Airflow elicits a spider's jump towards airborne prey. II. Flow characteristics guiding behaviour

Christian Klopsch, Hendrik C. Kuhlmann, Friedrich G. Barth


When hungry, the wandering spider Cupiennius salei is frequently seen to catch flying insect prey. The success of its remarkable prey-capture jump from its sitting plant into the air obviously depends on proper timing and sensory guidance. In this study, it is shown that particular features of the airflow generated by the insect suffice to guide the spider. Vision and the reception of substrate vibrations and airborne sound are not needed. The behavioural reactions of blinded spiders were examined by exposing them to natural and synthetic flows imitating the fly-generated flow or particular features of it. Thus, the different roles of the three phases previously identified in the fly-generated flow and described in the companion paper could be demonstrated. When exposing the spider to phase I flow only (exponentially increasing flow velocity with very little fluctuation and typical of the fly's approach), an orienting behaviour could be observed but a prey-capture jump never be elicited. Remarkably, the spider reacted to the onset of phase II (highly fluctuating flow) of a synthetically generated flow field with a jump as frequently as it did when exposed to natural fly-generated flows. In all cases using either natural or artificial flows, the spider's jump was triggered before its flow sensors were hit by phase III flow (steadily decreasing airflow velocity). Phase III may tell the spider that the prey has passed by already in case of no prey-capture reaction. Our study underlines the relevance of airflow in spider behaviour. It also reflects the sophisticated workings of their flow sensors (trichobothria) previously studied in detail. Presumably, the information contained in prey-generated airflows plays a similar role in many other arthropods.

1. Introduction

Spiders often sit motionless for long periods of time, but also show fast and precisely oriented behaviours when properly stimulated. Such reactions are demanding in regard to sensory control and the orchestration of motor activity. An impressive example can be watched when the hungry wandering spider Cupiennius salei [1] is jumping from its sitting plant into the air to catch flying insects.

Cupiennius is night-active. It has been known for years to be unable to execute such jumps when deprived of its airflow sensors, the trichobothria [2,3]. It also has been known for a while that it readily jumps even with its eyes blinded, i.e. without vision. However, there are still other potential sources of information. Both vibrations of the substrate the spider is sitting on and airborne sound generated by a nearby flying insect like a fly may indicate its presence to the spider. In the present study, we examined the relevance of such additional sources of information by detailed measurements of the spider's sensory environment and a comparison of the results with available data on the physiological sensitivity of the sensory organs involved. These organs are the metatarsal vibration receptors and the trichobothria (reviews in Barth [1]). As it turned out airflow signals alone suffice to tell the spider when and where to jump to catch its prey.

The main questions then to be asked were about the significance of particular cues contained in the fly-generated airflow pattern and received by the airflow sensors (trichobothria) on the spider's legs and pedipalps [4,5]. The data presented in Klopsch et al. [6] provide the first detailed and quantitative analysis of the airflow generated by a flying insect from a sensory point of view. They allow hypotheses on the significance of particular airflow characteristics measured at the site of the sensors for the sensory guidance of the spider's jump. Interestingly, apart from the wake proper, there is airflow ahead of the flying fly. We refer to it as phase I airflow. It is sufficiently strong to be detected by the spider. In the present paper, synthetic airflows imitating the natural flows occuring during phases I and II of the fly-generated airflow [6] were applied. This allowed us to directly see the spider's reaction to them and to judge the behavioural relevance of the cues contained in the natural flow field.

As it turns out, phases I and II of the fly-generated airflow play different roles. Together they contain all the cues needed to tell the spider when and where to jump. The present findings stress the importance of airflow sensing in biology. They may also stimulate and inspire the discussion on the guidance of man-made vehicles towards a source of medium flow.

2. Material and methods

2.1. Experimental animals

All spiders used in this work were C. salei [1] bred in the department of neurobiology in Vienna. Female juvenile and sub-adult spiders were used for all experiments. At this age, the females, in particular, show more reliable willingness to hunt [3]. For the behavioural experiments, five females were used. At the time of the experiments, they were between five and 12 months old.

The following endogenous and exogenous factors affect the responsiveness of the spiders and therefore have to be considered for the behavioural experiments. Cupiennius is night-active. For practical reasons, the day and night activity rhythm of the animals was reversed with the light on from 21.00 to 09.00 and off from 09.00 to 21.00. After 48 h, the animals had adjusted to the new rhythm [7]. Following Brittinger [3] and Ungersböck [8], the spiders were fed twice every week one fly only after the experiment to keep them hungry and reactive. Spiders close to a moult were not used because of their inactivity [9]. During the behavioural experiments, vibrations and airflows were avoided as much as possible in order to keep Cupiennius alert and motivated to react to prey-generated stimuli. For the experiments, the spider's eyes were covered with a mixture of wax, colophony and carbon to render them non-functional. With their eyes covered, the spiders were much calmer than before and appeared more concentrated on the hunt.

