Influence of caste and subcaste characteristics on locomotion in the ant Camponotus fellah

Walking in insects is of considerable interest in the study of locomotion (e.g. Hughes, 1952; Wilson, 1966), and ants have become important model organisms, providing a source of new biomechanical findings and inspiration for legged robot design (Dupeyroux et al., 2019; Yeoh and Yi, 2021). In most recent ant studies, workers of similar size were examined – soldiers and especially males were not considered (e.g. Reinhardt et al., 2009; Wahl et al., 2015; Pfeffer et al., 2019). Just a few insect locomotion studies have investigated the influence of body size (Duncan and Crewe, 1993; Morehead and Feener, 1998; Hurlbert et al., 2008; Tross et al., 2021), and in a number of these studies no significant effect of body size was observed (e.g. Lighton et al., 1987; Duncan and Crewe, 1993). However, partly as a result of small sample size and impact of factors such as temperature (Morehead and Feener, 1998; Lighton and Duncan, 2002), body size effects may have been obscured in many of these studies (see Hurlbert et al., 2008). In addition, body shape and allometric relationships between body parts often do not change uniformly with size. Such morphometric differences exist, for example, between castes and worker subcastes in several ant species. Size and shape variations within the worker caste occur in about 13% of ant species and about 16% of ant genera and thus are not uncommon (Hölldobler and Wilson, 1990; Fjerdingstad and Crozier, 2006; Wills et al., 2018). These variations enable division of labour within the colony and are probably under strong selection. Therefore, they should influence colony fitness through colony maintenance, reproduction and survival (Billick, 2002; Powell, 2009; Wills et al., 2018), and might also affect locomotion performance. The present study examined intraspecific differences in walking behaviour in the polymorphic ant species Camponotus fellah under precisely this perspective, as effects of caste and subcaste characteristics have rarely been investigated.

Camponotus fellah has evolved a polymorphic worker caste with subcastes of minor and major workers (Boulay et al., 1999; Laciny et al., 2019) and a spectrum of transitions, often called media workers (medias) (Boulay et al., 2000; Lenoir et al., 2001). The worker subcastes differ significantly in their morphometry (Laciny et al., 2019), showing considerable distinction in mesosoma and head shapes (Fig. 1). In some studies, the category ‘medias’ is seen to be developmentally undefined (Tschinkel, 1988), but it is specified here as a subcaste because of discrete morphological differences compared with the minor and major characteristics (see also Powell and Franks, 2006; Wills et al., 2018). Minors and medias are described as primarily engaged in foraging, brood care and nest cleaning, while majors (often called ‘soldiers’) are characterised as specialised in nest defence (Hölldobler and Wilson, 1990; Laciny et al., 2018, 2019). Contrary to this outline, three worker groups are described in C. fellah by Mersch et al. (2013): nurses, foragers and cleaners; surprisingly, without consistent size differences between groups (Mersch et al., 2013). This division of tasks is relevant for smooth functioning of the colony, whereby division can be the result of polyethism or polymorphism (Espadaler et al., 1990; Galbán et al., 2021). In addition to the workers, the queen and the winged males form the reproductive caste (Hölldobler and Wilson, 1990). The queen is the largest individual, with large eyes and a well-developed gaster and reproductive system (Peeters and Ito, 2015), but was not examined in this study to preserve viability of the colony. Males usually have considerably smaller heads and bodies, fragile legs, and wings (Hölldobler and Wilson, 1990; Laciny et al., 2018, 2019). Their short lifespans associated with mating flights complicate their investigation (Heinze, 2016; Laciny et al., 2019), making the present analysis special.

Locomotion analyses have rarely been performed in this polymorphic ant species, which seems perfectly suitable for analysing influences of caste and subcaste characteristics. In the present study, we thus performed an extensive locomotion analysis along gradients in both body size and (relative) walking speed for C. fellah worker subcastes and males, and emphasis was placed on morphometry and allometric relationships within the species; that is, on caste- and subcaste-specific characteristics.

Camponotus fellah (Dalla Torre 1893) ants were obtained from their natural habitat near Mahrés, Tunisia (34.53°N, 10.49°S) in 2017, and have since been kept at Ulm University, Germany (48.42°N, 9.95°S). Ants were reared at 27°C in two connected nests made of plaster (to maintain constant humidity); the nests were connected to a foraging arena (50×50×20 cm) with water and honey water ad libitum, and locusts were supplied twice a week. To analyse the locomotion of C. fellah castes and subcastes, high-speed videos were recorded during 2020 at Ulm University (around noon), under both laboratory conditions and open sky, whereby no notable differences could be observed between the different environmental conditions. Ants were collected from the foraging arena, directly filmed, killed by freezing and dissected afterwards (body parts were kept in numbered tubes with 70% ethanol).

We analysed the walking behaviour in the worker subcastes of C. fellah (minor workers: N=32 ants, n=130 top-view runs; medias: N=29 ants, n=93 runs; major workers: N=28 ants, n=116 runs) and compared it with the males' locomotion (N=11 ants, n=25 top-view runs) to examine subcaste- and caste-specific differences. Locomotion was additionally analysed by lateral recordings (n=15 runs per subcaste and caste). The experiments were conducted in compliance with the current laws, regulations and ethical guidelines of Ulm University.

