Predatory behaviour of some Central European pselaphine beetles (Coleoptera: Staphylinidae: Pselaphinae) with descriptions of relevant morphological features of their heads

The Pselaphinae is a large subfamily of staphylinid beetles with a characteristic habitus and small body size. Detailed morphological and behavioural studies on these beetles are scarce. In this study, specimens of Bryaxis puncticollis (Denny, 1825), Bryaxis bulbifer (Reichenbach, 1816), Bythinus burrelli (Denny, 1825), Brachygluta fossulata (Reichenbach, 1816), Rybaxis longicornis (Leach, 1817), Pselaphus heisei (Herbst, 1792) and Tyrus mucronatus (Panzer, 1803), all collected in Northern Germany, have been examined with regard to their sensory organs (eyes and antennae), mouthparts and method of capturing prey. Scanning electron microscope studies revealed sex-specific differences in the numbers of ommatidia in Bryaxis puncticollis. A multitude of different sensilla on the antennae and great differences in the shape of the mouthparts were observed and peculiarities of the antennae and maxillary palps (e.g., the segment-like appendage) were examined using scanning and transmission electron microscopy. The prey-capture behaviour of these species is described in detail for the first time based on laboratory experiments using Heteromurus nitidus (Templeton, 1835) (Collembola) as prey. This behaviour seems to be tribe specific, ranging from simple seizure with the mandibles (e.g., Rybaxis longicornis, tribe Brachyglutini) to the employment of raptorial legs (Tyrus mucronatus, tribe Tyrini). The two Bryaxis species (tribe Bythinini) even employ their apparently sticky maxillary palps to capture prey. The assumption that a viscous secretion is used by these species is supported by the finding of glandular structures in the interior of their maxillary palps. Prey-capture is preceded by a complicated preparation phase in most of the species and followed by a sequence of preyhandling movements that seem to be adapted to restrain prey such as Collembola. In simple prey-choice experiments the beetles of several species preferred small prey, irrespective of their own body size. In these experiments, Bryaxis bulbifer and Brachygluta fossulata were more successful in capturing prey than Bryaxis puncticollis and Pselaphus heisei. This might be related to their different sensory equipment and different methods of capturing prey.


INTRODUCTION
The Pselaphinae is a globally distributed and very diverse subfamily of the beetle family Staphylinidae, comprising 9267 described species (A.F. Newton, pers. comm.). They are small predators (0.5-5.5 mm in body length) with a characteristic appearance. They possess compound eyes with only a few ommatidia. The mouthparts are prognathous, often with prominent maxillary palps and strong mandibles, indicative of a predatory life style. Saprophagy is recorded from some species of the genera Batrisodes, Claviger and Adranes, but not mycoor phytophagy (Thayer, 2005).
Pselaphinae are most species-rich and diverse in the tropics, but also abundant in temperate regions. They usually occur in moist forest leaf litter or debris, or moist mosses at the margin of water bodies. Some pselaphines display unusual biological adaptations (e.g., myrmecophily), which sometimes also involve interesting behavioural features (especially remarkable in Claviger testaceus: Cammaerts, 1991).
Research on pselaphines has mostly focused on their systematics and geographical distribution, whereas studies on their behaviour (De Marzo, 1985, 1988Engelmann, 1956;Poggi, 1990) and ecology (Reichle, 1967) are scarce. Furthermore, there are only a few general morphological studies on this group (e.g., Chandler, 2001;Jeannel, 1950;Sabella et al., 1998;Thayer, 2005). Most of the morphological data on Pselaphinae are general taxonomic morphological descriptions, often with notes on the conspicuous structures used for identification.
The aim of this publication is to extend the knowledge on the behaviour and morphology of Pselaphinae by examining several Central European species. Detailed information on the way that adult pselaphines catch and handle their prey, and comparative descriptions of the morphology of their sensory organs and mouthparts are provided. It was examined whether different speciesspecific strategies or structures are used in this process. Although the assembled information is limited, we attempt to establish correlations between certain aspects of behaviour and particular morphological traits. The observations and conclusions are discussed in the context of earlier observations made by Engelmann (1956).

MATERIAL AND METHODS
Adult male and female beetles were collected in wet habitats mostly around the city of Kiel (Schleswig-Holstein, Northern Germany) by searching the ground litter at the margins of water bodies using a Reitter's sieve or "sifter".
The animals were transferred in small containers (4.5 cm in diameter) with a floor of moist plaster of Paris, and kept at 14-20°C in an outdoor shelter. The insects were maintained on a diet of small collembolans [Heteromurus nitidus (Templeton, 1835)]. The photoperiodic conditions were that of natural day length.
The specimens examined (in all 144) belong to 7 species in 6 genera, representing four tribes in two supertribes ( Table 1). The smallest were Bryaxis puncticollis, Bryaxis bulbifer and Bythinus burrelli, all around 1.3 mm in length (from base of labrum to abdominal tip). These have maxillary palps with a voluminous fourth segment. Brachygluta fossulata (ca. 1.8 mm) and Rybaxis longicornis* (ca. 2.0 mm) have less well developed maxillary palps compared to the previously mentioned species. Tyrus mucronatus, a large pselaphine beetle of about 2.3 mm, which is found beneath bark of dead trees, has comparatively small maxillary palps. Finally, Pselaphus heisei (1.8 mm long) has extremely long maxillary palps with a slender and terminally clubbed fourth segment. These species will be referred to by abbreviations in the following text: Bp = Bryaxis puncticollis, Bb = Bryaxis bulbifer, Bf = Brachygluta fossulata, Rl = Rybaxis longicornis, Ph = Pselaphus heisei, Tm = Tyrus mucronatus.