The blowflies used as prey insects were Calliphora erythrocephala. For details, see Klopsch et al. [6].

2.2. Airborne sound: laser Doppler vibrometry

A laser Doppler vibrometer (LDV; Polytec PDV-100) was used to measure the angular deflection and the angular velocity of trichobothria upon exposure to the airborne sound generated by a flying blowfly. For that purpose, a living spider was mounted on a support in the hunting position and immobilized to ensure that only the motion of the trichobothria was measured. Because of their specific significance, the trichobothria on the spider's tarsus were in the focus of our investigations (figure 1). They form the most numerous group with the highest density of hairs. More importantly, they are the trichobothria receiving the airflow generated by an approaching fly first owing to their most peripheral location.

Figure 1.

All trichbothria on a leg of an adult Cupiennius salei. Trichobothria on the tarsus are shown enlarged in the inset and the investigated trichobothrium is highlighted by the red dot of the LDV (adapted from Barth et al. [4]).

The analogue signal was digitized using the CED Type 1401 A/D board (Cambridge Electronic Devices). Both the angular deflection and angular velocity were analysed as described for the sound pressure measurement [6].

2.3. Synthetic fly-like airflow

To simulate the flow velocity pattern generated by a freely flying blowfly close to the spider's trichobothria, a particular mechanical device was developed. The approaching blowfly induces a circulating flow field around itself. Of this flow, the spider receives a largely horizontally oriented velocity signal whose vectors point into the flight direction and increase exponentially with time (phase I of the fly-generated flow). A cylinder, rotating at 66 revolutions per minute, produced a rotational flow field which induced a horizontally oriented flow near the ground and around the spider (figure 2a,b). The magnitude of the velocities in the flow field could be measured using the particle image velocimetry system and calibrated by adjusting the revolutions per time of the cylinder. In the original, fly flow phase I was followed by the much more fluctuating phase II. This phase corresponds to the fly's wake which points backwards and downwards from its body. A fender above the rotating cylinder prevented radial airflow on the upper side of the cylinder while creating a synthetic downward pointing wake (figure 2c,d). The cylinder was moved horizontally at fly-like velocities of around 0.4 m s−1 [6]. The horizontal flow vectors near the tarsi of the spider legs then increased exponentially with time, as they do during phase I and with decreasing distance from the fly. At the same time, the wake's angle decreased from 90° to around 45° (figure 2d). The resulting synthetic flow field was similar to that generated by a freely flying blowfly [10].

Figure 2.

(a) DPIV vector map of the flow field generated by the clockwise rotating cylinder without fender. A meshed fence (mf) was placed 360° around the cylinder to protect the spider. (b) Clockwise rotating cylinder placed above the spider support with the spider sitting on it. With the presence of the horizontal Perspex plate, the rotational flow field of the rotating cylinder forms a horizontal flow above the surface around the spider. (c) A fender (fe) around the upper half of the clockwise rotating cylinder collects the airflow from above and generates a wake pointing downwards. The horizontal flow below the cylinder (owing to the rotational flow field of the cylinder itself) is still present around the spider. (d) The flow field generated by the cylinder (rotating clockwise) in combination with the fender and horizontally moving from right to left. The flow field is similar to that generated by a freely flying blowfly [6].

Results described in Klopsch et al. [6] suggested that the spider uses the flow signal of phase I to detect, recognize and localize the flying prey, whereas the fluctuating phase II triggers the prey-capture jump. To test this hypothesis, we investigated the behavioural reaction of the spider towards a stimulus simulating only phase I of the fly signal. For this purpose, the fluctuations of phase II were eliminated by a deflector at the rear of the cylinder, which directed the wake horizontally away from the cylinder. Thereby only phase I of the fly signal reached the spider (figure 3).

Figure 3.

Flow field around the spider generated by the device shown in the inset and moved horizontally from right to left with the cylinder rotating clockwise. Because the wake is guided away from the spider by a deflector (de), only the horizontal flow components pointing in the direction of the device's movement remain.

Digital particle image velocimetry (DPIV) measurements were used (set-up as described in Klopsch et al. [6]) to adjust the synthetic airflows to the actual fly flow by the empirical selection of the cylinder's rotation frequency and of the velocity of its horizontal movement.

2.4. Behavioural experiments

The spider was exposed to different natural and synthetic airflow signals in behavioural experiments in order to see its specific reaction to them.

