Comparison of smooth and hairy attachment pads in insects: friction,adhesion and mechanisms for direction-dependence

Many arthropods and vertebrates possess tarsal attachment systems, which broadly divide into two principal groups, namely hairy (fibrillar) and smooth adhesive pads (Scherge and Gorb,2001). Fibrillar adhesive systems have recently attracted much attention from the engineering and physical sciences because they are considered to be promising models for novel, biomimetic adhesives(Aksak et al., 2007; Glassmaker et al., 2004; Gorb et al., 2007; Hui et al., 2004; Jagota et al., 2007; Kim and Sitti, 2006; Lee et al., 2008; Menon and Sitti, 2006; Schubert et al., 2008). The fibrillar design is thought to convey a number of specific advantages, such as superior performance on rough substrates(Persson and Gorb, 2003),effortless detachment (Autumn and Hansen,2006; Autumn et al.,2006a; Autumn et al.,2006b; Federle,2006), self-cleaning properties(Hansen and Autumn, 2005) and increased adhesion due to contact splitting(Arzt et al., 2003). However,hairy and smooth adhesive pads have evolved repeatedly in different taxa(Beutel and Gorb, 2001; Gorb et al., 2002), and it appears that both designs fulfil the requirements for successful climbing on diverse substrates. Surprisingly, it is still unclear whether the performance of the two designs is any different and, if so, in what respect. Smooth and fibrillar adhesive systems have been the subject of recent reviews(Autumn, 2007; Barnes, 2007; Federle, 2006; Gorb, 2007; Persson, 2007). However, no previous study has explicitly compared the performance of these systems under controlled conditions, and there exists a lack of information regarding the forces that can be supported by these systems in insects. This study attempts to address this by comparing a representative example of each, the hairy pads of the leaf beetle Gastrophysa viridula De Geer (Coleoptera) and the smooth arolium of the stick insect Carausius morosus Brunner(Phasmatodea) (Fig. 1).

Unlike the `dry' adhesives of spiders and geckos, both smooth and fibrillar adhesive systems in insects require the use of a secreted fluid(Betz, 2003; Eisner and Aneshansley, 2000; Federle et al., 2002; Gorb, 1998; Langer et al., 2004). This mediates attachment through capillary and viscous forces and may help overcome the problem of rough substrate attachment by filling in surface crevices. It has been shown, however, that a build-up of fluid is detrimental to the performance of smooth pads on smooth substrates due to the smaller forces resulting from a thicker fluid layer(Drechsler and Federle, 2006). As such, in the present study, we tested whether a similar effect occurs in the hairy system.

Insects with hairy and smooth pads can generate very high attachment forces(Eisner and Aneshansley, 2000; Federle et al., 2000). While strong adhesion may be beneficial in many situations, it can make locomotion more difficult. The problem of how to effect a controlled, energy-efficient detachment is of particular importance to leaf-dwelling insects, which must be capable of adhering to a range of demanding substrates while still being able to rapidly detach (for example during the pursuit or evasion of other animals). A fundamental property of adhesive structures that helps to achieve rapid and controllable adhesion during locomotion is their direction-dependence. It has been observed that friction and adhesion forces of most animal attachment organs are higher when pulled towards the body (i.e. proximally) rather than pushed away from it (i.e. distally). Previous studies have shown this to be the case in hairy adhesives of geckos, spiders and flies(Autumn et al., 2000; Autumn et al., 2006a; Hill, 1977; Niederegger and Gorb, 2003) as well as in smooth adhesive pads of ants, bushcrickets and cockroaches(Clemente and Federle, 2008; Federle et al., 2001; Federle and Endlein, 2004; Gorb and Scherge, 2000).