Prey-capture behaviour
Observations in the laboratory: The beetles were placed in a cuvette or a small labyrinth of plaster of Paris and were presented with different-sized specimens of Heteromurus nitidus. The behavioural responses were filmed at 25 frames/s with conventional digital video cameras (Sony Digital Handycam, DCR-TRV345E, and Sony Handycam, DCR-HC17E) fitted with close-up lenses (hama Nah (Close-Up) +1, +2, and +4). Light was provided by cold-light illuminators (Schott KL105B). Actual data were obtained by single frame analysis of the video recording.
Evaluation of the preferred prey size and the prey-capture success: Beetles that had been starved for two days were placed in small 5 by 5 cm containers with a base of moist plaster of Paris under natural light conditions, along with defined numbers and size classes of Heteromurus nitidus for two hours [presented prey-size classes: (I) 0.6 mm, (II) 0.6-0.9 mm, (III) 0.9-1.2 mm, and (IV) >1.2 mm]. To determine prey-capture success, the absolute number of Collembola caught during these two hours was recorded, and the proportion of the size classes in this catch was used to determine the preferred prey size. In order to avoid any external disturbances, this was conducted as a "blackbox" experiment, i.e., captured Collembola were not replaced. However, since only in a small percentage of these experiments all the Collembola of one size-class were captured, the preference for a specific prey type could be clearly established.
Statistical analysis: If multiple data sets per specimen were available, species-specific grand means were calculated, summarizing the mean values for several individuals. All means and standard deviations refer to these grand means. There was only a single specimen of Tm available for behavioural observations (marked as "one individual" in the tables), so that in this case only the means for this specimen together with their standard deviations are given.
All the statistical analyses were performed with SPSS 11.0 (SPSS Inc., Chicago), generally employing univariate analyses of variance (ANOVA) followed by pairwise comparisons employing Student t-tests with a Bonferroni correction.

Morphology
Sensory organs (i.e., the compound eyes and sensilla on the antennae and maxillary palps) and the mouthparts were examined using scanning and transmission electron microscopy.

890
head and attached at their base to SEM stubs by double-sided sticky carbon tape. The objects were then left to dry in a desiccator. Afterwards, the stubs were sputter-coated with gold (Bal-Tec Sputter Coater SCD 050) and examined in a LEO S 402 SEM.
In order to count the ommatidia and measure their diameters, the heads of the beetles were cut in half longitudinally. Each half was then attached to SEM stubs, so that the eye faced upwards. Additionally, the corneae of some individuals were detached, photographed under a light microscope and measured with the help of the software TPSdig (Rohlf, 2004).
For transmission electron microscopy (TEM), individuals were fixed in glutaraldehyde (2.5% solution in 0.1M cacodylate buffer, pH 7.4) and osmium tetroxide (1% solution in buffer; 2 h), gradually dehydrated in isopropanol and propylene oxide, and embedded in agar (Agar-100 Resin Kit, Plano). Serial semithin and ultra-thin sections were cut with a diamond knife on a Reichert Ultracut S. The semi-thin sections (0.5 µm) were then stained with Richardson solution (Robinson et al., 1985), whereas the ultra-thin sections (60 nm) were stained with lead citrate (Reynolds) and uranyl acetate (Robinson et al., 1985), and examined using a Philips TEM 208 S. Semi-thin sections were examined using an axioscope (Zeiss) and digitally photographed (Zeiss Axiocam).
Body lengths were measured on dead or stunned individuals by using a binocular and a stage micrometer (Wild, Heerbrugg, Switzerland, No. 310345).

Behavioural observations Prey-capture
The prey-capture behaviour consisted of several steps: (i) searching behaviour, (ii) detection of prey, (iii) approach towards the prey, (iv) prey-capture and finally (v) prey-handling in which the prey is oriented for feeding (species in Table 1). However, this pattern can vary depending on the species, the specific situation and motivation of the individual, with steps omitted or performed in a different way. To facilitate the understanding of the following descriptions pre-defined body angles ( The labyrinth was searched for prey using slight lateral swinging movements of mainly the head (angle d in Fig.  1B; only slightly accompanied by the prothorax, angle f in Fig. 1B) in Bb, Bp, Bf and Rl. Ph beetles did not move their head or prothorax laterally relative to their abdomen. Tm beetles moved the whole body axis laterally, plus additional lateral head movements of about 10° to either side. In Ph and Tm, the movements performed while searching for prey were slower than in the other species (Table 2).
Depending on the species, the antennae are spread at different angles with respect to the median axis (angle e in Fig. 1B). Usually, they are horizontally (angle e in Fig.  1B) and vertically (angle c in Fig. 1A) swung in a pendulum-like manner (Ph moved them only slightly vertically). On account of the slow and often irregular antennal movement in the two Bryaxis, Ph and Tm beetles, it was almost impossible to measure the frequencies of movement. This was only possible in Bf and Rl, which move their antennae much faster (up to 11Hz in Bf and 12.5Hz in Rl, in the vertical direction) in a highly regular manner (Table 2). For the Bryaxis species we can only state that they move their antennae faster than Ph and Tm. Rl and Bf also use their antennae for testing irregularities in the surface of the ground.  The head is bent downward at different angles relative to the longitudinal body axis in a species-specific manner (angle b in Fig. 1A; Table 2).
The maxillary palps are retracted during searching. They may perform only slight movements and may occasionally be protracted. Bp and Tm usually perform rapid vibrating movements with the last palpomere during this time. Specimens of Ph sporadically stop moving and test the ground slowly with their maxillary palps.
All these movements result in different areas being searched ("width of search path" in Table 2). In contrast to the other species, specimens of Ph pursue a slightly meandering path of differing widths.