The following situations were examined. (i) Tethered flying fly experimentally moved above the spider sitting on a bromeliad leaf or (ii) on a stiff and heavy metal plate to avoid substrate vibrations; (iii) synthetic fly-like airflow moved above the spider on the metal plate; (iv) only phase I of the artificial fly flow moved above the spider on the metal plate; (v) control experiment with device shown in figure 3 (inset) moved above the spider on the metal plate. For this last experiment, the rotating cylinder was turned off to explore the effect of the movement of the device itself without generating fly-like flow patterns.

The general set-up shown in figure 4 was the same for all behavioural experiments. The flow source, either a blowfly or the synthetic fly flow generator, was fixed to the rod (see ‘signal source’ in figure 4) and moved above the spider 5 cm above the substrate. The spider was released the same way either on the stiff and heavy metal plate or the leaf of a bromeliad. A LDV (Polytec LDV-100) was used to document the oscillations of the bromeliad leaf and to check for vibrations in the metal plate. With its working range down to theoretically 0.04 nm (200 Hz, 0 dB signal-to-noise ratio), the LDV is able to resolve vibrations smaller than those the spider vibration sensor (metatarsal organ) can detect at fly relevant frequencies up to 200 Hz [11]. A high-speed video camera (Indiecam GmbH, Vienna; up to 220 pictures s−1) was used to determine the point in time of the jump in regard to the position of the fly.

Figure 4.

Set-up used for behavioural experiments. The spider was released on a solid metal plate (mass ca 20 kg; dimensions 60×40×1 cm³). Rubber foam and air tubes disconnected the metal plate from vibrations of the building. The source of the flow signal was fixed to a rod (diameter 3 mm) which was connected to the frame by linear bearings and could be easily moved to all horizontal positions in the same plane.

Five spiders were used for the behavioural experiments. Each individual was tested in about 15 sessions. During each session, the spider was exposed to the same stimulus 10 consecutive times. The result of a session was then classified according to the spider's most active response. We distinguished between no reaction, slight reaction and jump. A slight reaction was defined as any movement towards the source of airflow except for a jump, that is, as moving one or more legs or walking a few steps. A session was classified as ‘jump’ if Cupiennius responded to the 10 stimuli with a prey-capture jump at least once. The spider rested for at least 5 min between the sessions.

2.5. Statistical tests

The experimental results were evaluated using two non-parametric tests. The Mann–Whitney test (U test) was applied to compare independent samples whereas the Wilcoxon signed-rank test served to interpret two related samples or repeated measurements on a single sample. Both tests were performed using the software XLSTAT 2009 (Addinsoft). The chosen level of significance was 5 per cent.

3. Results

3.1. Behavioural reaction to an experimentally moved humming blowfly

To learn more about the type of signals (airflow, airborne sound or substrate vibration) used by Cupiennius to detect, localize and catch flying prey, the fly's position at the moment of the spider's jump was studied with a high-speed video camera.

A fly was mounted to the set-up described in figure 4 and moved in different altitudes above the blinded spider. In all cases of successful captures (figure 5), the jump occured when the approaching fly was above the tarsus of the leg pointing towards the fly at a horizontal distance of −0.022±13.4 mm.1

Figure 5.

(a) High-speed video picture (taken at ca 220 frames s−1) showing a spider catching a blowfly. (b) Positions of flies in relation to the spider at the time of a successful jump (N = 6, n = 7). (Online version in colour.)

For a quantitative analysis of the behaviour, five spiders were tested in a total of 75 sessions. The experiments were done under red light and again the spider's eyes were covered to exclude vision. The spider jumped on average in 20 per cent of the sessions towards the artificially moved fly. In 2.7 per cent of the sessions, the fly was caught. Altogether four of five spiders jumped at least once and two managed to capture at least one fly successfully (figure 6). Figure 5a shows a successful prey capture.

Figure 6.

Behavioural experiments with the spider sitting on a bromeliad leaf and exposed to airflow generated by a tethered humming blowfly moved horizontally above it. Far right: mean values for five individuals shown separately on the left. Each individual was tested in 15 sessions (see numbers at base of histogram bars) with 10 trials each. The session was then classified with regard to the spider's most active response. Responses were categorized as ‘no reaction’ (light blue), ‘slight reaction’ (light green) or ‘jump’ (green). The classified sessions were added and plotted in normalized form.

3.2. On the multimodality of natural stimuli

Several different types of stimuli may be involved in the spider's prey-capture behaviour. First, the spider might recognize the flying prey visually. There may also be vibrations owing to the blowfly passing by. Cupiennius might detect these with its slit sense organs [1]. In addition, the trichobothria may be stimulated by air particle oscillations owing to airborne sound and by the airflow that the fly generates. As shown in the following, it is indeed the airflow which informs the spider about the approaching fly.