For morphometric analysis, ants (Fig. 1) were dissected by cutting the head off, severing the six legs at the coxa–thorax joints and removing the gaster. Photos were taken through a dissection microscope (Stemi SV 6, Zeiss Microscopy GmbH, Jena, Germany) and measurements on the photos were performed with ImageJ (National Institutes of Health, Bethesda, MD, USA). We determined mesosoma length (ML) and width (MW), length of the leg segments (femur, tibia, basitarsus) (front legs: LL1, middle legs: LL2, hind legs: LL3), head length (HL), head width (HW), maximum eye distance (MD: the longest distance between the lateral eye margins), and total body mass (BM) of C. fellah. In addition to the ants used for high-speed recordings (N=89 workers: N=32 minors, N=29 medias, N=28 majors; N=11 males), we dissected N=30 workers (10 individuals per worker subcaste) and additionally determined mesosoma mass (MM), head mass (HM) and the (residual) mass of gaster and legs (RM) of those ants and of the N=11 males used for high-speed recordings (total: N=119 workers, N=11 males).

The respective length values of the femur, tibia and basitarsus were added to arrive at leg length (LL), and means were calculated for the three leg pairs [mean dispersion measure of 0.21 mm between left and right leg (s.d.)]. Mesosoma length was used to characterise body size (see fig. 8 in Andersen et al., 2020) and was one of the main characteristics to sort workers into subcastes of different body sizes and shapes. Total body length is in the range of around 2 to 3 times mesosoma length, depending on the subcaste and the state of the gaster. To define the subcastes of minor workers, medias and major workers accurately, we further used head size and body mass (Table 1, Fig. 2).

To describe the relationship between mesosoma and leg length and mesosoma length and head size within the worker caste and in males, we used the allometric equation y=a×x^b. Scaling factor a and exponent b were estimated for the male caste, the worker caste and the different worker subcastes (see also Sommer and Wehner, 2012; Tross et al., 2021).

High-speed recordings were made with two high-speed cameras (MotionBlitz EoSense Mini1 and Mini1-1, Mikrotron, Unterschleissheim, Germany; Nikon 105 mm 1:2.8 DG macro lens, Tokyo, Japan; and Tokina 100 mm 1:2.8 D macro lens, Tokyo, Japan) at sampling rates between 400 and 600 frames s⁻¹. Different frame rates were due to walking speed differences and recording window dependencies. Top-view recordings and lateral recordings were recorded separately.

For top-view recordings, a high-speed camera was mounted above an aluminium channel (30 cm long, 7 cm wide, and walls 7 cm high) and the ants were filmed while walking through that channel. The channel floor was coated with white paper to provide a slip-free ground and allow normal walking behaviour. For lateral-view recordings, the ants were filmed while walking through a channel with one transparent sidewall. For indoor illumination under laboratory conditions, two fibre optic cold light sources (Schott KL 1500LCD, 150 W, Schott AG, Mainz, Germany) were used, while no additional light source was necessary under the open sky. Ants were filmed at different controlled temperatures (10–40°C) to cover a broad walking speed range, as an increase in activity and speed could be achieved easily by increasing temperature (Fig. 3A). Channel temperature was regulated with both cooling pads and a heating light source and constantly measured with an infra red thermometer. Highest walking speeds were usually recorded at temperatures around 30°C; higher temperatures for more than a few minutes led to a decreased activity and lower walking speeds. Ants were not foraging and were not positively or negatively motivated. Sometimes, individuals were additionally spurred on by touching them with a fine paint brush (for activation), and thus we did not start the recording directly afterwards (random walks, no escape or foraging runs). Ants were filmed several times and up to five recordings of each individual were evaluated (each ant was filmed at low, medium and high temperatures, and we never used recordings of an ant with the same speeds). Furthermore, only straight and smooth running trajectories were analysed. Walking sequences with stops, decelerations, slipping steps or ants touching the channel walls were disregarded, and each evaluated high-speed video had to consist of a minimum of three complete step cycles per leg (and more steps for analysis of inter-leg coordination patterns). Only videos with nearly constant speed were evaluated (ensured by manual frame-by-frame analysis of speed and walking parameters during three consecutive steps). A piece of millimetre grid was used for calibration. All measurements were performed in a manual frame-by-frame video analysis with ImageJ (National Institutes of Health, Bethesda, MD, USA).

To compare the walking behaviour in C. fellah castes and subcastes, general walking parameters were analysed (Fig. 3; Figs S1, S2). Mean walking speed was calculated as the distance covered from the beginning of the first step cycle to the end of the third step cycle, divided by the time needed to cover this distance. Relative walking speed was defined as mean walking speed divided by mesosoma length. The timing of every lift-off and touch-down of the six tarsal tips was measured to obtain swing and stance phase durations. These values formed the basis for most further evaluations. Swing phase duration was calculated as the time between tarsal lift-off and touch-down. Stance phase duration was defined as the time when the tarsal tip touched the ground and did not move relative to the ground (Reinhardt and Blickhan, 2014; Wahl et al., 2015). Stride length was calculated for the respective leg as the distance between the tarsal lift-off and touch-down positions on the ground (no absolute stride length within the animal’s own coordinate system). To calculate stride frequency, we divided mean walking speed by stride length. Despite small differences between leg pairs, all parameters described above were averaged for all six legs to allow better comparison (see Fig. S2, single leg pair analysis).