Prey detection:
The tactile detection of prey takes place with only a slight contact with the antennae (Fig. 2B; regularly observed in Bp, Rl and Tm) or other parts of the body (observed in Rl). Bb, Ph and Bf beetles change their behaviour immediately prior to contact, so that they seem to detect the prey by olfactory or vibrational cues, the latter possibly transmitted via the substrate or the air.
Approach and preparation of predatory strike: Approach: Bp, Bb, Bf and Tm visibly slow down when approaching their prey. Tm even stops for a moment after the first contact with the prey and moves closer to the ground. In contrast Rl specimens rush forward with a velocity of up to 8 body lengths per second and, if necessary, pursue moving prey for 0.12 to 0.24 s (n = 2) at about 5.5 body lengths per second. Rl and Tm may turn towards the prey when it is located lateral to the body axis of the beetle. The prey is generally located between the antennae, with one or even both antennal tips often pointing directly towards the prey. The antennae are now moved more cautiously. Specimens of Bb, Bf and Tm recurrently seem to touch the prey lightly with the tips of their antennae, which might help them to adjust themselves better with respect to the position of the prey. The maxillary palps are usually retracted in Tm, and slightly or more strongly protracted in the other species.
Preparation for the predatory strike (for all species except Tm whose behaviour will be described in the following section): The beetles perform an upwardly directed movement with their fore bodies prior to the strike (angle a in Fig 1A; Fig. 3A). In the two Bryaxis species and Rl the body was moved upwards after having approached the prey. The latter species occasionally made this movement as it rapidly turned towards the prey. Specimens of Bf raise their fore body (by about 10°) after detecting prey, hold this posture as they approach the prey and complete the movement while making the last step. Ph always performs this upward movement while approaching the prey, and thus needs an extra second for preparation, which is about three to five times as long as for the other species. In both Bryaxis species the hind legs are often rapidly extended in order to perform a jump-like forward movement. The different angles of both the body and head are listed in Table 3. The head is brought into a more prognathous position immediately prior to the strike in all the species. The antennae are usually held diagonally forward and approximately parallel to the ground; only Bp ( Fig. 3A) and sometimes Rl hold their antennae  perpendicularly up. As the beetles raise their body, the maxillary palps are either pushed continuously forwards or are, initially, only slowly extended and finally thrust forwards (occurring in Bb and sometimes in Bp). At the end of the upward body movement, the maxillary palps are completely extended. In Rl and sometimes Bf, the palpal extension might in some cases occur only after the following downward movement of the body (predatory strike). Observations on Bf reveal that the beetles open their mandibles widely at the end of the upward body movement.

Tyrus mucronatus (one individual) Pselaphus heisei Rybaxis longicornis
Brachygluta fossulata Bryaxis bulbifer Bryaxis puncticollis TABLE 2. Searching behaviour of Pselaphinae. Searching velocities, the head angles (mean angles of deflection of the head horizontally and vertically, Fig. 1), the positions and movement of the antennae (average maximum and minimum angles with respect to the median axis during movement, and frequency), and the resulting widths of the search path. The numbers represent grand means [± standard deviations (SD)] and, in some cases, statistical ranges (labelled "range"). In Tyrus mucronatus (one individual), the statistics refer to repeated observations on one individual. "n" refers to the number of individuals, except in Tyrus mucronatus, where it is the number of observations on one individual. the beetle to locate the prey precisely and causes the prey to move gradually towards its mouthparts, often exactly below the widely opened mandibles (Figs 4B-C). If the prey has, however, moved on and is no longer within the reach of the beetle, the antennal inward and downward movement is continued, until an angle of about 13° to the median line (angle e in Fig. 1B) and a vertical angle of -53° with respect to a horizontal line parallel to the ground (angle c in Fig. 1A) is reached. Otherwise it stops once the prey is immediately below the widely open mandibles. The first complete antennal inward movement without contacting or driving the prey below the mandibles takes about 2.2 ± 0.2 s (n = 3, measured on one individual). The movement can then be repeated several times (often over only a smaller part of the area), and the time needed for these repetitions is very variable. If the prey is already located below the mouthparts, the antennae are not moved inwards.
In conclusion, the main difference between Tm and the other species in the preparation for the strike is that Tm beetles actively drive the prey towards their mandibles, whereas the other beetles move towards the prey.
Predatory strike: The predatory strike consists of a rapid downward movement of the fore body (angle a in Fig. 1A) in most of the species, and the prey is captured either with the mandibles, by means of the apparently sticky maxillary palps, or, in the case of Tm, by the fore and middle legs. In addition to the downward movement, Bp, Bf and Ph beetles bend their heads downwards with respect to the pronotum (angle b in Fig. 1A). Table 3 lists the different periods of time recorded for different species for this downward movement. Specimens of all the species, with the exception of Tm, stretched out their maxillary palps and placed them upon the dorsal surface of the prey (Figs 3B-C). In the two Bryaxis species, this seems to be sufficient to restrain the prey, although Bp beetles also sometimes additionally use their mandibles. Ph has an additional second phase of downward movement, during which they lower their opened mandibles towards the prey, which is Maximum: 30.0 ± 6 (n = 5); shortly before strike: 10.7 ± 4 (n = 7) 17.5 ± 3 (n = 3) 26.9 ± 1 (n = 3) 19.3 ± 10 (n = 6) 21.6 ± 7 (n = 3) 26.0 (n = 2) Body angle (°) (angle a in Fig.  1A)