However, the detailed mechanisms underlying direction-dependence are still not sufficiently understood in either smooth or fibrillar systems. It is unclear whether direction-dependence is achieved through changes in contact area or through a change in shear stress (i.e. friction force per unit contact area). Analysing the smooth euplantulae of the bushcricket Tettigonia viridissima, Gorb and Scherge proposed that friction is direction-dependent due to the bending and reorientation of the inner rods of the pad cuticle (Gorb and Scherge,2000). However, this hypothesis does not specify whether the action of the rods during a proximal pull is thought to increase pad contact area or shear stress. Large direction-dependent changes of contact area can occur in adhesive pads that can be unfolded. It has been shown for ants and bees that changes of adhesive contact area are mediated both by the action of the claw flexor muscle and by the passive unfolding of the pad when legs are pulled towards the body (Federle and Endlein, 2004; Federle et al.,2001). However, even in the absence of such an unfolding mechanism, direction-dependent changes of contact area can be brought about by the flexibility of the chain-like tarsus. This has been shown for the cockroach Nauphoeta cineraea, where a distal movement of the unrestrained tarsus caused a peeling detachment of the smooth arolium(Clemente and Federle,2008).

For the fibrillar adhesive system, previous explanations of direction-dependence have focused on the behaviour of individual setae. Adhesive hairs of geckos are non-symmetrical and feature distally pointing setae and spatulae, which have been shown to generate higher friction and adhesion when aligned with a proximal pull(Autumn and Hansen, 2006; Autumn et al., 2000; Autumn et al., 2006a; Gravish et al., 2008). Without a proximal pull, only the ends of the spatulae will contact the surface,representing only a small fraction of the total possible contact area. This results in a highly direction-dependent friction on the level of individual hairs. However, fibrillar adhesive systems have a hierarchical structure, and attachment and detachment may not only be controlled on the level of individual setae/spatulae but also on the level of setal arrays (pads) and the foot (tarsus) as a whole. It is still unclear what contribution each of these levels makes to directional dependence and to attachment and detachment during locomotion.

The fibrillar systems of beetles are similar to those of geckos in many aspects. Although several other hair types exist, spatula-tipped setae represent the prevalent design (Stork,1980) and may well exhibit similar properties to gecko setae. Tarsal movements involved in attachment and detachment have been recorded in flies (Niederegger and Gorb,2003). However, no observations have been made on the dynamic changes of adhesive contact area in any fibrillar system and the presence of a directional-dependence has yet to be confirmed for insects with hairy adhesive pads.

The aim of this study was therefore to compare the performance of smooth and fibrillar systems in insects and to clarify their mechanisms of direction-dependence. By measuring frictional and adhesive forces in two model organisms, the leaf beetle G. viridula and the stick insect C. morosus, we address the following questions: (1) how do smooth and hairy systems compare in terms of their adhesive and frictional performance, (2) how does the fluid pad secretion influence attachment in fibrillar systems, (3)are forces in smooth and hairy pads direction-dependent and (4) what is the mechanism for this direction-dependence, if present?

As the material properties of arthropod cuticle are strongly dependent on its state of hydration (Jiao et al.,2000; Vincent and Wegst,2004), the functional properties of insect adhesive pads can only be investigated in vivo. As such, live adult stick insects, C. morosus, and male dock beetles, G. viridula, were taken from laboratory colonies, weighed and mounted on glass cylinders. Stick insects(body mass 898±31 mg, mean ± s.e.m.) were enclosed inside the glass cylinder so that a single leg protruded. We isolated the leg using Plasticine™ and secured it to a firm piece of wire (attached to the glass cylinder) using dental cement (ESPE Protemp II, 3M, St. Paul, MN, USA). The claw tips were clipped with micro-scissors to prevent them from obstructing the arolium. Beetles (body mass 10.4±0.4 mg, mean ±s.e.m.) were immobilised by enclosing them in Blu Tack and Parafilm tape, with the Blu Tack used to isolate the leg. Front and rear legs were used.

The tarsus of C. morosus has five segments, the first four of which bear euplantulae on the ventral side. The leg terminates with a pretarsus that bears the claws and the smooth adhesive arolium (see Fig. 1D–F). The tarsus of G. viridula consists of five segments and a distal pretarsus bearing the claws. The fourth segment is reduced and sunken into the larger third tarsomere. The ventral sides of the first three tarsomeres are densely covered by adhesive setae. Setae are typically curved and oriented distally and belong to three principal types: (a) pointed, with a tapered end, (b) flat,spatula-tipped and (c) disk-tipped, with a marginal bulge. Due to sex-specific variation in the abundance of the different types of setae, only male beetles were used in this study. The distal-most pad on the third tarsomere(Fig. 1A–C) was used throughout this study for force measurements.