In a frame-by-frame analysis, the x- and y-coordinates of tarsal lift-off (posterior extreme position, PEP) and tarsal touch-down (anterior extreme position, AEP) were measured for each leg with respect to the centre of mass [COM, assumed to reside in the petiole (Reinhardt and Blickhan, 2014) and graphically determined (Weihmann et al., 2015)] to quantify footfall geometry (see Seidl and Wehner, 2008; Mendes et al., 2013). The x-axis is lateral (left and right side) to the COM (origin) and the y-axis is parallel to the animal's longitudinal axis in the walking direction. Footprint positions relative to the COM were normalised to total body length and the standard deviation of the footfall positions was used to illustrate the spread of footfall positions (Fig. 4). To track the movement of the legs and to analyse the footfall geometry, we used ImageJ (National Institutes of Health), Excel (Microsoft Corporation, Remond, WA, USA), Sigma Plot 11.0 (Systat Software Inc., San Jose, CA, USA) and Inkscape (Inkscape 1.0).

Tracking of COM movement was performed in the horizontal (Fig. 5A, top view) and lateral planes (Fig. 5B, lateral view) in a frame-by-frame analysis using ImageJ (National Institutes of Health). We analysed COM movement in C. fellah worker subcastes and males during one step cycle (beginning of swing phase in one tripod until end of swing phase of the following tripod; n=30 top-view and n=15 lateral-view step cycles per group) at the same walking speeds (average foraging speed range at 25°C, ∼90–125 mm s⁻¹). In Fig. 5, the x-axis represents one step cycle, and the y-axis shows lateral (Fig. 5A) and vertical (Fig. 5B) COM oscillations towards the longitudinal axis in the walking direction.

The x- and y-coordinates of six tarsal positions during one step were measured for the front, middle and hind legs with respect to the COM in a frame-by-frame analysis of the lateral high-speed videos (compare method of footfall geometry) (Fig. 5C): tarsal lift-off (PEP), tarsal touch-down (AEP), highest tarsus position (Max.) and three additional coordinates of tarsus position (P1, P2 between PEP and Max., and P3 between Max. and AEP; with equal distances in each case). Tarsus positions relative to the COM were normalised to mesosoma length (ML) and standard deviations illustrate the spread of tarsus positions (Fig. 5C). The x-axis is parallel to the animal's longitudinal axis in the walking direction; the y-axis is perpendicular to the x-axis, representing tarsus positions relative to the COM.

The onset of swing phase in the left hind leg (L3) was taken as the reference in phase analyses (Fig. 6). Phase plots show the coordination of the six legs in a circular step cycle diagram and were created with the ‘CircStat’ Toolbox in MATLAB (MathWorks, Inc., Natick, MA, USA) (Berens, 2009).

To evaluate the quality of tripod coordination, we calculated tripod coordination strength (TCS) (Wosnitza et al., 2013). For that, we measured t₁, the time period with all three legs of one tripod in swing phase, and t₂, the time period from the first swing onset in any of the legs in one tripod to the last swing termination. The ratio t₁/t₂ was taken as the TCS. Higher TCS values indicate a strong synchronisation of the three legs of a tripod group (L1–R2–L3 and R1–L2–R3), whereas lower values indicate that the temporal relationship of swing phases shifted to other coordination patterns (TCS values between 0 and 1, with a value of 1.0 indicating perfect tripod coordination) (Wosnitza et al., 2013).

We calculated duty factor, the ratio of stance phase to step cycle period (stance plus swing of a given stride) that describes the transition from walking to running in bipeds but is also a common parameter in insect locomotion studies. Considering a particular pair of legs, a duty factor of 0.5 signifies the point where aerial phases appear in the leg coordination pattern to achieve higher walking speeds (Alexander, 2003). Note that in animals walking on more than one pair of legs, phase shifts between leg pairs can still avoid aerial phases for the whole animal (Wilson, 1966; Hildebrand, 1985).

Inter-leg coordination patterns were assessed using a frame-by-frame video analysis for qualitative and quantitative leg coordination analyses. Podograms were created to visualise the sequence of swing and stance movements of the six legs. They illustrate the patterns of the six legs, with black bars indicating swing phases and white bars indicating stance phases. We further quantified the inter-leg coordination by classifying each frame according to its momentary leg coordination pattern (Mendes et al., 2013; Pfeffer et al., 2019). These data were compared quantitatively for different walking speed ranges in the three worker subcastes and in males of C. fellah. To this end, an index number and a respective index colour, corresponding to the momentary inter-leg coordination, were assigned to each frame. We considered the momentary coordination patterns hexapod (brown, −1), pentapod (dark purple, 1), tetrapod (lilac, 2), tripod (dark green, 3), bipod (yellow, 4), monopod (orange, 5) and aerial phase (red, 6), depending on specific inter-leg coordination patterns (Fig. S4). Other leg coordination patterns were classified as undefined (grey). For a more detailed description, see legends of Fig. 7 and Fig. S5, regarding particular inter-leg coordination patterns, Fig. S4. To calculate the podograms and inter-leg coordination index, we used a MATLAB environment (MathWorks, Inc., Natick, MA, USA).