Rybaxis longicornis
Brachygluta fossulata Bryaxis bulbifer Bryaxis puncticollis TABLE 3. Prey-capture behaviour of Pselaphinae. Body postures at the end of the upward movement prior to the strike (body angle a and head angle b in Fig. 1A) and time needed for the subsequent downward movement during the strike (Figs 3 B-C). The numbers usually represent grand means [± standard deviations (SD)] and statistical ranges (time only, in brackets). For Tyrus mucronatus (one individual), the statistics presented refer to repeated observations on only one individual (the maximum value is the maximum angle, as due to subsequent additional forward corrections of its body posture the previous angle is diminished shortly before the strike; for further information see text). "n" refers to the number of individuals, except for Tyrus mucronatus, where it is the number of observations on one individual. 115 ± 9 (n = 3) 100 (n = 1) (see above) 108 ± 18 (n = 5) 94 ± 12 (n = 3) 100 (n = 1) Head angle (°) (while prey is held sandwiched with the forelegs; angle b in Fig. 1A) 6.22 ± 3.9 (n = 5) 1.04 (n = 1) 1.53 ± 0.6 (n = 3, prey only turned, not sandwiched)  finally seized and pressed on the ground. The prey sometimes gets stuck to their maxillary palps and can, in this way, be lifted towards the mandibles or escape is prevented to a certain degree. Rl and Bf beetles also place their maxillary palps on the dorsal surface of the prey (Fig. 3C), even though there is no clear indication of a sticky surface of the palps. These beetles push the prey down and then seize it with their mandibles. In contrast to Ph, beetles of these species perform the whole downward movement in one quick action. The maxillary palps of Rl are sometimes placed on either side of the prey, possibly to hold it in place.
Tm again differs in its method of capturing prey. At the beginning of the downward movement of the fore body, the maxillary palps may be widely extended for a short time. Thereafter, the pro-and mesothoracic tibiae and tarsi are pulled rapidly inwards, and the beetle crouches down. Similar to the predatory strike of a mantis, the prey is seized between the tibia and the trochantero-femoral spines and ridges of the fore legs and, to a lesser extent, the middle legs. Lastly, the head is bent downwards and the prey may also be seized by the mandibles.
In the case of an unsuccessful predatory strike, the two Bryaxis species, Bf and Rl often search the ground near their heads using their maxillary palps. They may even bite the plaster with their mandibles. Ph beetles show a special searching pattern, which resembles the behaviour of Tm when preparing to strike. If after the downward movement the maxillary palps or the mandibles do not contact the prey, both the maxillary palps and the 895 Fig. 6. Preferred size of the collembolan prey (Heteromurus nitidus). Proportion of each of four prey sizes in the total amount of prey caught (points = grand means). The error bars mark the standard error (SE). (Size classes: I = 0.6 mm, II = 0.6-0.9 mm, III = 0.9-1.2 mm, IV = > 1.2 mm. Sample sizes: Bryaxis puncticollis n = 29, Bryaxis bulbifer n = 8, Brachygluta fossulata n = 18, Pselaphus heisei n = 9). Different small letters above the error bars indicate statistically significant interspecific differences (Student t-test with Bonferroni correction; p < 0.05). antennal tips are swept over the ground, distally to proximally. If this is not successful, it may be repeated several times. Each cycle takes about 1.5 s. This behaviour may locate nearby prey, often causing it to move towards the predator. Finally, the maxillary palps are placed on the dorsal surface of the prey, so that the second phase of the downward movement (as described above) can take place, ending in prey capture.
Prey handling and final seizure: Immediately subsequent to the strike individuals of all the species except Tm raise their fore bodies in order to lift the prey off the ground (Fig. 5A). While doing this, they manipulate the prey with their fore legs (tibiae and tarsi), while the middle and hind legs ensure a firm stance. The different periods of time spent in this position are given in Table 4. The struggling prey may generate abnormally steep angles of the beetle's body with respect to the ground (generally, a maximum angle of 70° is observed, although in Bp it can be 105°; see Table 4 for means). Sometimes the beetle is even knocked over. The prey is held in the mandibles (Bf, Rl, and Ph) or stuck to at least one maxillary palpus in Bryaxis spp. (sometimes also in Ph). Even when the prey was seized using the mandibles, the maxillary palps of Bf and Rl sometimes remain in contact with the prey. The antennae are extended anteriorly in Bb and Bf, and seem to form a mechanical barrier preventing the escape of the prey.
Most species drew the prey towards their ventral side and sandwiched it between both the tibiae and the femora of the fore and often the middle legs (depending on prey size), at the same time lowering the anterior part of their body (Fig. 5B). They now take a posture similar to that adopted by Tm beetles after their strike. All species, including Tm, continue in approximately the same manner. Further adjustments of the prey are often necessary. To attain this the head is usually bent by about 100° (  Table 4) relative to the longitudinal body axis. Finally a permanent grasp with the mandibles is achieved. The grasp of the legs is then relaxed and the prey lifted up in the mandibles (see Table 4 for the specific durations of the period the prey is held with the legs).
Rl beetles often omit the first part of the prey-handling behaviour sequence (i.e., the upright posture) and even reduce the second part (prey held with the legs), just turning but not grasping the prey with the forelegs. The mandibular grasp is supported by pressing the prey to the ground with the mandibles. In Bp passive prey might only be manipulated using the maxillary palps and the fore legs without sandwiching it. In this case the maxillary palps are only moved away after achieving a firm hold with the mandibles.
Finally, all the beetles usually hold the collembolan so that its extended furca points upwards (Fig. 5C), irrespective of the initial orientation of the prey. This prevents the prey from harming the predator and makes it more difficult for the prey to escape. Some beetles do not show a specific prey-handling behaviour if the prey is exceptionally passive (Bp) or small (Rl, see above), and as a result the furca of the collembolan does not necessarily point upwards.  In all species the maxillary palps are not involved in feeding and are usually held laterally. Even during feeding the beetles might react to other prey animals (observed in Bp and Rl). In most cases the antennae, the maxillary palps, and at least the tarsi of the fore legs are cleaned after feeding. This might also include selfgrooming of other parts of the body.
Preferred prey size and prey-capture success All species tested (both Bryaxis, Bf, and Ph; see Table  1) prefer the smallest and the second smallest (< 0.9 mm) of the prey offered (Fig. 6).
Bb and Bf were more successful predators than Bp and Ph (Fig. 7) with respect to the prey species used. This is again not related to the size of the predator.