To investigate the role of both the tarsal chain and the adhesive pads, two restraining conditions were used for the legs; either the `footloose'condition, where the legs were fixed only up to the tibia, or the`immobilised' condition, where the legs were fully encased (including the dorsal side of the pretarsus in Carausius and the dorsal side of the tarsus in Gastrophysa), leaving only the pad free(Fig. 2). Additionally, for G. viridula, lateral instabilities became apparent during footloose slides, causing the pad to rotate significantly. Therefore, further observations were taken whereby the tarsus was fixed laterally, as for the immobilised condition, but left movable in the dorsal–ventral direction.

Following a previous study (Drechsler and Federle, 2006), a force transducer setup was used to measure friction forces of the pad while simultaneously recording contact area(Fig. 3). Forces were measured with a self-built, two-dimensional force transducer employing 350Ω foil strain gauges (1-LY13-3/350, Vishay, Malvern, PA, USA) and fixed to a three-dimensional DC motor stage (M-126PD, Physik Instrumente, Karlsruhe,Germany). The force transducer was calibrated with calibration weights and by applying defined displacements to obtain the spring constant at different lever arm lengths. The stage was controlled with custom-made LabVIEW (National Instruments, Austin, TX, USA) software that allowed a precise set of user-defined movement patterns. Voltage output was amplified(ME-Meßsysteme, Henningsdorf, Germany) and sampled at 1000 Hz with an I/O board (PCI-6035E, National Instruments). The LabVIEW programme included a normal force feedback mechanism that allowed friction experiments to be performed while keeping the normal force constant. The force feedback mechanism consists of a 50Hz feedback loop, in which the programme computes the deviation between a set-point force and the actual force and passes this on to a discrete PID control algorithm to compute a displacement, which would compensate the error. The distal-most footpad was brought into contact with a glass plate (18 mm×18 mm×0.1 mm) attached to the strain-gauge transducer. Contact area was visualised using a coaxially illuminated stereomicroscope, which shows actual contact as a high-contrast silhouette(Federle et al., 2002). Images were recorded using either a Redlake PCI 1000 B/W camera (for smooth pads)(Tallahassee, FL, USA) or a high-speed digital HotShot PCI 1280 B/W camera(NAC image technology, Simi Valley, CA, USA) (allowing the higher resolution necessary to image hairy pads) and were analysed with MATLAB (The Mathworks,Natick, MA, USA) scripts. For the hairy pads, a `projected' pad area was also measured by manually plotting a solid polygon around the outermost setae in contact to allow the basic frictional force per total pad area to be compared between hairy and smooth systems.

For both animals, proximal friction slides (corresponding to a pull of the leg towards the body) were performed at 500 μ ms^–1 over 10 mm. The relatively large sliding distance was chosen to ensure that the pads were sliding at the same velocities and to be able to test the effects of fluid accumulation and depletion. Distal friction slides (corresponding to a push of the leg away from the body) were done in the same way but were preceded by a short, 0.5 mm proximal slide. This was done because previous studies (Autumn et al., 2000; Gravish et al., 2008) and preliminary observations had suggested that a proximal movement following contact was beneficial in aligning the footpads and ensuring proper contact. For the beetles, the normal force feedback kept the load constant at 0.1 mN during the slide, corresponding to 98% of the body weight of the beetle and to a load stress (force per projected contact area) of 1.7 kPa. This was raised to 1 mN for the stick insects (corresponding to 11% of the body weight and a load stress of 9.8 kPa) to achieve a compromise between a comparable fraction of the insect's body weight and a comparable load stress. Our results show that this difference of normal forces and load stresses has a negligible effect on friction and shear stresses (see below). Otherwise, conditions were kept identical for both insects during all slides. For the footloose condition, no feedback was used during the slide because otherwise the flexibility of the tarsal chain caused the leg to bend and bring other pads into contact. Adhesion area was not recorded for the distal slides as, in most cases, the visible contact area at the end of the slide had dropped below a range that could produce meaningful, noise-resistant results. Similarly,projected area was not calculated for distal slides as the pad outline was often small and irregular.