Box-and-whisker plots, normality tests, and statistical pairwise and multiple comparisons were generated in SigmaPlot 11.0 (Systat Software Inc.) and in Excel (Microsoft Corporation). Box-and-whisker plots show the median as the box centre, the 25th and the 75th percentiles as box margins and the 10th and 90th percentiles as whiskers. Data were tested for normal distribution with the Shapiro–Wilk test and Levene's mean test was used to assess equal variance. The t-test was used for pairwise comparison of normally distributed data. For multiple comparisons of normally distributed data, we used a one-way ANOVA with Holm–Šidák's method as post hoc test, and for non-normally distributed data, we used an ANOVA on ranks with Dunn's method as post hoc test. For comparison of circular data, we employed the MATLAB ‘CircStat’ Toolbox (Berens, 2009). Differences between phase relationships were tested for significance using the Kuiper two-sample test (circular analogue of Kolmogorov–Smirnov test). Tests for correlation were performed in Excel (statistical analysis of, for example, walking parameters, TCS values and duty factor). Final editing of the figures and tables was performed in Inkscape (Inkscape 1.0).

For morphometric analysis, ML, MW, HL, HW, MD and LL were measured. Furthermore, the ratios ML/MW and HL/HW, and BM were determined (Table 1). With regard to body size, the entire size range was covered, although the queen was not measured to preserve viability of the colony. The queen caste was the largest by eye, with large eyes and ocelli and a well-developed reproductive system, and to our knowledge the largest body mass and size (see Peeters and Ito, 2015). We followed the division into reproductive caste (queen and males) and (non-reproductive) worker caste, with further division into subcastes of minor, media and major workers, based on our data and the following pre-established criteria. Animals with a clear head modification and ML of at least 4.0 mm were classified as majors. Ants with ML below 3.2 mm or BM lower than 10 mg were considered minors. The class of medias consists of a multitude of transitions between minors and majors (ML: 3.2–4.0 mm, no head modification). In comparison, males exhibited markedly different body size and shape, with small heads, fragile legs, a different mesosoma shape and wings (Fig. 1).

ML of workers ranged from about 2.8 to 5.0 mm (minors: 3.07±0.14 mm, medias: 3.59±0.26 mm, majors: 4.36±0.19 mm), with head size (HW) increasing out of proportion towards the majors (minors: 1.34±0.09 mm, medias: 1.80±0.33 mm, majors: 2.97±0.30 mm). Positive allometric relationships were observed when considering ML in relation to HL, HW and MD, illustrating this change in head shape and especially the increase in HW. Majors had a larger and relatively wider head while medias and minors had more elongated head shapes (Table 1, Fig. 2). Allometric relationships between ML and LL in workers were slightly negative, with b-values for front, middle and hind legs <1. That is, with increasing body size, LL increased at a slower rate than ML, with majors possessing relatively shorter legs (Fig. 2A; for a detailed analysis, see Table S1). In comparison, males had a different body shape with ML comparable to that of medias (3.54±0.25 mm) but with very small heads (HW: 1.17±0.09 mm), longer gasters, fragile legs and noticeably shorter hind legs (Fig. 2; Table S1).

BM largely supported division into minors (9.29±2.80 mg), medias (17.81±7.82 mg) and majors (37.96±5.27 mg). BM of males was 10.28±1.23 mm. The relationship between HM, MM and residual BM was also determined by BM measurements. While the ratio of MM to total BM in workers and males was nearly isometric, HM exhibited a positive allometry in all ants (Fig. 2C).

In summary, the strong intraspecific size polymorphism was remarkable, with different allometric relationships between body parts, concerning in particular head size and leg length (Fig. 2; Table S1). Considering the entire colony, there were clearly more minors and even more medias than majors.

Highest walking speeds were achieved by minors and a few medias (319 mm s⁻¹ or 107 ML s⁻¹ and 330 mm s⁻¹ or 95 ML s⁻¹, respectively), with highest relative speeds in minors. Majors achieved considerably lower speeds (199 mm s⁻¹ or 46 ML s⁻¹) and males achieved maximum speeds as low as 120 mm s⁻¹ or 37 ML s⁻¹. All further parameters were evaluated in relation to both relative (Fig. 3) and absolute speeds (Fig. S1). Medias reached speeds above 200 mm s⁻¹ in only a few cases and therefore cannot be compared meaningfully with minors in higher speed classes.

It is remarkable that minors and medias reached considerably higher absolute and relative walking speeds than majors. With increasing speed, stride length and stride frequency increased in all worker subcastes. Majors tended to achieve the largest stride lengths (up to 13 mm) and showed higher absolute stride lengths than the other subcastes over the entire speed range (Fig. S1). In individual cases, minors and small medias increased stride length by the insertion of aerial phases at the highest speeds (Fig. 7). Aerial phases started to emerge at speeds around 250 mm s⁻¹ or 80 ML s⁻¹. Shorter stride lengths at similar speeds were compensated for by higher stride frequencies in smaller ants. Maximum frequencies were 25 Hz in minors and medias and 18 Hz in majors.