Morphology
Mainly Bp, Bb, Bf, Rl, and Ph were used for the morphological studies (Table 1). Tm and Bythinus burrelli were only partially studied (eyes resp. antennae).

Sensory organs
As described earlier, pselaphine beetles seem to rely on tactile and olfactory cues to find prey and orient themselves. The following section presents information on the various sensory organs (eyes, antennae, maxillary palps) likely to be important for prey-finding.

Eyes:
The compound eyes are composed of only a few ommatidia, which are relatively large in size, convex and distinguishable at low magnifications (10×) (Fig. 8A). The range in numbers of ommatidia is small and speciesspecific. The males and females of Bp have significantly different numbers of ommatidia, with 23 in males and 14 in females (Table 5). Such intraspecific differences were not found in any of the other species in this study.
Pselaphines have apposition ommatidia of the acone type (Meyer-Rochow, 1999; Fig. 8B), i.e., without a distinct crystal cone. The corneal lenses (c in Fig. 8B) consist of a multitude of separate layers, which are separated from each other by unmodified cuticle (cu in Fig. 8B).  Pigment granules occur beneath the cornea, appearing as black spherical components that enclose the rhabdome (pg in Fig. 8B). Particularly large granules in the apical regions of the ommatidia (primary pigments, pp in Fig.  8B) occur in the Bryaxis species. Two to three nuclei of Semper cells (sc in Fig. 8A-B) occur immediately below the cornea.
The rhabdome is closed (i.e., the rhabdomeres are fused), and consists of eight rhabdomeres (Fig. 9). A wedge-shaped rhabdomere (labelled 2 in Fig. 9Ba) is present in the distal part of the rhabdome, which extends in a proximal direction and is surrounded by two other rhabdomeres (labelled 1 and 3 in Fig. 9Bb). The basal part contains only three rhabdomeres, two of which occur along the entire length of the rhabdome (labelled 1 and 4 in Fig. 9Bd). The third rhabdomere (labelled 8 in Fig.  9Bd) does not extend far distally.  Different symbols represent the different types of sensilla; their sizes and numbers reflect the relative sizes and the proportional quantities of sensilla. Large black circles = large sw-wp sensilla; small black circles = small sw-wp sensilla; small grey circles = dw-wp sensilla; triangles = long projecting sensilla; large grey circle = trichobothrium-like sensillum; small black square = probably simple mechanoreceptive sensilla; long black bar = crescent-shaped aggregation of sw-wp sensilla. Abbreviations: dw-wp = double-walled wall pore sensillum, sw-wp = single-walled wall pore sensillum. Antennae: The pselaphine antennae, which can be described as clubbed in the species studied, are composed of 11 antennomeres: a scape, a pedicel and a flagellum of 9 segments (Fig. 10). The club is composed of the terminal three antennomeres (9-11), and is especially conspicuous in Bryaxis species. The relative length of the antennae varies with body length and is significantly larger in the larger species, i.e., Bf, Rl and Ph. The antennal length can reach 50% of the body length or more (Fig. 11) in these species. The absolute length of the antennae ranges from 0.44 mm (Bp; SD = 0.034, n = 10) to 0.96 mm (Rl; SD = 0.068, n = 8).
SEM studies show that the antennae differ greatly in the number and types of sensilla present on their surface ( Fig. 12 for a representation, Fig. 13A). Several distinct types of sensilla revealed in this study are described below [double-and single-walled wall pore sensilla, following the terminology of Altner (1977) and Steinbrecht (1997)].
Long projecting sensilla (Fig. 15): Originate from a cuticular protuberance. A tubular body is present and the sensillum is much longer than other sensilla. Characteristically it is bent away from the antennal surface and possesses a knob-like ending, visible under the SEM at high magnifications (~70,000×, Fig. 15B). The average lengths vary from about 54 µm (Bp, Bb and Ph) to 77 µm (Bf), or even 85 µm (Rl). There are few of them and the number   was species-specific in our sample, with the largest number (16) in Bp.
Trichobothria-like sensilla (Fig. 16): Long and extremely fine recognizable by the rhombic shape of the cross-section of their hair shafts and small diameter (Fig.  16C) in ultra-thin sections. There are three of them on the tip of the 11th antennomere in the Bryaxis species. They are arranged in a triangle when viewed apically (Fig.  16B). They arise from a cuticular depression on the 11th antennomere (Fig. 16A). Several pores are present at their base, which were not investigated further (Fig. 16A).
Peculiar structures on the antennae (Fig. 17): Stampand ridge-like modifications occur on the lower inner side of both the scape and the pedicel in males of Bb (Figs 17C-E), and a crescent-shaped modification on the pedicel of males of Bythinus burrelli (Figs 17A-B). SEM has revealed the presence of pores (Figs 17A, D, E) along the ridges on the pedicels of both species and on top of the stamp-like structure in Bb (Fig. 17E).
Furthermore, Rl has a distinct broad flattened seta on the tip of its 11th antennomere (Fig. 13D); its function is unknown.