To investigate the function of the pad secretion in the hairy system (and to control for its effects), repeated proximal slides were performed as above for the immobilised dock beetle. Nine consecutive slides (separated by a 3 s pause following each pull-off) were carried out either on the same area of the glass plate (intended to allow the fluid to build up) or on a fresh area(intended to allow the fluid to deplete). The fluid accumulation on the glass substrate was visualised using Interference Reflexion Microscopy (λ=546 nm, ×20 magnification, Leica DRM, Wetzlar, Germany). Consecutive slides were statistically analysed using Page's non-parametric L test(Page, 1963), where the indices L_m,n indicate the number of conditions(m) and the sample size (n).

Due to the considerable influence of the amount of secretion on friction forces, all `immobilised' slide movements were repeated following two regimes:(1) `little secretion' – each slide was performed on a clean area of glass plate and (2) `accumulated secretion' – four consecutive, proximal slides were performed first on the same area of the glass plate to allow the pad secretion to build up. We recorded forces in both conditions in order to reduce variation caused by variable amounts of secretion present in the contact zone. At the end of every slide, a 5 s pause was left to allow friction to drop before performing a 500 μms^–1perpendicular pull-off. This allowed adhesion forces to be measured.

To test the effect of applied normal force on sliding friction and shear stress, we measured forces for G. viridula and compared this with already published, analogous measurements for stick insects(Drechsler and Federle, 2006). Proximal and distal slides (little secretion) were performed as above but with the applied normal force varied at 0.1, 1.0 and 5.0 mN.

To measure frictional and adhesive forces in pads of stick insects (C. morosus) and beetles (G. viridula), we performed long-distance sliding movements of adhesive pads on glass. When pads were sliding in the proximal direction, friction and shear stress increased steadily in the course of each slide and slowly tended to reach a plateau after approximately 15 s(see Fig. 4). However, when moved in the distal direction, pad forces increased only briefly and then remained at a very low level (Fig. 4). Pads started to slide within 1 s of the beginning of the motor movement; there was no friction force peak at the onset of sliding.

To evaluate the effect of the amount of adhesive secretion in the hairy system, we measured friction and adhesion during proximal slides under two different regimes in Gastrophysa. When slides were repeated on the same area of glass (fluid build up), friction and adhesion strongly decreased from slide to slide until they approached a plateau (mean shear stress decreasing to 42% from slide one to nine; adhesive stress decreasing to 52%). As contact area remained largely unchanged, this effect was due to a highly significant decrease of both adhesive stress (adhesion per unit area) and shear stress (friction per unit area) (Page's L test, adhesive stress, L_9,5=1310, P<0.001; shear stress, L_10,5=1888, P<0.001)(Fig. 5A,C). On the contrary,repeated slides, each time on a fresh area of glass (fluid depletion), showed no significant drop and in fact showed a significant upwards trend in shear stress (Page's L test, adhesive stress, L_9,5=1138, P=0.396; shear stress, L_10,5=1665, P=0.007)(Fig. 5B,D). The increase in shear stress mainly occurred from the first to the third slide (Page's L test from first to third slide, L_3,5=68, P=0.0089), and subsequent forces remained constant, producing highly reproducible curves between consecutive slides (Page's L test from third to tenth slide, L_8,5=816, P=0.44). This indicates that the amount of fluid in the contact zone was depleted over the first three slides and reached a constant level after that. An example image of the build up of pad secretion is presented in Fig. 6. It can be seen that there are many more fluid droplets deposited on the glass surface in the`accumulated' condition. These droplets represent the persistent, hydrophobic component of adhesive secretion (Federle et al., 2002). Fluid build up and fluid depletion have a similar effect in C. morosus (see Table 1) (Drechsler and Federle,2006), and the results presented in the current study demonstrate that this is equally present and conspicuous in the fibrillar pads of Gastrophysa.