3.2.1. Visual stimulus

According to earlier work [3], spiders do not need vision to successfully catch flying prey. Therefore, vision is not under investigation in the present study. The eyes of the spiders were covered to exclude vision in all behavioural experiments.

3.2.2. Substrate vibration

Cupiennius has highly sensitive vibration receptors [11]. We, therefore, analysed the vibrations of a bromeliad leaf (with Cupiennius sitting on it), possibly induced by a humming blowfly passing by above the spider.

Using LDV, we measured the leaf oscillations in the x, y and z direction with the measurement point on the leaf directly underneath the spider (figure 7). Low-frequency vibrations of up to 50 Hz were detected in all directions considered (figure 7). However, these vibrations were present both before and after the humming fly was pulled above the leaf and identified as noise of no use for information about the approaching fly. While the humming fly was pulled above, the leaf narrowband vibrations at the fly's wing beat frequency (between 150 and 170 Hz) and induced by airborne sound could be measured in the x and z direction (figure 7a,c). The magnitude of these vibrations was 5.64±1.15 nm in the x-direction and 7.24±1.01 nm in the z-direction (N = 5, n = 15). In the y-direction, no clear signal could be measured (figure 7b). The second harmonic of the vibrations induced by the airborne sound of the fly (between 300 and 400 Hz) was too weak to be resolved for all directions (figure 7).

Figure 7.

Vibrograms showing the movement of a bromeliad leaf in the x-, y- and z-direction while a humming blowfly was pulled along a straight line in the x-direction 5 cm above the spider (see black arrow in the inset in c). When the humming fly passed by the leaf (between 0.2 and 0.5 ms), the first harmonic of the fly's wing beat frequency (between 150 and 170 Hz) induced vibrations of the leaf in the x- and z-direction (a and c) (N = 5, n = 15). Low-frequency vibrations up to 50 Hz (in a, b and c) were present both before and after the fly was pulled above the leaf and identified as noise representing background substrate vibrations.

Considering the threshold sensitivity of the spider's vibration receptor (see §4), Cupiennius does not detect either the vertical or the horizontal leaf vibrations induced by a blowfly humming 5 cm above it. The spider should, therefore, still be able to catch flying prey when sitting on a non-vibrating stiff and heavy plate.

A 20 kg metal plate resting on a mechanically damped table (figure 4) ensured that no vibrations were induced by the flying fly. This was verified by measuring vertical vibrations. As seen from figure 8, the spider still successfully captured the fly as predicted. All five individuals investigated jumped towards the prey and three of them caught at least one fly (figure 8). There was no significant difference between the percentage of jumps towards the prey during this experiment and that with the spider sitting on a bromeliad leaf (p = 0.794, Mann–Whitney test, null hypothesis: no difference). When sitting on the stiff plate, the spider answered with a prey-capture jump in 19.3 per cent of the sessions compared with 20 per cent when sitting on the bromeliad leaf. The rates of successful jumps did not differ either (p = 0.762, Mann–Whitney test, null hypothesis: no difference). Evidently, the substrate, on which the spider sits, has no significant influence on its prey-capture behaviour under the given conditions.

Figure 8.

Behavioural reactions of a spider to airflow under different experimental conditions. No vibration: tethered humming fly moved horizontally above a spider which sat on a non-vibrating metal plate; vibration: same as in previous experiment but with the spider sitting on a bromeliad, its natural dwelling plant; synthetic: the spider sitting on the metal plate and exposed to synthetic ‘fly-like’ airflow; fly: the spider sitting on the metal plate and exposed to a humming fly moved horizontally above it; only phase 1: the spider sitting on the metal plate exposed to synthetic phase I of fly airflow; all phases: the spider sitting on the metal plate and responding to synthetic airflow consisting of all phases of the fly-generated airflow. Light blue, no reaction; light green, slight reaction; green, jump.

3.2.3. Airborne sound

As the trichobothria are deflected by drag forces owing to airflow, they potentially also respond to the sound radiated from a blowfly. To evaluate this possibility in a first step, the deflection of the tarsal trichobothria was examined under a light microscope and the mechanically most sensitive trichobothrium selected for detailed analysis. The trichobothrium that was deflected most was 750 µm long and the most sensitive to frequencies in the range of the first harmonic of a humming blowfly (see Barth et al. [4]). Both the angular deflection and the angular velocity of this hair were measured with the LDV and compared with known physiological threshold curves [10].

A sample ‘sonogram’ up to 1000 Hz and containing the harmonics of the angular deflection and angular velocity owing to the humming fly signal is shown in figure 9a,b. Similar to the sound radiation of the blowfly, the first harmonic contained larger deflection and velocity values than the other harmonics. Its peak frequency ranged between 130 Hz and 200 Hz, depending on the individual fly and its wing beat frequency.

