Gabbiani, F. (2009). A 103–111. Kavaliers, M., and Choleris, E. (2001). Nakagawa, H., and Hongjian, K. (2010). novel neuronal pathway for visu- Harvey, J., Brunger, H., Middleton, C. Antipredator responses and defen- Collision-sensitive neurons in the ally guided escape in Drosophila A., Hill, J. A., Sevdali, M., Sweeney, sive behavior: ecological and etho- optic tectum of the bullfrog, Rana melanogaster. J. Neurophysiol. 102, S. T., Sparrow, J. C., and Elliot, C. logical approaches for the neuro- catesbeiana. J. Neurophysiol. 104, 875–885. J. H. (2008). Neuromuscular control sciences. Neurosci. Biobehav. Rev. 25, 2487–2499. Fotowat, H., and Gabbiani, F. (2007). of a single twitch muscle in wild type 577–586. Oliva, D., Medan, V., and Tomsic, D. Relationship between the phases of and mutant Drosophila, measured Kovac, M. P., and Davis, W. J. (1977). (2007). Escape behavior and neu- sensory and motor activity dur- with an ergometer. Invert. Neurosci. Behavioral choice: neural mecha- ronal responses to looming stimuli ing a looming-evoked multistage 8, 63–70. nisms in Pleurobranchaea. Science in the crab Chasmagnathus granu- escape behavior. J. Neurosci. 27, Hassenstein, B., and Hustert, R. (1999). 198, 632–634. latus (Decapoda: Grapsidae). J. Exp. 10047–10059. Hiding responses of locusts to Krasne, F. B., and Wine, J. J. (1984). Biol. 210, 865–880. Fotowat, H., Harrison, R. R., and Gab- approaching objects. J. Exp. Biol. “The production of crayfish tailflip Pearson, K. G., and O’Shea, M. (1984). biani, F. (2011). Multiplexing of 202, 1701–1710. escape responses,” in Neural Mecha- “Escape behavior of the locust. The motor information in the discharge Heitler, W. J., and Burrows, M. nisms of Startle Behavior, ed. R. C. jump and its initiation by visual of a collision detecting neuron dur- (1977). The locust jump. I. The Eaton (New York: Plenum Press), stimuli,” in Neural Mechanisms of ing escape behaviors. Neuron 69, motor programme. J. Exp. Biol. 66, 179–211. Startle Behavior, ed. R. C. Eaton 147–158. 203–219. Kristan, W. B. (2008). Neuronal (New York: Plenum Press), 163–178. Friedel, T. (1999). The vibrational star- Heitler, W. J., and Edwards, D. H. (1998). decision-making circuits. Curr. Biol. Pflüger, H. J., Duch, C., and Hei- tle response of the desert locust Effect of temperature on a voltage- 18, 928–932. del, E. (2004). Neuromodulatory Schistocerca gregaria. J. Exp. Biol. sensitive electrical synapse in cray- Kristan, W. B., and Gillette, R. (2007). octopaminergic neurons and their 202, 2151–2159. fish. J. Exp. Biol. 201, 503–513. “Behavioral choice,” in Invertebrate functions during insect motor Gillette, R., Huang, R. C., Hatcher, Hemmi, J. M., and Tomsic, D. (2012). Neurobiology, eds G. North and R. J. behaviour. The Ernst Florey mem- N., and Moroz, L. L. (2000). Cost- The neuroethology of escape in Greenspan (New York: Cold Spring ory lecture. Acta Biol. Hung. 55, benefit analysis potential in feed- crabs: from sensory ecology to neu- Harbor Laboratory Press), 533–553. 3–12. ing behavior of a predatory snail rons and back. Curr. Opin. Neuro- Kupfermann, I., and Weiss, K. R. Power, M. E. (1948). The thoracico- by integration of hunger, taste, and biol. 22, 194–200. (1978). The command neuron con- abdominal nervous system of pain. Proc. Natl. Acad. Sci. U.S.A. 97, Herberholz, J., Antonsen, B. L., and cept. Behav. Brain Sci. 1, 3–10. an adult insect, Drosophila 3585–3590. Edwards, D. H. (2002). A lateral Kupfermann, I., and Weiss, K. R. (2001). melanogaster. J. Comp. Neurol. Glanzman, D. L., and Krasne, F. B. excitatory network in the escape Motor program selection in simple 88, 347–409. (1983). Serotonin and octopamine circuit of crayfish. J. Neurosci. 