The effect of normal force was investigated in G. viridula by performing proximal slides at three different applied forces (0.1, 1.0 and 5.0 mN). Despite a 50-fold variation of load, we did not observe any significant change in friction, contact area or shear stress(Fig. 7). The lack of an increase in contact area with load differs markedly from our previous findings for stick insects (Drechsler and Federle,2006) (see Table 1). This finding is consistent with the morphology of both types of attachment pad (Fig. 1). The seta tips in G. viridula are almost coplanar so that, even at very small normal forces, all setae make contact if the pad is properly aligned with the substrate. By contrast, the arolium of C. morosus is hemispherical, resulting in an increase of contact area with load as predicted by the JKR theory (Johnson et al.,1971). However, in both G. viridula and C. morosus, shear stress was independent of load, confirming that friction forces are fully determined by contact area in both systems(Table 1). As a consequence,the comparison of shear stress between the smooth and hairy adhesive systems in this study is not affected by changes to normal force.

We quantified the effect of sliding direction in smooth and hairy systems by performing proximal and distal slides in a randomised order. We evaluated both the maximum friction during the slide and the adhesion force peak during the pull-off at the end of each slide. As friction forces of beetles and stick insects were strongly dependent on the amount of fluid present in the contact zone [Carausius (Drechsler and Federle, 2006); Gastrophysa, see above], slides were performed in both the `little' secretion and the `accumulated' secretion regimes (see Fig. 8). In both animals and both conditions, maximum friction was significantly lower in the distal direction (Table 2; Fig. 8). However, frictional forces decreased more strongly in the hairy system (mean little secretion,7.8-fold in G. viridula vs 2.3-fold in C. morosus; mean accumulated secretion, 3.4-fold in G. viridula vs 2.7-fold in C. morosus).

Contact area was measured simultaneously, allowing friction forces to be normalised for area (see Table 2). In both systems, contact area visibly decreased during distal pushes (most differences were significant; Table 2). The contact area of the hairy system showed conspicuous changes during the slides. During proximal slides, adhesive contact area was maximal and any hair tips not already in contact after the initial preload were brought into full contact at the start of the slide. However, during distal slides, hair tips appeared to peel off individually and decreased in contact area(Fig. 9). This resulted in the distal slides taking place with only what appeared to be the setal tips in contact. The tips appeared to remain in contact with the surface, the hairs bending or tilting to allow this. The deflection of setae was manifested visually by a movement of the contact zones relative to the pad. This displacement was measured by comparing the positions of the distal end of the contact zone during the short proximal and subsequent distal slides (measured at the force peaks) using a MATLAB script. Excluding three slides where the movement of the distal edge was difficult to track reliably, the hair contacts moved by 68.1±4.4 μm in the proximal–distal direction (little secretion, mean ± s.e.m.). This is greater than the length of the setae(40–50 μm) (Orso et al.,2006) and corresponds to a large bending or rotation of the hairs from being distally to being proximally angled.

Shear stress was computed using real observed contact area. For stick insects, shear stress showed no significant change between proximal and distal slides (Table 2; Fig. 8). This shows that the higher friction forces in the proximal direction are not explained by shear stress but by an increase in contact area. However, for the beetles, shear stresses were higher in the proximal direction (significant difference in the little secretion regime; see Table 2). Thus, the frictional direction dependence in Gastrophysa is not only based on a higher contact area but also on an increased shear stress during proximal slides.

Similar to the friction forces, adhesion was much smaller after a distal slide in both smooth and hairy systems (differences were highly significant; Table 2). This clearly confirms that adhesion is strongly influenced by shear forces towards or away from the body, thus providing a way of controlling attachment and detachment.

It should be noted that despite the 3 s stop of the motor movement at the end of each slide, there was still a significant shear force present during the pull-off movement. Thus, the effective force vector was not perpendicular to the surface but the mean detachment angles were: C. morosus20.61±2.64 deg. proximal, 168.44±3.91 deg. distal; G. viridula 32.36±2.76 deg. proximal, 114.52±4.80 deg. distal(all presented for little secretion).

For both animals, the frictional force generated by a single pad in the proximal direction was more than sufficient to support the body mass of the animal. However, this was not the case for the distal adhesion of C. morosus (mean body weight, C. morosus, 8.81±0.30 mN; G. viridula, 0.102±0.004 mN).