Mouthparts
The following brief descriptions of the mouthparts of pselaphines generally refer to Bp, Bb, Bf and Ph (Table 1).
Clypeo-labrum (Fig. 18): Trapeziform and free, i.e., connected with the clypeus by a membrane. The anterior margin features two medial peg-like sensilla (ps in Fig. 18). Several hair-like sensilla of different lengths can be found on its dorsal and lateral sides: the long lateral sensilla are bent toward the medial axis, so that they come in contact with the prey seized by the mandibles. The epipharynx (investigated in Bf only) has several disc-like sensilla and two rows of medially directed scales or trichomes (Fig. 18E).  Mandibles (Figs 19A-B): More or less symmetrical. The apex is simple (unidentate) and several subapical teeth are present (st in Fig.  19A). A prostheca is lacking. The mola is distinct and firmly united to the rest of the mandible. The mola differs in size from well-developed in the Bryaxis species to reduced in Ph. It has small nipple-like grinding cones (Fig. 19B) on its medial surface. The mandible lengths, measured from the abductor to the tip of the mandible, vary from 0.137 mm in the Bryaxis species (n = 13; SD = 0.01 mm) to 0.167 mm in Bf (n = 3; SD = 0.01 mm) and 0.207 mm in Rl (n = 1). Electron-micrographic images of the Bp mandibles have revealed the presence of a single sensillum on the outer margin (s in Fig. 19A). It is particularly conspicuous and long (up to 70 µm), and laterally flattened. The sensilla present on the mandibles of the other species (partly visible in Figs 18B-D) are inserted on the dorsal side and are less distinct and often much smaller and thinner (Ph has several short sensilla). An interaction with other mouthparts has not been observed, and the sensillum is usually positioned relatively close against the mandible. Its function is unknown.  The cardo is transverse. The stipes is subdivided into basi-and mediostipes, the latter forming the base of both the galea and lacinia. The apices of both are differentiated into mesal brush-like arrays of curved hair-like trichomes (Fig. 19C). The maxillary palps are four-segmented (Fig.  20). The external appearance of the palpomeres is at least genus-specific (Fig. 21). The first palpomere is small (only Ph beetles have a strongly elongated first palpomere), the second relatively long and the third short. The fourth palpomere is generally large and varies greatly in appearance between species (Fig. 21).
Both Bryaxis species seem to capture their prey with the help of their maxillary palps. SEM analyses of their fourth palpomere have shown that the dominant type of setae resembles the adhesive structures that are often found on insect tarsi, with the distal parts being spoonshaped (ssh in Fig. 22). The fourth palpomere of Bp bears on average 356 (female, n = 4, SD = 14) and 441 (male, n = 3, SD = 57), respectively, of these setae. Semi-thin sections of the maxillary palps have revealed the presence of glandular structures (Fig. 23). The outlet was not visible but is possibly situated on the apical part of the spoonshaped hairs, thus supporting the hypothesis that the bee-tles capture their prey by means of the adhesive surface of their palps. Additionally, another type of sensilla was found (lps in Fig. 22), which resemble the long projecting sensilla on the antennae.
A common peculiarity of pselaphine maxillary palps is a segment-like appendage at the tip of the fourth palpomere (Newton & Thayer, 1995;Fig. 24). This appendage measures 25-35 µm in length and ca. 5 µm in diameter, and is inserted either in a cuticular depression or on a cuticular socket (Ph). The distal part of this appendage features longitudinal striae, which only become visible at high magnification (7,500 ×). The apex is oblique (Fig. 24B) and transmission electron micrographs reveal several cuticular cavities and pore tubules in the cuticle on the slanting mesal side (Fig. 24C). The space between two cavities is filled with small, almost rectangular, cuticular structures, which are separated by small interstices (Fig. 24C). The depth of these cavities varies between 400 and 800 nm. Receptor lymph cavities are visible further proximally, as are several dendrites inside the appendage (Fig 23D). These originate in the fourth palpomere and extend into the appendage, thus indicating its sensory function. A tubular body, which is characteristic of mechanosensilla, was not observed. Labium (Fig. 25): The mentum is large and plate-like and separates the maxillae (Fig. 20). The palps are three-segmented, with   the last palpomere sensillum-like (Fig. 25). At its tip the prementum bears prominent apical lobes with a fringe of stout bristles and trichomes (al in Fig. 25). These lobes result from the fusion of the glossa and paraglossa forming a synglossa (Nomura, 1991).

DISCUSSION
All the species investigated in this study are distributed throughout most of Europe (Besuchet, 1974), with all but Tyrus mucronatus living in grass-and leaf-litter, and in moss growing close to waterbodies and bogs. Tyrus mucronatus live under the bark of dead trees, a habitat where only a small number of species of the subfamily are found.
Pselaphines (more than 9000 described species worldwide) are generally predacious as larvae and adults (Chandler, 2001;Thayer, 2005), the commonest feeding mode among staphylinids (Thayer, 2005). Pselaphines were initially reported to feed on earthworms, insect larvae, small flies and especially mites (Park, 1932(Park, , 1933(Park, , 1942(Park, , 1947aPearce, 1957), whereas in later studies Collembola were used as the main prey (De Marzo, 1985, 1988De Marzo & Vit, 1982;Engelmann, 1956). Other potential prey animals (mites, ant larvae and Drosophila larvae) were used in this study, but rejected or only reluctantly ingested. However, all of the beetles eagerly preyed upon Collembola (Entomobryidae, whereas the isotomid Folsomia was rejected). Therefore, the collembolan Heteromurus nitidus was used in the prey-capture experiments. Hence, recorded here is the performance of the different beetle species when attempting to catch a rather elusive prey. They might perform differently on other prey types encountered in situations different from those in our laboratory experiments.
Park (1932, 1933, 1947a) worked especially on those species facultatively associated with ants, focused on their feeding behaviour but did not describe their method of capturing prey, possibly because ant larvae and mites are easy to locate and not elusive. Park (1947b) includes a short note indicating that free-living pselaphines mainly feed on mites, which after capture they press to the ground using their fore tarsi. De Marzo & Vit (1982) report Batrisodes occulatus "trying to catch Collembola with their mandibles", and that fragments of Collembola were found in the intestines. A short account of the preycapture behaviour of Pselaphus parvus larvae is given by De Marzo (1988), who reports that they capture their prey by means of protrusible viscous head organs, with the viscous secretion originating from glands in the head (De Marzo, 1988). According to DeMarzo (1985DeMarzo ( , 1988, this structure is found in all investigated pselaphine larvae; however, it is absent in larvae of Faronitae (Newton, 1991). The most detailed description of the prey-capture in adult Pselaphinae to date is given by Engelmann (1956).