22, model systems. Curr. Opin. Neuro- Rangel, A., Camerer, C., and Montague, have opposite modulatory effects on 9078–9085. biol. 11, 673–677. P. R. (2008). A framework for study- the crayfish’s lateral giant escape Herberholz, J., Issa, F. A., and Edwards, Liden, W. H., and Herberholz, J. (2008). ing the neurobiology of value-based reaction. J. Neurosci. 3, 2263–2269. D. H. (2001). Patterns of neural cir- Behavioral and neural responses of decision making. Nat. Rev. Neurosci. Glimcher, P. W., and Rustichini, A. cuit activation and behavior during juvenile crayfish to moving shadows. 9, 545–556. (2004). Neuroeconomics: the con- dominance hierarchy formation in J. Exp. Biol. 211, 1355–1361. Riede, K. (1993). Prepulse inhibi- silience of brain and decision. Sci- freely behaving crayfish. J. Neurosci. Liden, W. H., Phillips, M. L., and Her- tion of the startle reaction in the ence 306, 447–452. 21, 2759–2767. berholz, J. (2010). Neural control of locust Locusta migratoria (Insecta: Gray, J. R., Blincow, E., and Robertson, Herberholz, J., Sen, M. M., and Edwards, behavioural choice in juvenile cray- Orthoptera: Acridoidea). J. Comp. R. M. (2010). A pair of motion- D. H. (2004). Escape behavior and fish. Proc. Biol. Sci. 277, 3493–3500. Physiol. A 172, 351–358. sensitive neurons in the locust escape circuit activation in juvenile Lima, S. L., and Dill, L. M. (1990). Rind, F. C., Santer, R. D., and Wright, encode approaches of a looming crayfish during prey-predator inter- Behavioral decisions made under G. A. (2008). Arousal facilitates object. J. Comp. Physiol. A 927–938. actions. J. Exp. Biol. 207, 1855–1863. the risk of predation: a review collision avoidance mediated by a Frontiers in Neuroscience | Decision Neuroscience August 2012 | Volume 6 | Article 125 | 14 Herberholz and Marquart Decisions underlying escape looming sensitive visual neuron in preparation: the neuroethology of short- and long-term visual mem- Neural Mechanisms of Startle Behav- a flying locust. J. Neurophysiol. 100, visual startle in locusts. J. Insect ory in an arthropod. J. Neurosci. 23, ior, ed. R. C. Eaton (New York: 670–680. Physiol. 56, 876–883. 8539–8546. Plenum Press), 133–161. Rind, F. C., and Simmons, P. (1992). Stern, M., Thompson, K., and Zhou, Trimarchi, J. R., and Schneiderman, Ydenberg, R. C., and Dill, L. M. Orthopteran DCMD neuron: a P. (1995). Octopaminergic neurons A. M. (1995a). Initiation of flight (1986). The economics of fleeing reevaluation of responses to mov- in the locust brain: morphologi- in the unrestrained fly, Drosophila from predators. Adv. Study Behav. 16, ing objects. I. Selective responses to cal, biochemical and electrophysi- melanogaster. J. Zool. 235, 211–222. 229–249. approaching objects. J. Neurophysiol. ological characterisation of poten- Trimarchi, J. R., and Schneiderman, A. 68, 1654–1666. tial modulators of the visual sys- M. (1995b). Different neural path- Conflict of Interest Statement: The Santer, R. D., Rind, F. C., Stafford, R., tem. J. Comp. Physiol. A 177, ways coordinate Drosophila flight authors declare that the research was and Simmons, P. J. (2006). Role of an 611–625. initiations evoked by visual and conducted in the absence of any com- identified looming-sensitive neuron Sugrue, L. P., Corrado, G. S., and New- olfactory stimuli. J. Exp. Biol. 198, mercial or financial relationships that in triggering a flying locust’s escape. some, W. T. (2005). Choosing the 1099–1104. could be construed as a potential con- J. Neurophysiol. 