Footloose slides (where the tarsal chain was left free to move) were performed both proximally and distally. This was done in order to investigate whether and how the flexible tarsus contributes to the observed direction-dependence of friction forces. Slides were performed in the same way as for the immobilised condition, with the exception that the feedback had to be left out to prevent the proximal tarsus or the tibia from touching the substrate. We therefore performed footloose trials with a constant z-position of the motor after an initial force feedback preload. During a proximal pull, good contact was made and was maintained throughout the slide in both systems (Fig. 10A,C). However, during a distal push, tarsal instabilities were apparent. For C. morosus, a distal push caused the tarsus to buckle upwards, thereby peeling off the arolium from the proximal side and detaching it rapidly from the surface (Fig. 10B). For G. viridula, the lateral flexibility of the tarsal chain caused the leg to bend, mainly laterally, rotating the foot by almost 180 deg. and preventing it from detaching(Fig. 10D). However, this behaviour is never observed in freely walking beetles and may be an artefact resulting from the relaxed claw flexor muscle and the fixed tibia. As such,additional observations were made where the beetle was mounted in Blu Tack as for the immobilised condition but with greater freedom in the dorsal–ventral direction. In this condition, a proximal slide showed good contact, as before (Fig. 10E, as imaged from above, given that the side view was obstructed with Blu Tack), whereas a distal push caused the entire pad to peel off from the proximal to the distal side (Fig. 10F). As for C. morosus, this prevented any recording of friction forces in the distal direction. Adjacent setae peeled and detached together and we observed the propagation of `peeling fronts' across the pad contact zone as a whole. However, this propagation was very fast and peeling of individual setae occurred almost simultaneously over large contiguous areas of the pad contact zone. This detachment of the whole pad was apparently caused by a rotation of the tarsal segment within the sagittal plane due to the torque introduced by the distal push.

We used the data for proximal slides presented above to compare the performance between smooth and hairy pads(Table 3). Shear stress was approximately 1.7× greater in G. viridula when calculated as force per unit real contact area of the hairs (significant difference; Table 3). For the hairy system,a projected area of the pad was also measured to allow comparisons with smooth pads in terms of force per available pad area. Projected area was approximately 2.6× larger than spatula contact area (little secretion),corresponding to an area fraction of the seta tips of 38%. When shear stress was calculated from the projected pad area (see Table 3), it was 1.5×lower in the fibrillar system (not significant; Table 3). This is similarly true for adhesive stress, which was 1.9× higher in the hairy system for real contact (significant difference; Table 3) but was 1.3× lower for projected contact (not significant; Table 3).

While obviously limited to only one representative of each adhesive system,insight can be gained through a comparison of the hairy system of G. viridula with the smooth system of C. morosus. Our findings indicate that in both systems, the fluid secretion plays a similar role. Forces in the smooth and the hairy system decreased when fluid had accumulated over several slides on the same area. Forces, however, increased or stayed constant when fluid was depleted. Our recent study on stick insects showed that this effect is only present on smooth surfaces(Drechsler and Federle, 2006). However, the opposite effect was found on a rough substrate, suggesting that the fluid mainly serves to maximise contact on rough substrates. We believe that this conclusion is also likely to hold for hairy pads of insects, where the contact of relatively large seta tips is supplemented by a fluid secretion.

To compare adhesive and frictional performance between G. viridulaand C. morosus, the most direct contrast is of the shear and adhesive stresses supported by each adhesive. Table 3 shows that stresses were slightly higher in G. viridulawhen calculated from spatula contact area, suggesting that the hairy system may represent a more efficient attachment mechanism. However, it is biologically more relevant to compare forces per projected pad area because this is the area available to the animal for generating adhesion and friction. Shear and adhesive stresses calculated from projected pad area, for the hairy system, were in most cases no longer significantly different and even slightly lower.

