The behavioral responses to a wind stimulus were studied in the freely moving cricket, Gryllus bimaculatus. The observed responses included walking, running, jumping, turning, withdrawing, abdominal lifting, hind-leg lifting, kicking, cercal cleaning, antennal swinging, and “no action” and were classified into five behaviors as functional categories: escape, evasion, offense, surveillance and “no action”. The elicitation of each type of behavior by an identical stimulus was variable and unpredictable. However, the stochastic approach showed that the probabilities of the behaviors converged in a series of responses obtained from one cricket and in mass response data collected from 48 crickets. These probabilities, the values calculated from the different populations, were the same. The statistical analysis, using the probabilities, revealed that the probabilities of these wind-evoked behaviors were affected by the intensity of the wind stimulus. The escape and “no action” were dominant, together constituting more than 60% of the total responses. The stimulus method and the animal conditions determining a high probability of escape are also discussed.
Escape systems have attracted interest in the study of neural mechanisms of behavior because of the clear causality between stimulus and response. For example, the giant interneuron systems of the crayfish and the cockroach were revealed to elicit the stereotyped motor output of escape in response to a particular key stimulus (e.g., Wine and Krasne, 1982; Camhi, 1980).
The cercal wind sensory system of the cricket has also been studied in connection with escape behavior. Studies on the cercal sensory system of the cockroach (Camhi and Tom, 1978; Camhi et al., 1978), whose response to wind is primarily escape running, prompted studies of the corresponding cricket system. The cercal sensory giant interneurons of the cricket resemble those of the cockroach in morphology and physiology (Edwards and Palka, 1974; Levine and Murphey, 1980; Baba et al., 1991; Hirota et al., 1993), and crickets also show escape running to response to wind (Tauber and Camhi, 1995). The neural activities, signal processing, connection plasticity and other general properties of the neurons of the cercal sensory system of the cricket have also been studied, not necessarily in connection with the escape behavior (e.g., Matsumoto and Murphey, 1977; Kanou and Shimozawa, 1984; Chiba and Murphey, 1991; Baba et al., 1995).
In the cricket, however, wind-evoked responses other than escape, such as kicking (Dumpert and Gnatzy, 1977), anti-wasp stand-still posture (Gnatzy and Heußlein, 1986) and the neural responses of the giant interneurons related to mating songs (Kämper, 1984) have been reported. From daily observations of our breeding colony, it has become clear that crickets do not respond to wind only by escape, and that most rutting male crickets respond to wind by back-walking to females.
We thus hypothesized that wind stimulus could drive diverse responses as well as escape, and several questions arose: first, how many responses does a wind stimulus to the cerci evoke? Second, are the diverse responses related to the stimulus parameters? Third, how can the diverse responses be analyzed? Finally, are the responses of wind-sensitive interneurons related to escape behavior?
In the present study, we used a video camera to observe the responses of freely moving crickets to a wind stimulus. We observed ten types of responses and a “no action” response and we classified the responses into five behaviors as functional categories. The responses elicited by an identical stimulus were multitudinous, variable and unpredictable but probabilistic. The probabilistic properties of the wind-evoked responses are described and the particular stimulus driving the escape behavior is discussed.
MATERIALS AND METHODS
Adult male crickets (Gryllus bimaculatus) with intact cerci were used 3–30 days after the imaginal molt. They were raised in a 12–12 hr light-dark cycle at the constant temperature of 28°C in our laboratory. The compound eyes of a cricket were removed a few days prior to the behavioral observation. This cricket did not have the shadow reflex.
The observation arena was 18 cm in diameter surrounded by a transparent plastic 10 cm-high wall (Fig. 1). Each cricket was placed in the arena and allowed to explore the area for more than 2 min. An identical wind puff was then applied 10–15 sec after the cricket stopped moving. The duration of the wind stimulus was 2 sec. To observe a series of responses (one series consisted of 50–60 wind stimuli), the same stimulus was applied again 10–15 sec after the insect stopped at one place. All observations were conducted at room temperature (20–28°C) between 9:00 a.m. and 8:00 p.m.