Figure 9.

(a) ‘Sonogram’ of angular deflection and (b) angular velocity of a trichobothrium (length 750 µm) on the tarsus of a living spider in response to a humming blowfly located 30 mm ahead and 10 mm above the trichobothrium. (c) Angular deflection and (d) angular velocity of a trichobothrium (length 750 µm) on the tarsus exposed to the sound (first harmonic) radiating from a humming blowfly. The fly was positioned at various altitudes above and at various horizontal positions in front of the examined trichobothrium. (c) The grey horizontally orientated planes represent the angular deflection and (d) angular velocity thresholds previously determined electrophysiologically [10]. Note inverted scale on x-axes of (c) and (d).

The evaluation of the first harmonic of the movement of the trichobothrium, which reflects the fly's wing beat frequency, is shown in figure 9c. At most positions of the fly, both the deflection magnitude and the angular velocity are far below the electrophysiological threshold. Only deflection and velocity values above the grey plane drawn into figure 9c,d and representing the respective thresholds of the trichobothria, are potentially effective stimuli.

In the behavioural experiments, where the fly was caught successfully, the blowfly was located 50 mm above the spider. At this altitude, both the mean values of the deflection and the angular velocity reach just 8 per cent and 6 per cent, respectively, of the physiological threshold values (figure 9c,d). The spider was unable to sense the fly by the sound it emitted (first harmonic). As the deflection magnitude as well as the angular velocity of the second and larger harmonics were less than those of the first harmonic at all measured positions and the trichobothrium's physiological threshold increases for frequencies above 200 Hz [10]. We conclude that the second and higher harmonics of the sound generated by a humming blowfly could not be sensed by the spider either.

3.2.4. Prey-capture behaviour induced by synthetic airflows

In the companion paper [6] of this study, the flow field generated by a freely flying blowfly close to the trichobothria was examined in detail. To prove that the airflow indeed represents a sufficient stimulus to elicit prey capture, the spider's reaction to synthetic flows was studied. These included flows lacking the fluctuating pattern of phase II which previous observations had suggested to elicit the spider's jump.

Complete signal

Figure 10a shows the similarity of the artificial fly-like airflow measured above the spider leg to the natural airflow. As seen from the frequency spectra (figure 10b), the structures of phase I and II are very similar indeed [6]. In both cases the actual fly and the synthetic device phase I reached the maximum at 8 Hz, with a spectral density below 0.0001 (m s−1)2 Hz−1. In phase II, the signals reach their maxima at 17.9 Hz (fly) and 15.65 Hz (synthetic), respectively, with spectral densities around 0.0006 (m s−1)2 Hz−1.

Figure 10.

(a) Normalized velocity magnitude of the artificial airflow signal generated by the ‘fly flow generator’ (black line) at the location of the trichobothria when pulled over the spider at fly-like velocities ranging from 13 cm s−1 to 81 cm s−1. The results of the 10 flights of the freely flying blowfly shown in fig. 7a in Klopsch et al. [6] are plotted in grey for comparison. The graphs are normalized as in that figure. (b) Spectral density of the velocity magnitude (vel. mag. spec. dens.) of phases I and II obtained for one measurement of the synthetic fly flow above the tarsus of leg 1. The values on the y-axis are equivalent to the squared velocity magnitude relative to the spectrum's bin width.

In the behavioural experiments, the spider was stimulated by pulling the synthetic airflow generator horizontally at 5 cm (measured from the lower edge of the cylinder to the floor) above the substrate (like the fly in experiments described in Klopsch et al. [6]). Three out of four spiders jumped during the sessions at least once. The percentages of jumps towards the synthetic stimulus (23%) did not differ significantly (p = 0.683, Mann–Whitney test, null hypothesis: no difference) from those obtained in experiments using the actual fly-generated flow (19.3%). Owing to the bigger size of the target, the ratio of successful jumps (see example in figure 8) is more than three times higher in case of the synthetic flow (18.3%) than in case of the natural fly signal (5.1%). The fly stimulus and the synthetic stimulus were answered in 49.9 per cent  and 46.7 per cent of the cases, respectively, by either a slight reaction or a jump. Thus, the spider is equally attracted (p = 0.556, Mann–Whitney test, null hypothesis: no difference) to both types of flow.

On the significance of phase I

In the quasi-natural situation, the spider jumped towards the fly when it was directly above the tarsus of the spider leg closest to it. At this instant, the flow sensors on the tarsus were exposed to the transition of the airflow from phase I to phase II. To decide whether the onset of phase II triggers the jump the spider's reaction to a stimulus representing phase I only was examined (figure 11).

Figure 11.