95, 3391–3400. greater of two goods: neural curren- Wiersma, C. A. G. (1947). Giant nerve flict of interest. Santer, R. D., Simmons, P. J., and Rind, cies for valuation and decision mak- fiber system of the crayfish; a contri- F. C. (2005a). Gliding behaviour ing. Nat. Rev. Neurosci. 6, 363–375. bution to comparative physiology of Received: 07 March 2012; paper pend- elicited by lateral looming stimuli Sun, H., and Frost, B. (1998). Computa- synapse. J. Neurophysiol. 10, 23–38. ing published: 23 April 2012; accepted: 08 in flying locusts. J. Comp. Physiol. A tion of different optical variables of Wiersma, C. A. G. (1952). Repetitive dis- August 2012; published online: 28 August 191, 61–73. looming objects in pigeon nucleus charges of motor fibers caused by a 2012. Santer, R. D., Yamawaki, Y., Rind, F. C., rotundus neurons. Nat. Neurosci. 1, single impulse in giant fibers of the Citation: Herberholz J and Marquart and Simmons, P. J. (2005b). Motor 296–303. crayfish. J. Cell. Physiol. 40, 399–419. GD (2012) Decision making and behav- activity and trajectory control dur- Sztarker, J., and Tomsic, D. (2008). Wine, J. J. (1984). The structural basis ioral choice during predator avoid- ing escape jumping in the locust Neuronal correlates of the visually of an innate behavioural pattern. J. ance. Front. Neurosci. 6:125. doi: Locusta migratoria. J. Comp. Physiol. elicited escape response of the crab Exp. Biol. 112, 283–319. 10.3389/fnins.2012.00125 A 191, 965–975. Chasmagnathus upon seasonal vari- Wine, J. J., and Krasne, F. B. (1972). The This article was submitted to Frontiers Santer, R. D., Yamawaki, Y., Rind, F. C., ations, stimuli changes and percep- organization of escape behaviour in in Decision Neuroscience, a specialty of and Simmons, P. J. (2008). Prepar- tual alterations. J. Comp. Physiol. A the crayfish. J. Exp. Biol. 56, 1–18. Frontiers in Neuroscience. ing for escape: an examination of the 194, 587–596. Wine, J. J., and Krasne, F. B. (1982). Copyright © 2012 Herberholz and Mar- role of the DCMD neuron in locust Sztarker, J., and Tomsic, D. (2011). Brain “The cellular organization of cray- quart. This is an open-access article dis- escape jumps. J. Comp. Physiol. A modularity in arthropods: individ- fish escape behavior,” in The Biol- tributed under the terms of the Creative 194, 69–77. ual neurons that support “what” but ogy of Crustacea, eds D. C. Sande- Commons Attribution License, which Schall, J. D. (2001). Neural basis of not “where” memories. J. Neurosci. man and H. L. Atwood (New York: permits use, distribution and reproduc- deciding, choosing and acting. Nat. 31, 8175–8180. Academic Press), 241–292. tion in other forums, provided the original Rev. Neurosci. 2, 32–42. Tomsic, D., Berón de Astrada, M., Wyman, R. J., Thomas, J. B., Salkoff, authors and source are credited and sub- Simmons, P. J., Rind, F. C., and Santer, R. and Sztarker, J. (2003). Identifica- L., and King, D. G. (1984). “The ject to any copyright notices concerning D. (2010). Escapes with and without tion of individual neurons reflecting Drosophila giant fiber system,” in any third-party graphics etc. www.frontiersin.org August 2012 | Volume 6 | Article 125 | 15

Journal

Frontiers in NeuroscienceUnpaywall

Published: Jan 1, 2012

There are no references for this article.