The wind puff was an air jet discharged through a fine nozzle (1 mm in inner diameter) connected to a compressor pump through an electrical solenoid valve. The tip of the nozzle was placed 2 cm above the cercus. Seven different stimulus intensities were prepared by using air guide tubes of different sizes. Wind velocities of 38, 66, 83, 195, 351, 427 and 770 cm/s (a range of 23-fold) were calibrated by a hot-wire anemometer (HC-24, Hayakawa-keisoku co., Japan). The theoretical estimation of the wind velocity from the volume of ejected air and the inner diameter of the nozzle with a parabolic velocity profile assumption (Kanou and Shimozawa, 1983) also confirmed the calibration. We used the wind velocity for the measure of stimulus intensity. We did not monitor the temporal waveform, acceleration or spatial distribution of each wind stimulus at the cercus. According to fluid dynamics theory, the acceleration and the spatial size of wind increases with its velocity (Kanou and Shimozawa, 1983).
The reaction of the cricket was recorded by a video camera. A light indicating the duration of the stimulus was recorded in the corner of the frame. The physical delay of the wind stimulus, after this indicator was turned on, was measured by the hot-wire anemometer 2 cm away from the nozzle. The wind velocities reached 10% and 90% of their final values at 10 and 20 msec, respectively, after the electrical stimulus onset. The frame rate of the video camera was 30 Hz, and one frame had 33 msec uncertainty. Response latencies were measured by assuming that the response and stimulus duration light indicator onset started at the beginning of the first recognizable frame. Therefore, the latency contained a margin of error of 33 msec.
The classification of the responses was based on the finding of previous reports (Table 1). The terms “head stand”, “still-stand”, and “kicking” in Gnatzy and Heußlein (1986) were replaced by “abdominal lift”, “hind-leg lift”, and “anti-wasp kick”, respectively, in the present study. We also coined new terms when we did not find any appropriate terms in the previous reports. The ten responses and “no action” we observed were classified into five behaviors as functional categories. This classification was used for the statistical analysis.
Responses of crickets to wind stimuli: previous observation
We used the chi-squared test to test significant differences. All tests were performed with the raw data of the behaviors. If the expected value was less than 5, some behaviors were analyzed together to produce a value of 5 or more; usually evasive behavior, offensive behavior, and surveillance behavior were analyzed together.
Responses elicited by the wind stimulus
We observed approximately 5,000 responses to the wind puffs in more than 900 crickets.
Ten motor responses (walk, run, jump, turn, withdrawal, kick, abdominal lift, leg lift, cercal cleaning, antennal swing) and “no action” were recorded. The stimulus evoked all of the responses reported in previous studies except for the anti-wasp kick (Table 1). Additionally, the stimulus elicited three responses not mentioned previously: withdrawal, wherein the cricket moved its body backward and the angle between the tibia and femur decreased; cercal cleaning, in which the cricket wiped its cercus(i) with the hind-leg(s); and antennal swing, i.e., the cricket moved its antennas (Fig. 2).
The responses (other than the antennal swing) required the movement of the leg(s), and thus none of these responses could be performed simultaneously. In contrast, the antennal swing occurred simultaneously with 60% of the kick (n=10), 63% of the walk (n=81), 70% of the run (n=20), 75% of the leg lift (n=4), 78% of the abdominal lift (n=9), 80% of the turn (n=50), 87% of the cercal cleaning (n=8) and 90% of the withdrawal (n=5). We were unable to tell whether the jump with the antennal swing occurred or not, because the jump was too rare to observe.
Sometimes the responses involving leg movement, however, occurred in a sequence: a walk or a run followed a turn; and a cercal cleaning, a kick, or an abdominal lift followed a walk. Following a turn, 90% of the crickets remained in place and 10% showed a subsequent walk or a run response (n=530). Other serial responses were observed only three times out of all the experiments.
The latency periods of overt responses after the stimulus onset were unexpectedly long, as shown in Table 2 (we did not measure the latency periods of the jump and the hind-leg lift). The mean latencies and the range of SDs were more than the 8th frame (about 264 msec) and the 6–11 th frame. The shortest latencies of most of the overt responses (the responses other than “no action”) were less than the 3rd frame, but that of cercal cleaning was the 10th frame. When the antennal swing occurred with another response, it always appeared at least one frame earlier than or within the same frame as the other responses.
Latencies of overt responses by crickets to a wind stimulus
Probabilistic property of the behaviors elicited by an identified stimulus
The responses were classified into five categories according to cricket's function for the sake of statistical analysis: four overt behaviors (escape, evasive, offensive, and surveillance) and “no action”. The criteria for the behaviors are listed in Table 3. When the antennal swing occurred with another response, we classified the responses according to this particular response.