This concept has been used to explain differences in contact size and density across a range of animals with different body sizes because larger animals with relatively less available surface area (such as geckos) are expected to require a more effective adhesive system per unit pad area than smaller animals [such as insects (Arzt et al., 2003; Spolenak et al.,2004)]. The fact that fibrillar adhesive pads of lizards are characterised by a much higher contact density than those of beetles has been seen as a confirmation of the `force-scaling' hypothesis(Arzt et al., 2003). However, Eqn 1 predicts the adhesive stress of G. viridula (spatula width approximately 6 μm) to be 30× smaller than that of a Tokay gecko [Gekko gecko, spatula width approximately 0.2 μm (Williams and Peterson, 1982)]. The mean adhesive stress of arrays of gecko setae has been measured as 53±7.6 kPa(Gravish et al., 2008), which is only slightly higher than our values for G. viridula and C. morosus. As the adhesive stress values compared in the present study were measured under different conditions, any conclusions have to be treated with caution. The comparison is also being made between different species and adhesive systems (wet vs dry) and this is an important caveat. However, despite this, these values certainly suggest that gecko pads do not in fact have a much higher efficiency per unit attachment area, in contrast to the prediction from Eqn 1. This conclusion is consistent with recent data on the scaling of adhesive hair dimensions across different taxa (Peattie and Full, 2007), which suggest that differences in seta density are mainly explained by phylogenetic background and the presence or absence of an adhesive fluid (as is the case here for the dry gecko and the wet insect system) rather than by force scaling. Eqn 1 is also inconsistent with the almost identical adhesive stress in a smooth (C. morosus) and a hairy system (G. viridula),as found in the present study.

A possible explanation is that the assumptions of the force-scaling model do not hold for animal adhesive pads. First, load may not be `shared' equally by all the setae of an array. If hairy pads detach from a surface by peeling(as observed during the footloose experiments), stress is concentrated at the edge of the pad so that only a small number of setae contribute to the total force (Hui et al., 2004). In this case, pad pull-off forces would not be correctly predicted by the force-scaling model. Second, the adhesive forces of individual subcontacts might not scale with their width or radius but might scale with contact area,removing any scaling advantage from seta miniaturisation(Gao and Yao, 2004; Spolenak et al., 2004). This could be achieved, for example, through spatulae that have an optimised concave shape, giving rise to a uniform stress distribution in the adhesive contact zone at pull-off (Gao and Yao,2004; Spolenak et al.,2004). In fact, adhesive setae in several insects (including G. viridula) are known to have concave spatulae(Haas and Gorb, 2004; Langer et al., 2004).

If the efficiency of gecko and beetle pads is indeed of a similar magnitude, it is unclear how geckos compensate the size-related loss of mass-specific adhesion. Assuming isometry, the surface-to-volume ratio of G. gecko can be estimated to be approximately 16× smaller than that of G. viridula [body masses, 10.4 mg vs 43.4 g(Irschick et al., 1996)]. Geckos may partly compensate for this through disproportionately larger adhesive pads [estimated total pad areas, 0.47 mm²vs227.1 mm² (Irschick et al.,1996)]. However, given that the pull-off forces of G. viridula are extreme, with the force of a single pad on a smooth surface corresponding to more than 10× the body weight of the beetle(Table 2), geckos may simply have a smaller `safety factor' and still adhere perfectly well.

The above arguments apply only to smooth substrates. Most biologically relevant substrates, however, possess some degree of surface roughness. Theory predicts that fibrillar systems, and in particular arrays of branched setae with fine endings as found in geckos, should make better contact to rough substrates (Persson, 2003; Persson and Gorb, 2003). However, it still remains to be investigated experimentally whether the performance on rough substrates differs between smooth and wet or dry fibrillar pads.

In order to consolidate fast running with effective attachment, an adhesive system must allow rapid and energy-efficient detachments. Both smooth and hairy adhesive pads of insects possess this ability. The direction-dependence of adhesive and frictional forces is probably a key adaptation for the dynamic control of surface attachment. Our findings demonstrate that smooth and hairy systems both showed this anisotropy when comparing proximal (a pull of the leg towards the body) and distal (a push of the leg from the body) slides. Distal friction forces were always much lower in both little and accumulated secretion regimes and showed a significant decrease in contact area. Adhesion forces were also greatly reduced following a distal slide and, for the conditions used, demonstrate an increased ease of detachment.