(a) Velocity magnitude of the airflow signal generated by the fly flow generator with the deflector attached in order to exclude the fluctuating phase II (black line) of the natural fly flow signal (grey lines). Flow measurement at the location of the spider's sensors when the flow source was pulled over the spider. Values for 10 flights of the freely flying blowfly taken from fig. 7a in Klopsch et al. [6] are plotted in grey for comparison. The graphs are normalized in the same way as in that figure. (b) Spectral density of the velocity magnitude (vel. mag. spec. dens.) of phases I and II shown in (a). The values on the y-axis are equivalent to the squared velocity magnitude relative to the spectrum's bin width. For a better comparison, the same scaling as in figure 10b was used for this graph in combination with an inset to enlarge the first section of the curve.

With the deflector attached to the stimulating device, the frequency content of phase I (velocity increase) still nicely resembled that of phase I of the naturally generated fly flow and of the complete synthetic fly-like airflow. The same was found for the structure, peak frequency and the spectral density of the velocity magnitude (figure 11b). At the end of phase I (transition to phase II), the frequency content of the velocity signal did not change as it would in case of both the natural and the complete synthetic fly-like airflow (figure 10a). As the spider does not jump later than during phase II, phase III of the natural fly signal can be excluded as the releaser of the jump. The fluctuations which occur in phase III of the natural fly signal were absent in the synthetic flow (figure 10a).

Without the fluctuations of phase II, no prey-capture jump could ever be elicited in 75 sessions. In 23.3 per cent (average) of the sessions, prey-capture jumps were recorded when exposing the spider to the complete synthetic fly-like airflow (figure 8). The percentage of slight reactions was the same in case of exposure to the complete synthetic fly-like airflow (23%) and that to phase I only (25.3%) (figure 8, p = 0.968, Mann–Whitney test, null hypothesis: no difference).

According to the video film recordings taken of all 32 slight reactions triggered by the phase I flow stimulus, the observed motions were directed towards the approaching flow source in all cases. In 16 out of the 32 cases, the spider twitched with those legs first which were closest to (and pointed towards) the approaching flow source. In the remaining 16 cases, the spider responded in the direction towards the stimulus as well but twitched with more than one leg during the 40 ms between two frames of the video recording.

In 73.3 per cent of the sessions with the spiders exposed to the synthetic fly flow without wake, they did not react at all.

Control experiment

To ensure that the spiders' reactions to the synthetic stimulus were elicited by the airflow only, the fly flow generator was pulled over the spider with the rotating cylinder turned off. Although the pulling itself caused airflow velocities of up to 0.025 m s−1 at the spider sensors, the characteristic fly-like flow pattern was lacking. Therefore, the spider should not react to it. The control experiment also explored whether the spider can sense the device in another way (substrate vibration or airborne sound, e.g. induced by the linear bearing when the device was pulled horizontally; vision was excluded anyway by covering the eyes).

In contrast to the experiments both with the natural and the artificial fly flow, the spider did not assume its characteristic hunting position in the control experiments. It wandered around in different directions, obviously not being aroused by the control stimulus. None of the four animals tested ever jumped and significantly fewer slight reactions were observed than during exposure to phase I of the synthetic flow stimulus (p = 0.009, Mann–Whitney test, null hypothesis: no difference).

4. Discussion

4.1. Substrate vibration not needed to capture flying prey

It has long been known that Cupiennius is very sensitive to substrate-borne vibrations and indeed uses vibratory signals (and cues2) in both its elaborate courtship behaviour and the capture of prey crawling around on its sitting plant (reviews in Barth [1]). Leaf vibrations induced by an insect passing by in flight might inform the spider about the presence of airborne prey as well. However, our present results demonstrate that, for the capture of airborne prey, such vibrations are not needed, whereas airflow signals alone are sufficient. This conclusion rests on a comparison of the magnitude of the plant vibrations and previously determined threshold curves of the metatarsal vibration receptor organ [11]. As it turns out, the threshold values in the relevant frequency range are 10 to 42 times higher than the transversal vibrations owing to the flying fly. Even the largest leaf vibration measured (9.18 nm) is much smaller than the lowest sensory threshold (50 nm) measured at frequencies around 1 kHz. Similar arguments apply to the horizontally oriented vibrations of the leaf. The conclusion that the substrate vibrations under discussion are sub-threshold and not necessary in the given prey-capture context is strongly supported by an additional finding. The spiders successfully catch the flies even when standing on a damped, heavy and stiff metal plate (instead of on the plant), where no vibrations could be measured at all with the laser vibrometer.