Detection of prey
The pselaphine beetles studied have relatively reduced eyes, with only a few convex ommatidia. The presence of large numbers of screening pigment granules between the ommatidia is found only in scattered taxa among Coleoptera, and within Staphyliniformia only in Micropeplinae and some Scydmenidae (Caveney, 1986). The Semper cells are probably the remains of a crystal cone (cf. Meyer-Rochow, 1999). The reduced eyes and the lack of a crystal cone are indicative of a life in habitats where light intensities are low (Lawrence & Britton, 1991), such as litter and rotting logs. These structures clearly suggest that the beetles are not visual hunters. Nonetheless, they respond to light, possibly even ultraviolet light, since many species in tropical and some in temperate regions are collected at ultraviolet lights (Wolda & Chandler, 1996). Our observations suggest that Brachygluta fossulata, Rybaxis longicornis, and Tyrus mucronatus, which have a higher number of ommatidia than other species, might be able to perceive and follow the movement of their prey visually. The higher number of ommatidia in the males of Bryaxis puncticollis is presumably advantageous when searching for females and a rather common sexual dimorphism known from a number of Pselaphinae (Chandler, 2001;Jeannel, 1950).
In accordance with their poor visual capabilities, tactile and chemical cues seem to play a principal role in prey detection and several types of sensilla may be involved. Of special importance in the search for prey are the antennae, which have a wide range and multitude of sensilla (much more so than on the rest of the body). In the species studied 42-50% of the antennal sensilla were concentrated on the terminal antennomere (Figs 10, 12). The structural peculiarities of the antennae of Bryaxis bulbifer and Bythinus burrelli are obviously related to mating, as they occur only in males. These structures are rarely located on scapus and pedicel in Pselaphinae, but almost always on the flagellomeres (Chandler, 2001). Sensilla described as single-and double-walled wall pore sensilla have an olfactory function (cf. Altner, 1977;Steinbrecht & Gnatzy, 1984) in other insects. The position of the olfactory sensilla at the apex of the antennae allows the assessment of the direction of the source of olfactory cues (Skilbeck & Anderson, 1996), and the constant move- Fig. 25. SEM-micrograph of the right labial palpus of Bryaxis puncticollis. Bar = 10 µm. Abbreviations: al = apical lobe, lp 1-3 = labial palpomere 1-3. ment of the antennae promotes the perception of chemical cues according to phasic receptor physiology (e.g., Zacharuk, 1985). The antennal tips of the Bryaxis species are additionally equipped with specific trichobothria-like sensilla (Fig. 16). They probably function as receptors of air vibrations. Brachygluta fossulata, Rybaxis longicornis and Pselaphus heisei lack this type of sensilla. Notwithstanding, these species were capable of identifying prey, even without mechanical contact.
Like the antennal movements, the maxillary palps, vibrated in Bryaxis puncticollis and Tyrus mucronatus, also seem to be involved in the perception of olfactory cues. This is also described for Batrisodes globosus, Euplectus sp. and Bibloplectus sp. (Engelmann, 1956). Testing the ground with the maxillary palps as observed in Pselaphus heisei can be interpreted in terms of searching for gustatory chemical cues of prey.
The fourth maxillary palpomere has a characteristic appendage (a in Fig. 20; Figs 21, 24) at its distal end, a synapomorphy for the subfamily (Newton & Thayer, 1995). Thayer (2005) describes this appendage of the maxillary palps as an "unsclerotized digitiform segmentlike appendage", whereas Jeannel (1950) and Pearce (1957) consider it to be a fifth palpomere. Others have identified it as an ordinary sensillum. The origin of this structure remains unclear. As it has several cuticular sheaths it appears to be a compound sensillum formed by the fusion of several sensilla. That it has a sensory function can be deduced from the dendrites and dendritic sheaths within the appendage (Fig. 24). The morphology suggests a contact chemoreceptor.
The long projecting sensilla (Fig. 15) on the antennae seem to have a bimodal mechanoreceptive function. They are characterized by their shape (usually S-curved), with the most distal part being bent away from the antennal surface (G. Alberti, pers. comm.). This type of sensilla lacks lateral wall pores. Instead they have a single pore at the apex (cf. Ozaki & Tominaga, 1999), which suggests a gustatory function. We have found similar sensilla on the fourth maxillary palpomere (between those with spoonlike apices, Fig. 22), which might have the same modalities. The many small, thin sensilla on the antennae are probably simple mechanoreceptors for tactile perception. Tactile contact (generally with the antennae) seems to be necessary to induce preparation for the predatory strike in at least Bryaxis puncticollis, Rybaxis longicornis and Tyrus mucronatus.