With increasing speed, swing phase and stance phase durations levelled off with a power function. Swing phase duration roughly plateaued towards 22 ms in all worker subcastes at speeds of around 150 mm s⁻¹ or 40 ML s⁻¹, and higher speeds resulted in hyperbolically decreased stance phase durations. Stance phase duration was as short as 12 ms in minors, at speeds around 300 mm s⁻¹ or 100 ML s⁻¹, and was slightly higher in medias and majors (minimum values 15 ms and 26 ms, respectively). Stance phase duration was longer than swing phase duration at low speeds, and this relationship reversed at higher speeds, though only in minors (∼139 mm s⁻¹) and medias (∼156 mm s⁻¹).

Overall, our relative walking speed (ML s⁻¹) data suggest that majors have relatively shorter legs than their smaller kin, as demonstrated by the slightly negative leg allometry of the worker caste (see morphometric analysis above; Fig. 2). At the same relative speeds, majors showed lower swing and stance phase durations than minors and medias; their relative stride lengths were lower and their stride frequencies were higher (Fig. 3).

Surprisingly small differences were observed in general walking parameters between subcastes and males, with respect to both absolute and relative speeds. However, significantly higher swing and shorter stance phase durations of the males' hind legs, a stronger increase in stride length with speed, and clearly lower maximum speeds of males were noticeable (Fig 3; Figs S1, S2).

Footfall patterns of the tarsi in PEP, shortly before lift-off at end of stance, and in AEP, shortly after touch-down at the beginning of stance, were analysed for worker subcastes and males. Compared with minors and medias, majors consistently showed footfall positions of all six legs, in both AEPs and PEPs, that were closer to the body over the entire speed range (Fig. 4), in agreement with their relatively shorter legs (Fig. 2A). Surprisingly, males showed footfall positions clearly more distant from the COM, compared with the worker subcastes. A significant influence of speed on footfall position could not be observed in the different groups, even if points of tarsal ground contact tended to be minimally closer to the body at low speeds.

In one step cycle, the COM covered distances between 8 and 13 mm in the anterior–posterior direction in all ants (minors: 9.53±0.80 mm, medias: 11.01±0.86 mm, majors: 10.38±1.03 mm, males: 10.98±1.59 mm; average foraging speed at 25°C). As described by Reinhardt and Blickhan (2014), forward COM speed decreased in the double support phase and increased during the rest of the step cycle. In the lateral direction (top-view videos) COM oscillated around zero, with maximum lateral displacements on a relative scale in minors (∼0.25 mm, ∼0.08 ML), followed by medias (∼0.26 mm, ∼0.07 ML) and majors (∼0.24 mm, ∼0.05 ML) (Fig. 5). In comparison, maximum lateral displacement of the petiole in males was around 0.40 mm (0.1 ML). Maximum values were reached at the beginning of swing phase in L1 and L3, or R1 and R3, respectively, and stance initiation in R2 or L2. On average, COM was at least 1.9±0.2 mm (0.64±0.07 ML), 2.6±0.3 mm (0.72±0.09 ML), 3.2±0.3 mm (0.73±0.06 ML) and 2.2±0.2 mm (0.61±0.07 ML) above the ground at the beginning of swing in a tripod in minors, medias, majors and males, respectively. The highest oscillation amplitude was observed in minors and males, with maximum values of 0.38 and 0.34 mm (0.12 and 0.10 ML), compared with 0.21 and 0.22 mm (0.06 and 0.04 ML) in medias and majors, respectively (Fig. 5).

Regarding body posture, minors walked with a more erect posture with their heads held higher than for medias and majors (Table 2; Movie 1). Lateral recordings showed that all ants, and especially males, were repeatedly out of balance for short time periods, stumbling, which affected the results of vertical COM movement and lateral tarsus tracking.

Tarsus trajectories of the front, middle and hind legs differed during a step, with maximum relative tarsus movement in minors (Fig. 5C). Maximum tarsus elevation tended to be highest in minors and males (0.57 and 0.76 ML above COM in front legs), while medias and majors rarely showed maximum relative tarsus elevation above the COM. Variance (s.d. bidirectional) was highest in minors and males and lowest in medias and majors. In males, dragging of the hind legs across the ground was noticeable.

TCS provides a measure of tripod gait quality, with a value of 1 characterising exact synchrony in the movements of the legs in a tripod. TCS values of workers were mostly between 0.5 and 0.8 and decreased with higher speeds (Fig. 6E). TCS values were higher in minors and medias than in majors (Fig. 6; test for correlation, P<0.05), and highest values were achieved at speeds of 50–100 mm s⁻¹ (20–40 ML s⁻¹). It is noticeable that TCS values of minors and medias rose again at the highest speeds, as relative swing phase duration increased, in conjunction with the occurrence of aerial phases. By contrast, TCS values were lowest in males and nearly constant at around 0.5 over their speed range (compare phase deviations, duty factor).

Phase plots illustrate variability of inter-leg coordination. They show the onset of swing phase during the step cycle in a circular diagram, with the left hind leg (L3) serving as reference. For all groups, a clear antiphase relationship between the two tripods L1–R2–L3 and R1–L2–R3 was evident. This tripod coordination was observed over the entire speed range, with small but consistent phase shifts between the tripod legs (significant shifts only in males: L3–R2 (P<0.001), R3–L2 (P<0.02, Kuiper two-sample test).