4.2. The importance of fluctuations

Our present results underline the importance of fluctuations contained in the stimuli of biological relevance for the spiders. It had previously been demonstrated that the airflow sensors, filiform hairs or trichobothria, form a system particularly well adapted to the reception of transitory as opposed to static stimulation. This adaptedness applies to both the morphology and mechanics, and to the physiology of the sensilla. Their nervous response to the deflection of the hair shaft is strictly phasic. The same response type was also found for neurons in the central nervous system integrating the data provided by the sensors in the periphery ([10,12,13]; for reviews, see Barth [1] and Humphrey & Barth [5]). Being movement detectors, the sensors do not respond to the position of the hair shaft but to changes in position. This makes sense in light of the fact that the hair shaft is deflected by the frictional forces owing to the airflow which correlate with velocity. Interneurons in the central nervous subesophageal mass were shown to predominantly respond to the sudden increases of flow velocity contained in the ‘fly signal’ and to be well suited to encode the fluctuating time course typical of prey signals [12]. Our present results extend these findings to the behavioural level. They demonstrate the particular relevance of the onset of phase II of the airflow and of its prominent fluctuations for the release of the spider's jump.

The characteristic features of substrate-borne vibrations used by the spider to identify prey seem to be remarkably similar. The vibrations need to be non-sinusoidal and ‘noisy’, covering a relatively large frequency range, and to exhibit a rather irregular temporal structure. Both the slits of the metatarsal vibration receptor organ and interneurons in the spider's central nervous system show a substantial increase in threshold sensitivity when stimulated with small frequency bands (1/3 octave) instead of purely sinusoidal vibrations. The drop in sensitivity may be as large as 20 dB [14,15]. Likewise, prey-catching behaviour can be elicited significantly more often with broad-band vibratory signals than with sine waves [16].

4.3. Orientation towards the stimulus source

Establishing the presence of a prey animal and identifying it as such with the help of a highly evolved system of airflow sensors [5] is only the first part of what the spider has to accomplish. The subsequent task is to jump at the right moment and to actually seize the prey. The jump needs to be oriented well enough in regard to the signal source. The fly's translational velocity has to be taken into account. We are still far from understanding details of the neuronal networks and their operations which underlie the spider's remarkable performance. One characteristic of the spider's behaviour makes the task easier in regard to the necessary precision: the spider uses its front legs (leg span in adult animals up to more than 10 cm) instead of its chelicerae to grab its prey from a rather wide space. Surprisingly, it is able to adjust its body and leg positions even when already in the air. Possibly, the spider thereby fine-tunes its orientation towards the prey. However, these movements have still to be analysed in detail.

When prepared for prey capture while still sitting on its dwelling plant, Cupiennius keeps its eight legs in stereotypical positions forming almost uniformly distributed radii of a full circle [3,17]. We therefore speculate that the predatory space of the spider with its ring-shaped arrangement of some 900 trichobothria is represented as a cylindrical rather than a Cartesian coordinate system in its central nervous system. If this was true, the spider would use radial horizontal distances, angular coordinates (direction) and height (altitude) instead of the Cartesian horizontal distances x and y and height z. The fact that the spider can detect flies equally well in all horizontal directions seems to support the hypothesis [2,3].

In the following, we summarize arguments and experimental findings at least in a fragmentary way to indicate possible ways of extracting information on the position of the signal source from the airflow produced by the flying insect.

4.3.1. Horizontal distance

When the fly comes close to the spider, the successive occurence of the three phases of the airflow close to the trichobothria indicates both the fly's approach and its horizontal distance from the spider. The start of the signal and the instant of transition from phase I to phase II of the flow (which triggers the jump) occur independently from the fly's altitude within the limits of 20–60 mm. Although the spider responds to the onset of phase I (e.g. by turning towards the leg stimulated first), it does not jump into the air yet. Obviously, only the transition to phase II informs the spider that the distance to the prey is short enough for a jump. The high-frequency content of the wake behind a fly rapidly decreases with distance. The results of an experiment where the spider was exposed only to the fly wake at increasing distances underline the overall conclusions. The spider did not jump towards the wake anymore as soon as it began to look like ‘background noise’ at a distance of some 25 cm. The airflow then showed a very narrow frequency spectrum at very low frequencies. In this case, ‘too far away’ is equivalent to ‘not identified as prey generated flow’, because the high-frequency components are all gone [10].

Presumably, fish are able to predict the horizontal distance of prey by evaluating the positions at the lateral line canal where either extreme values (minimum and maximum, respectively) or zero crossings of water flow velocity occur [18,19]. The distance to their prey linearly depends on the distance between these points along the lateral line. This hypothesis is not applicable to Cupiennius, where neither extreme values of airflow velocities nor zero crossings are found at more than one position simultaneously.