Criteria of behaviors in response to wind
The appearance of the behavior was variable and was not predictable. The series of behaviors elicited by an identical stimulus (one series was comprised of 55 stimuli; number of crickets=12) was variable (Fig. 3). The probabilities of the behaviors showed stochastic independence. For example, the probability of the “no action” was P(na)=18/54(1/3), and the conditional probabilities were P(na | na)=8/18, P(na | escape)=7/26, P(na | offensive)=1/3, P(na | evasive)=1/4, and P(na | surveillance)=1/3. These conditional probabilities were not significantly different from the P(na). All the data on the “no action” and the escape behavior were stochastically independent. The other behaviors were also thought to be stochastically independent; we had too little data to perform a stochastic analysis of them. Thus, we did not find any patterns in the appearance of the behaviors.
However, the probabilities of the series of responses in the total data converged (Fig. 4). These values were almost identical to the data collected from 48 individual insects tested with the same wind intensity (Table 4). The probability value of the “no action” behavior in the series response of the weak wind intensity tended to be larger than that of the collected data for the same intensity. This tendency is thought to have been caused by habituation (Table 4). In addition, two collected data, which were obtained from different insects in another time, were not significantly different (Table 5). According to these data, the behavior could not be predicted but the probability of a certain behavior evoked by an identified stimulus had to be determined, and the probability percentages drawn from the collected data from 48 crickets were regarded as the probabilities of each behavior.
Raw data of wind behaviors
Raw data of wind behaviors in “group C”
Probabilistic property of the behaviors with wind intensity
We then examined the probabilistic properties of the behaviors. The data of 48 crickets were analyzed. As the stimulus intensity increased, the rate of the total overt behaviors, all behaviors without “no action”, increased. The probability of the total overt behaviors showed a linear regression with the logarithms of wind velocity (Fig. 5). The wind intensity at which there was a 50% probability of the occurrence of the overt responses was 129 cm/s.
The occurrence probability of each overt behavior is shown in Fig. 6. The occurrence of each overt behavior depended on the wind intensity. The escape and “no action” behaviors were dominant, together comprising more than 60% of the total responses (Fig. 6A). As the stimulus intensity decreased, the probability of the no-action behavior increased and that of the escape behavior decreased. The gradient of the “no action” behavior and the escape behavior were diametrically opposite. In contrast, the probabilities for the evasive, offensive, and surveillance behaviors showed an optimum range of stimulus intensity and were less than 0.4 (Fig. 6B).
Furthermore, although the total number of the overt behavior responses depended on the wind intensity, the composition of the overt behaviors in the range from 38 to 195 cm/s seemed to be uniform (Fig 7). At more than 195 cm/s, the escape behavior increased in frequency and the offensive and surveillance behaviors decreased. The evasive behaviors peaked at 427 cm/s.
Classification of the responses and behaviors
The crickets responded to the wind stimulus with eleven responses. Our classification of these responses was based on the responses observed in previous studies (Table 1) and the three new responses: withdrawal, cercal cleaning, and antennal swing (Fig. 2). The terms used for the responses used in the previous and present studies denote serial movements such as walking, running, jumping, etc. This classification was simple, convenient, and useful.
We further classified the responses into five behaviors of crickets. It was not possible to completely eliminate our subjectivity from this classification, e.g., in labelling a behavior as “escape”. This classification, however, seemed appropriate for analyzing the behaviors statistically and for clarifying the wind responses in the large context.
It is important to note that the terms used for the responses and behaviors do not correspond precisely to the cricket's neural network. Our study was an attempt to understand the wind responses, and we acknowledge that this classification could be challenged and improved.
Responses of wind to the cerci
A remaining issue is about the number of the types of responses of the cerci evoked by a wind stimulus in a cricket. Our stimulus evoked all of the previously reported responses except for the anti-wasp kick. We think that the cricket has yet other wind responses which were not elicited by our stimulus and experimental conditions. The back-walk and singing the mating song responses could be added to the twelve wind responses (our eleven responses, and the anti-wasp kick). Kämper (1984) suggested that the cricket's giant interneurons contributed to its mating behavior, and we found in a preliminary study that a wind stimulus could induce a rutting cricket to back-walk and perform a mating song. Although the frequency of each response is not yet established, crickets are thus seen to have at least fourteen wind responses in their repertory.