At low speeds (0–20 ML s⁻¹), phase shifts within one tripod group were smaller compared with those at higher speeds (20–40, 40–60 ML s⁻¹), which is consistent with the TCS values above. With increasing speed, an inter-leg coordination sequence hind leg–middle leg–front leg within a tripod group was established, with shifts between these legs of about 10% in minors (leg 3–2: 3.8%, 2–1: 6.3%), 8% in medias (leg 3–2: 3.1%, 2–1: 4.4%) and up to 15% (leg 3–2: 7.3%, 2–1: 7.8%) in majors (with no significant differences between subcastes) (Fig. 6A,B). Males, by comparison, exhibited the largest phase shifts of around 18% (leg 3–2: 15.4%, 2–1: 2.4%) with highest variance in the front legs (Fig. 6C,D). Furthermore, contralateral shifts between leg pairs (L3–R3, L2–R2, L1–R1) were around 50%, while ipsilateral shifts (L3–L2–L1, R3–R2–R1; consistent with shifts within tripod groups) varied between 42% and 65%, with higher phase shifts between leg 2 and leg 1, than between leg 3 and leg 2 in all subcastes (Table 3).

Duty factor characterises the transition from walking to running, which might occur at values around 0.5 (strictly for bipedal locomotion only; Alexander, 2003). With increasing speed, duty factor decreased non-linearly for all leg pairs. Only in the case of minors and medias did all three leg pairs fall below a duty factor of 0.5 (minors: front legs: 148 mm s⁻¹, middle legs: 206 mm s⁻¹, hind legs: 119 mm s⁻¹; medias: front legs: 165 mm s⁻¹, middle legs: 255 mm s⁻¹, hind legs: 131 mm s⁻¹) and these ants may thus be considered to run (Fig. S3). As the middle legs were the last to reach a value below 0.5 (middle legs with shortest swing and longest stance phases; Fig. S2), they determined the threshold at which the transition from walking to running might occur (minors: 206 mm s⁻¹ or 65 ML s⁻¹, medias: 255 mm s⁻¹ or 71 ML s⁻¹). All three leg pairs of minors fell below the value of 0.5 earlier in relation to absolute and relative speeds (test for correlation, P<0.05; Fig. S3). In comparison, majors and males did not make the transition from walking to running in their observed speed range.

We analysed inter-leg coordination patterns of worker subcastes and males by assigning an index number and colour to each video frame, according to its momentary leg coordination (number, colour codes of specific coordination patterns in Fig. S4). Podograms, with colour and number indices of worker subcastes and males at lowest and highest relative speeds were created (qualitative analysis, Fig. 7A,B) and the proportion of each coordination pattern was determined by indexing in adjacent speed bins (quantitative analysis, Fig. 7C; Fig. S5), for both relative (ML s⁻¹) and absolute speed (mm s⁻¹).

At low speed, coordination patterns with three or more legs in stance phase dominated in all groups, and phases with all six legs on the ground occurred. Fractions of tripod coordination increased in all groups up to medium speeds of 100–150 mm s⁻¹ (20–40 ML s⁻¹) and decreased thereafter (Fig. S5). Patterns with four or five legs in stance were replaced at medium and high speeds by bipod and monopod situations. Majors exhibited a remarkably high fraction of tetrapod coordination (150–200 mm s⁻¹: 6.7%; 40–60 ML s⁻¹: 9.6%) (compare ipsilateral phase shifts), while minors showed distinctly smaller fractions [<1% above 150 mm s⁻¹ (60 ML s⁻¹)].

Speeds above 200 mm s⁻¹ (60 ML s⁻¹) were achieved only by minors and few medias. Tripod coordination decreased notably at highest speeds (200–250 mm s⁻¹: 54.7%, >250 mm s⁻¹: 40.6%) and specific combinations with fewer than three legs in stance increased correspondingly (200–250 mm s⁻¹: 29.1%, >250 mm s⁻¹: 58.3%). A rare but special feature observed in only a few minors and medias was the incorporation of aerial phases at speeds above 250 mm s⁻¹ (minors: 14.4%, medias: 1.2%). In males, a comparatively high fraction of bipod coordination (14.4%) at speeds between 50 and 100 mm s⁻¹ was striking (compared with <1% in worker subcastes). This may be related to their short hind legs, with bipod fractions representing combinations of front and middle legs of a tripod.

In summary, coordination patterns of majors changed only little over the entire speed range compared with those of minors, medias and males. Further, the constantly high tetrapod fraction in majors and the occasional incorporation of aerial phases in minors and medias are notable (Fig. 7; Fig. S5).

Camponotus fellah is a species characterised by a distinct intraspecific size polymorphism with different relationships between body parts in the worker subcastes and in males (Fig. 2). Such an investment in different castes and especially more than one worker caste may influence colony fitness, reproduction and the competitive ability of C. fellah and has evolved repeatedly in a number of ant species (Bershers and Traniello, 1994; Powell, 2009; Wills et al., 2018). In this context, we investigated the walking behaviour of C. fellah and the influence of caste and subcaste characteristics, as efficient locomotion is relevant for colony survival. Concerning allometric relationships between body parts, in summary, majors have relatively shorter legs and wider heads than minors and medias (Fig. 2). In comparison, the male reproductive caste is distinctly different, equipped with wings, a small head and wispy legs with comparatively short hind legs. Significant differences in body size, associated leg length and head shape should influence walking performance between subcastes, and in comparison to males.