4.3.2. Direction Time differences

The present study confirms the results of previous behavioural experiments showing that the spider always turns into the direction of the leg stimulated first [1,3]. Thanks to the dispersed and ring-shaped arrangement of the trichobothria all around the spider any azimuth of the stimulus source can be identified. In case of the artificial simultaneous stimulation of two adjacent legs, the spider chooses the compromise and turns into a direction between the two legs. The simultaneous stimulation of non-adjacent legs seems to be too unnatural; the spider does not respond at all in most cases. According to the behavioural experiments, a highly significant effect of ‘non-simultaneous’ is seen at a time difference of more than about 50 ms. However, even at a time difference of only 10 ms, the spider turns more often to the leg receiving the stimulus first [1,3]. With freely flying blowflies passing by above a spider of about 7.5 cm leg span (as used in the experiments of the present study), the differences in time of arrival measured 86±19 ms (N = 18) (see also fig. 11 in Klopsch et al. [6]).

Time differences are also used by the spider to orient towards a source of substrate vibrations. Again Cupiennius turns towards the leg first stimulated and at time differences of 4 ms only [17]. A scorpion was found to be even more time-sensitive, reacting to delays of 1 ms and less [20]. Intensity differences

In addition to time of arrival differences, Cupiennius may also use stimulus intensity differences to orient towards a stimulus source. We measured airflow-velocity ratios above different legs of up to 6.5:1 (mean: 4.04 ± 1.32 for 20 flights) during phase I when freely flying flies passed over the spider. The strongest airflow is also the one first appearing above the leg tarsus pointing in the direction where the stimulus flow comes from. Clearly, the role of airflow intensity differences needs further study. The experiments and measurements to be carried out will be very demanding, however.

When separately and simultaneously stimulated at two legs with substrate vibrations of different magnitude, Cupiennius turns towards the leg with the larger stimulus [17], as does the scorpion [20]. However, the vibration magnitudes measured on the leaves of plants preferred by Cupiennius as their dwelling plant show rather complex, frequency-dependent spatial patterns which should considerably complicate the determination of stimulus direction by the spider using intensity differences [1,21]. The use of time differences for the same purpose seems to be much more reliable.

4.3.3. Altitude

The question how the spider recognizes the altitude of the flying prey from properties of the airflow can only be addressed with a few speculative comments. Does it need to measure altitude at all? This may not be the case because the spider only needs to know when the fly is within the reach of a jump. This information is contained in the degree of fluctuation of phase II flow, whereas the transition from phase I to phase II, which triggers the jump, is independent of the fly's altitude (see Klopsch et al. [6]).

Like hydrodynamic stimuli, airflow stimuli get increasingly blurred with increasing distance. This is not too different from the blur of visual stimuli which serves as a cue for pictorial depth [19]. In case of airflow, the measure for blur could be the degree of velocity fluctuation. In phase I of the fly-generated flow, the fluctuation increases linearly with the fly's altitude. Theoretically then, it could serve to inform the spider about the altitude of the flying prey.

4.3.4. Signal source velocity

The difficulties of the spider's task are increased by the fact that the fly is moving fast. Knowledge of the fly's flight velocity should be helpful. Theoretically, such information is indeed available because the velocity correlates with the differences in time of signal arrival at the different legs. In addition, the steepness of the exponential increase of the airflow velocity in phase I increases linearly with horizontal flight velocity. Both measures are independent from the fly's altitude and the direction of its approach so that they may well be used to determine horizontal flight speed (see figs 9 and 11 in Klopsch et al. [6]).

The results presented here show that in principle information on five stimulus parameters suffices to detect, recognize and localize a flying prey and to trigger the actual jump towards it. These parameters are velocity magnitude, velocity fluctuation, velocity gradient, time of stimulus arrival difference and intensity difference. This and the knowledge about the arrangement of the trichobothria and their functional properties may also be helpful when designing an artificial sensor system detecting and localizing a source of air or water flow. It has to solve the same basic problems as the spider.


We are grateful to have been funded by the DARPA BioSenSE programme grant no. FA9550-05-1-0459 to F.G.B. We also thank the late J.A.C. Humphrey for his valuable advice and friendship and C. F. Schaber and T. Hoinkes for helpful discussions and assistance in the laboratory.


  • 1 The manipulation of the blowfly's natural flight characteristics by manually pulling it is thought to be the reason for the large standard deviation of this value. Supposedly then, when coming close to the tip of the spider leg the fly generates a signal (substrate vibration, airflow or sound) informing the spider about the right moment to jump.

  • 2 In the study of animal communication and its evolution signals are distinguished from cues. Here we use ‘signal’ in a technical sense and do not distinguish between the two terms.

  • Received October 7, 2012.
  • Accepted January 29, 2013.


View Abstract