Variety of the behaviors
The question arises as to whether the various responses of the cricket to wind are related to the stimulus method. The number of wind-evoked responses reported in previous studies was five or less responses (Table 1). Our present results showed that the crickets' behaviors were affected by the wind intensity (Fig. 5). The probability of each behavior can also be affected by the application of the stimulus and by the insect's condition. For example, the offensive behavior was not evoked with the wide wind stimulus (Tauber and Camhi, 1995), and the vibration generated by a wasp elicited the anti-wasp kick (Gnatzy and Heußlein, 1986) but the air current we used did not. Furthermore, freely moving crickets showed more varied responses than those of “fixed” cricket (Table 1). The differences in the response repertoire in each study are quite likely caused by differences in the stimuli used and in the experimental conditions.
We observed that crickets had multiple responses or behaviors to an identified stimulus. We do not know the exact cause(s) of the variety of responses to the same stimulus. Possible candidates for causes of the variety are the fluctuation of neural activity affected by internal and external noise such as circumstance noise around the cricket, the secretion of hormones and neural modulators, respiration, and pulsation. These factors could not be completely eliminated, and we thus could not create a stimulus and experimental conditions that would produce only one particular response.
However, our present and previous data showed that it is possible to apply an appropriate stimulus and experimental condition which elicit a particular response in high probability. For example, if researchers want to study escape behavior, they should use a high-intensity and wide-area wind stimulus, since our data showed that the escape behavior increased with wind intensity (the acceleration and stimulus area also increased with wind intensity; see Method's section), and Tauber and Camhi's (1995) data suggested that wide-area wind elicited escape behavior. The identification of the appropriate stimulus and conditions will enable researchers to study the neural circuits related to a particular response.
Stochastic approach to analyzing the wind responses
Another question concerns how to analyze the diverse responses. The stochastic approach was useful. We regarded the series of wind behaviors (Fig. 3) as discrete parameter sequences, and the series was then essentially a stochastic process; we analyzed the series with conventional stochastic tools. Our data showed that the probabilities of the wind behaviors were variable and stochastically independent, but converged (Figs. 3 and 4). Further, the series must be an ergodic process, because the converged value of each cricket was almost identical to the mean value of the data collected from 48 crickets (at the same wind intensity) (Table 4). Another advantage of the stochastic analysis was that it provided a way to compare the appearance probabilities with changes of a particular parameter such as the intensity of stimulus. We conclude that the stochastic approach is useful to analyze cricket wind responses.
Probabilistic property of the behaviors and neural activity
The final question was whether the responses of wind-sensitive interneurons were related to escape behavior. Many data of giant interneurons obtained using a wind puff are probably related to escape behaviors because the escape behavior was dominant as the response to a high-intensity stimulus (Figs. 6A, 7) and the wind stimulus which covered more than the area of the whole cerci tended to evoke an escape response (Tauber and Camhi, 1995). However, the responses of the giant interneurons are not related only to escape behavior. For instance, sinusoidal wind, which evoked the responses of giant interneurons, elicited anti-wasp behaviors (Kämper, 1984; Gnatzy and Kämper, 1990). We recommend the use of a wide area air current stimulus to study the neural mechanisms eliciting escape.
Although we do not know the exact relation between neural activity and the behavior observed in the present study, the behaviors reflected some type of the neural activity. For example, the probability of overt behaviors showed a linear relationship with the logarithms of wind intensity, and the wind response of sensory neurons and some interneurons also showed a similar curve with wind intensity (Miller et al., 1991; Landolfa and Miller, 1995). Besides, the composition of the overt behaviors in the range from 38 to 195 cm/s seemed to be uniform, and when it was more than 195 cm/s, the escape behavior increased in frequency and the offensive and surveillance behaviors decreased (Fig 7). This character also no doubt reflects the neural mechanism of behavior's choosing. By analyzing spike numbers and thresholds of the wind sensitive interneurons, we are able to take hints of the mechanism of decision making, which underlies its neural networks.
Finally, the previous studies about wind responses of the giant interneurons showed a mean spike number as representative of the response. Our data suggested that the fluctuation of neural activity was a factor in the variety of responses. The stochastic property of the wind behavior is thought to reflect the fluctuation of neural activity. If the series of neural activities is examined more carefully, we will identify the fluctuation. The probabilistic approach to the neural activity is a promising method for the investigation of the neural circuits related to behavior.
This study was supported by a Grant-in-Aid for Scientific Research 08740646 from the Ministry of Education of Japan and a Special Grant-in-Aid for the Promotion of Education and Science at Hokkaido University, provided by the Japanese Ministry of Education, Science and Culture.