We observed species-specific locomotor behaviour with comparably small differences in walking parameters, though notable differences in speed, COM movement, walking posture and coordination pattern between castes and subcastes. Surprisingly, males showed quite similar walking behaviour to female workers, but with lower speeds, wider footprints and deviating phase relationships.

Regarding the influence of caste and subcaste characteristics on speed, one could assume that larger ants with longer legs (majors) reach higher speeds through larger steps than their smaller kin, minors and medias. Contrary to this expectation, highest speeds were achieved by minors and medias (318 mm s⁻¹ or 107 ML s⁻¹ and 330 mm s⁻¹ or 95 ML s⁻¹, respectively), but it should be noted that medias achieved such high speeds rarely. By contrast, majors reached distinctly lower maximum speeds (199 mm s⁻¹ or 46 ML s⁻¹), which appears to be related to their function in the colony. An investment in achieving high speeds is not required for majors that are specialised in nest defence with their massive mandibles, in contrast to minors and medias that are mainly responsible for brood care and foraging (Hölldobler and Wilson, 1990; Laciny et al., 2018, 2019). Foragers have to walk as efficiently as possible across long distances to provide food and ensure colony survival. Camponotus fellah males reached considerably lower speeds (120 mm s⁻¹ or 40 ML s⁻¹), which nonetheless is a good performance level considering their wispy legs (Fig. 1). Compared with the workers, males have a different function and are thus equipped for a different form of locomotion (clearly not adapted towards running longer distances). They possess wings for mating flights, large eyes relative to head size, and low body mass. Nevertheless, walking is also an issue for males, as they stay in the colony while waiting for nuptial flights that are synchronised across nests (Boomsma et al., 2005).

In conclusion, division of labour is based to a large extent on different castes and subcastes, and this is reflected in locomotion behaviour: highest speeds were achieved by small ants that are responsible for foraging, while large majors, specialised for nest defence, are much slower. Winged males were again significantly slower, as they generally do not walk much outside the nest. The observed differences in maximum speed illustrate that scaling of speed depends on body size and shape (Hurlbert et al., 2008; Hirt et al., 2017), and in this context on caste and subcaste characteristics.

Converted to body lengths per second, maximum speeds of C. fellah correspond to 38 BL s⁻¹ (minors), 33 BL s⁻¹ (medias), 17 BL s⁻¹ (majors) and 14 BL s⁻¹ (males). Higher speeds are achieved by, for example, Cataglyphis desert ants (108 BL s⁻¹) (Pfeffer et al., 2019) and a few other arthropods, including Cicindela eburneola (171 BL s⁻¹) (Kamoun and Hogenhout, 1996). Camponotus fellah workers thus cannot compete with typical Cataglyphis desert ants in terms of speed, even though some of them (e.g. C. albicans, C. bicolor) are observed in the same habitats. The lower speeds are probably related to nocturnal foraging activity, first, as lower night temperatures reduce walking speed in heterotherms (Hurlbert et al., 2008) and second, as there is no heat stress during the night that would require higher foraging speeds (Wehner et al., 1992). In this context, we could observe an increase in speed with temperature: highest speeds were achieved at around 30°C; even higher temperatures led to an activity and speed decrease (Fig. 3A,B) (Hurlbert et al., 2008). Further, C. fellah’s morphology with relatively short and muscular legs and notable differences in head shape and size would appear to affect locomotor performance (for example, compare leg length of C. bicolor; Tross et al., 2021). Other ant species, for example Formica polyctena (Reinhardt and Blickhan, 2014) or Messor barbarus (Merienne et al., 2020), do not achieve speeds higher than 200 mm s⁻¹, and C. fellah appears to be more comparable to these species with regard to locomotion.

Different castes and subcastes of C. fellah use the same overall locomotion strategy, independent of their body characteristics, with a comparably strong decrease of swing phase duration in all groups (Fig. 3C). Nevertheless, there are certain differences between castes and subcastes. Minors and medias employed higher stride frequencies (up to 25 Hz), very rarely increasing stride length by the incorporation of aerial phases at the highest speeds. Majors, by contrast, achieved maximum frequencies around 18 Hz and were not able to increase stride length by aerial phases. The lower frequencies may be attributed to their larger body mass and the fact that stride frequency is limited by contraction and relaxation kinetics of leg muscles and tissue viscosity (Rubin et al., 2016). When considering the whole relative speed range, majors reached significantly lower speeds than their smaller kin with shorter relative stride lengths and higher relative frequencies in comparison, which might be partly attributed to the slightly negative leg allometry in the worker caste (majors have relatively shorter legs; Table S1). Surprisingly, despite distinct morphometric differences, general walking parameters of males were quite similar, except for significantly higher swing and lower stance phase durations of the hind legs and a stronger increase in stride length (Fig. 3). In this context, gait variability and also phase shifts do not differ significantly between sexes in stag beetles, indicating that they also move in a dynamically similar way (Goyens et al., 2015). Overall, C. fellah ants, like most insects, maintain a typical tripod(-like) coordination pattern without aerial phases (tripod >50%; Fig. S5) (Hughes, 1952; Full and Tu, 1991; Reinhardt and Blickhan, 2014), with partly desynchronised tripods in males. Locomotion of C. fellah might be characterised as a form of ‘grounded running’ (Reinhardt and Blickhan, 2014; Merienne et al., 2021), a statically stable type of gait in small insects with comparable high speeds and without aerial phases.