Therefore, the mechanism for this direction-dependence makes for an interesting contrast. For the smooth pads, although forces decreased when comparing proximal to distal slides, the contact area also decreased. This resulted in no significant change in shear stress (in either little or accumulated secretion) and argues against a change in the inherent pad efficiency. Thus, a drop in contact area due to the flexibility of the pad is the only explanation for the direction-dependence, consistent with findings in the smooth pads of cockroaches (Clemente and Federle, 2008). However, in the hairy system of the beetles,the higher friction in the proximal direction was not only due to a strongly increased area of contact but also due to a higher shear stress. This suggests that the `quality' of the adhesive contact differed between proximal and distal sliding.

Fig. 9 shows that the changes in contact area occurred at the level of each individual hair. During the proximal pull (Fig. 9A),all hairs made good contact with the surface, and a high resultant force was observed. However, during the start of a distal push(Fig. 9B–E), the spatulae of each hair began to lose contact. They appeared to peel from the surface and remain with this small fraction of contact during the slide. The hairs are typically angled in the distal direction(Beutel and Gorb, 2001) and,as such, the resulting steeper peel angle may aid detachment during a distal push, allowing individual hairs to peel from the proximal side. By contrast, a proximal pull would put the hairs into tension, the shallow angle acting against contact peeling (Autumn et al.,2006a; Federle,2006) (see Fig. 11). For immobilised pads, the fibrillar design had a more pronounced direction-dependence, with a 7.8-fold drop in friction, a considerable decrease compared with the 2.3-fold drop for the smooth system(little secretion). Unlike the smooth system, which can detach at just one peel edge, the hairy system in beetles has several hundred contacts that can peel independently and almost simultaneously, which may aid rapid detachment.

The observed decrease in shear stress may be partly a result of overestimating the area of the seta tips that is in close contact. This could arise from fluid-filled, near-contact being observed in the coaxial illumination and being included in the contact area calculation (see Fig. 11). This idea gains support when taking into account the calculated change in contact area for both secretion regimes. The change for accumulated secretion was considerably less than the corresponding drop for little secretion (a 1.5× decrease compared with a 3.7× decrease), implying that the increased presence of fluid may well contribute to the measured area.

The footloose slides of the stick insect showed that the tarsus itself contributes to a loss of contact area in the distal direction through buckling. Instabilities in the tarsal chain rapidly increase the angle between the leg and the surface, allowing the pad to peel from the proximal side. This adds to previous observations of the same effect in cockroaches(Clemente and Federle, 2008)and implies that, due to its construction, the tarsus, and to some degree the pad itself, is unstable when pushing. The foot as a whole also contributed to pad detachment in the beetles, adding to the direction-dependence of individual setae. This was less clearly demonstrated by the original footloose experiments but further observations made with a semi-restrained tarsus showed a similar proximal–distal peeling detachment as observed for the footloose stick insects.

Our results show that adhesion strongly depends on the sliding direction before pull-off. When pull-off and proximal/distal shear forces act simultaneously, the effect is very similar. Recent work on geckos has shown the presence of a critical detachment angle at the level of single setae,arrays of setae and the whole toe (Autumn and Peattie, 2002; Autumn et al., 2006a; Gravish et al.,2008). Detachment occurs as soon as the angle of the force vector exceeds the critical angle. As some proximal shear is required for the setae to adhere, this effect has been termed `frictional adhesion'(Autumn et al., 2006a). Direction-dependence of adhesive structures is the precondition for controlling attachment and detachment via the amount of proximal or distal shear force. The present results suggest that both C. morosusand G. viridula might control adhesion in a similar way as the gecko. However, further work is needed to determine the detailed angle dependence of adhesion in both systems.

The directional behaviour of the hair tips represents a specialised passive mechanism to control adhesion. Biomimetic directional adhesives have many possible applications and first prototypes have already been fabricated(Autumn et al., 2006a; Lee et al., 2008; Schubert et al., 2008). However, before design principles can be effectively transferred into technical applications, a greater understanding is needed of their precise function in the natural systems. Our study illustrates that there is still much to investigate about both smooth and fibrillar adhesive systems in animals.

We wish to thank Andreas Eckart and Filip Szufnarowski for their help in the development of the LabVIEW motor control programmes. This study was funded by research grants of the Deutsche Forschungsgemeinschaft (Emmy-Noether Fellowship FE 547/1 to W.F.), the UK Biotechnology and Biological Sciences Research Council and the Cambridge Isaac Newton Trust.