Predatory strike
All the beetles investigated slightly touch their prey during the preparation phase prior to the strike, probably in order to ascertain its exact location and identify it as prey using gustatory sensilla. These contacts are usually so gentle that they do not initiate the escape mechanism of the Collembola, and therefore the beetles do not have to strike faster than the prey can initiate the escape (in Collembola about 26 ms subsequent to the stimulus, Bauer, 1978).
In all the pselaphines studied there occurred an upward movement of the body prior to the strike (Fig. 3A). This might function to accelerate the strike, so that the maxillary palps of the Bryaxis species hit the prey at a relatively high velocity. If any adhesive secretion is involved, this would result in a better adhesion to the surface of the prey (Betz & Kölsch, 2004). The role of the palps of Pselaphus heisei in prey-capture is not easy to determine, since ultrastructural examinations revealed no evidence of an adhesive surface. Hence, in this case the prey might simply be captured by entanglement among the massive setae on the maxillary palps (Fig. 21E).
The broad conclusion of this limited comparative study is that the prey-capture behaviour seems to be similar within certain tribes of Pselaphinae (classification into tribes according to Löbl &Besuchet, 2004, andNewton &Chandler, 1989). The prey-capture behaviour in all the species studied is substantially different from the relatively simple behaviour described for Batrisodes globosus (tribe Batrisini), Euplectus sp. (tribe Euplectini), and Bibloplectus sp. (tribe Trichonychini) by Engelmann (1956). In these three species the beetles briefly stop their forward movement, then lunge forward and grasp the collembolan with their mouthparts. Of the species investigated in the present study, the behaviour of the members of the Brachyglutini is most similar.
Other behavioural features and structures seem to be more derived compared with this general pattern. (1) The presumably sticky maxillary palps of the Bryaxis species (tribe Bythinini) are a specific prey-capture device not previously described.
(2) The behaviour of Tyrus mucronatus beetles seems to resemble that of Cedius spinosus (both species belonging to the tribe Tyrini) described by Engelmann (1956). The latter species slowly approaches the prey, "steadies itself upon its meso-and metathoracic legs, and rears up slightly" before seizing the prey (Engelmann, 1956). Hence, both of these species capture their prey by means of their forelegs, sandwiching it between the femur and tibia, an action supported by the mouthparts in Cedius spinosus (Engelmann, 1956). Engelmann (1956) observed similar prey-capture behaviour in Tmesiphorus costalis (tribe Tmesiphorini, closely related to Tyrini). All three species have specific morphologically modified forelegs, which supposedly improve their predatory strike, i.e., (i) spines on the femur and tibia in Cedius spinosus, (ii) trochantero-femoral spines and ridges in Tyrus mucronatus and (iii) stiff brushes of setae on the femur and tibia in Tmesiphorus costalis. Such prey-capture behaviour employing raptorial fore legs appears to be rather unique within the Coleoptera and has only recently been described for two other groups of staphylinids, i.e., the genus Philonthus subgenus Onychophilonthus Neresheimer & Wagner (Betz & Mumm, 2001) and Nordus fungicola (Sharp) (Chatzimanolis, 2003).
Complex preparation behaviour prior to the strike and directing the prey towards the mandibles with the help of the antennae was not recorded for Cedius and Tmesiphorus by Engelmann (1956), possibly because of the poor technical facilities available at that time.

Prey-handling
Another previously undescribed behaviour in pselaphines is prey-handling. Only Engelmann (1956) records that a Bibloplectus beetle held down a large isotomid collembolan with its fore tarsi. The specific behaviour described in the present study, which involves the manipulation of the position of the prey in the mouthparts, might be a special adaptation for catching elusive prey such as Collembola.
The maximum prey size manageable by a beetle depends on the prey's combativeness and the predator's ability to hold and subdue it. As far as prey size is concerned, the two Bryaxis species, Brachygluta fossulata and Pselaphus heisei were similar, i.e., they preferred small to large prey (Fig. 6). However, these species differed significantly in their general prey-capture success (Fig. 7). The absolute length of their antennae and their pattern of movement did not seem to be related to preycapture success. Apart from its lower strike velocity, the reason for the lower success in Pselaphus heisei might be the lower number of mechanoreceptors on the antennae, possibly resulting in a less accurate location of the prey and an imprecise strike. Moreover, Pselaphus heisei has relatively short mandibles (ratio mandible length/body length, 0.086), which could also decrease its capture success. The difference in the capture success of the two Bryaxis species might be caused by the different adhesive properties of the maxillary palps.
Based on the behavioural observations on one individual, Tyrus mucronatus seems to be an extremely effective predator. Its complex behaviour prior to the strike permits accurate location of the prey, and its predatory legs allow the strike to be especially precise and successful.

CONCLUSIONS
This study provides new information on the behaviour and morphology of a hitherto largely neglected group, the Pselaphinae. These data improve our knowledge of the prey-capture behaviour of this very diverse group of beetles and should spur further comparative behavioural analyses.
The pselaphines studied differ significantly in the array of sensilla on their sensory organs. In particular, the differences in the numbers and types of sensilla on the antennal tip might correspond to differences in preyfinding success. Pselaphines do not seem to be visual hunters, but rather employ chemical and mechanical cues, which is to be expected of litter-dwelling organisms.
Although they vary in size and have different sized mandibles, all of them prefer the smallest prey-size classes offered. Notwithstanding, the beetles differ in their specific prey-capture strategies. These might be tribe-specific and are highly elaborate in some species, leading to increased prey-capture success, for instance the use of sticky maxillary palps (e.g., Bryaxis bulbifer) or raptorial legs (Tyrus mucronatus), or certain behavioural patterns prior to the strike. Beetles of other species simply use their mandibles, probably the ancestral method of capturing prey.
Because of their great ecological and morphological diversity, this subfamily seems to have great potential for future integrative studies on the evolution of functional and ecological diversity in soil organisms.