Negative leg allometry in the worker caste – more precisely, relatively shorter legs of majors – were also observed in footprint positions, as their relative footprint positions are closer to the body than for medias and minors. The most striking differences in footfall position were observed between workers and males, indicating increased static stability in males (Ting et al., 1994; Zollikofer, 1994). It should be noted, however, that footprints of worker subcastes support statically stable walking as well, and broader footfall positions of males might be related to their spindly legs. In general, legs placed in more sprawled positions around the body provide a more stable posture for resistance against disturbances (Alexander, 1971) and can offer static stability during locomotion, especially at low speeds (Ting et al., 1994). In this context, it appears that continuous static walking in C. fellah results in less strict tripodal leg coordination with increasing speed (partly desynchronised tripods) – as indicated by phase shifts within tripods and lower TCS values – except for the few ants that incorporated aerial phases (more dynamic walking). When considering phase relationships, phase shifts between leg pairs were apparent in all ants, with a clear tendency for front legs to lift off last and strongest phase shifts in males (Fig. 6). Phase shifts increased with speed, which is reflected in decreasing TCS values, leading to poor tripod coordination accuracy, particularly in males (see Weihmann, 2018). In animals walking on more than one leg pair (multi-legged), phase shifts can avoid aerial phases for the whole animal up to a certain point (Hildebrand, 1985), explaining the connection between missing aerial phases and phase shifts at the velocities observed. Comparable shifts for front legs occurred in C. bicolor (Tross et al., 2021), and even more distinct phase shifts could be observed in connection with a metachronal gait during fast runs of the cockroach Nauphoeta cinerea (Weihmann et al., 2017). A velocity-dependent gait continuum with tripod coordination as the major pattern (Hughes, 1952; Wilson, 1966; Schilling et al., 2013; Dürr et al., 2018; Pfeffer et al., 2019; Tross et al., 2021) was observed in all ants, with a large fraction of combinations with fewer than three legs in stance. In this context, it was striking that majors did not show considerable changes in leg coordination patterns within the continuum, and had a consistently high tetrapod fraction compared with their smaller kin.

We further report that lateral and vertical COM positions are characterised by a periodic pattern with two peaks corresponding to the two tripods in a step cycle. Highest relative and absolute lateral and vertical oscillations were observed in minors and males, and it appears that minors are walking with a clearly more erect posture compared with their larger kin (Table 2). This has been described for Messor barbarus and might be related to a more excited state of minors in response to manipulation and environmental conditions and to the division of labour within a colony (Merienne et al., 2021). Differences in response to a threat between differently sized ants have also been described in Atta capiguara ants (Hughes and Goulson, 2001). However, different postures could also result from their mass distribution – majors have large and heavy heads – and the comparatively small body mass of minors. On average, majors are about 4 times heavier than minors. Moreover, tarsus positions relative to the COM, leg and COM movement indicated a spatial consistency during walking in majors and medias (Fig. 5). Maximum relative tarsus movement was observed in minors, and maximum tarsus positions above the COM in the front and middle legs of minors and males. Lowest variance in tarsus position was observed in majors, which also showed the lowest COM oscillations, indicating a comparatively uniform locomotion pattern with consistent leg movements (Movie 1). The noticeable dragging of the hind legs across the ground in males is related to their much shorter hind legs. In summary, C. fellah ants walk with relatively high speeds and periodic lateral and vertical COM oscillations, mainly to ensure ground contact and stable locomotion (Blickhan and Full, 1993; Reinhardt and Blickhan, 2014), but with the occasional bipod and monopod situations in their coordination patterns.

Although the polymorphic workers and males of C. fellah show distinct differences in body size, shape and mass, intraspecific differences in locomotion are less pronounced than morphological differences may suggest. Comparing C. fellah with Cataglyphis desert ants, for example, the latter employ distinctly higher speeds, with a strong increase in stride length, more than quadrupling in C. bombycina, in combination with aerial phases (Pfeffer et al., 2019). Camponotus fellah do not even double stride length over their speed range and hardly ever use aerial phases. Other ant species, such as Formica polyctena (Reinhardt and Blickhan, 2014) or Messor barbarus (Merienne et al., 2021), both of which perform a form of ‘grounded running’ with speeds comparable to those of C. fellah, also show walking patterns different from those of C. fellah. Formica polyctena, for example, performs a tripod gait with no more than three legs in the air at any time (Reinhardt and Blickhan, 2014), while all C. fellah groups inserted bipod and monopod situations (Fig. 7; Fig. S5). Relative speeds comparable to those of C. fellah males are achieved by Drosophila melanogaster (∼15 BL s⁻¹) (Wosnitza et al., 2013), which are adept at walking and flying locomotion, with tripod gait at high speeds and tetrapod and pentapod patterns at lower speeds. Studies on larger insects, such as cockroaches or beetles (Hughes, 1952), also demonstrated that tripod coordination is used at the highest speeds and main walking patterns are more variable at lower speeds, including tetrapod and pentapod patterns. In this context, a slow and fast gait (ambling and trotting) was found in Blaberus discoidalis (Bender et al., 2011) and a fast metachronal gait in Nauphoeta cinerea (Weihmann et al., 2017). In conclusion, C. fellah exhibit a more species-specific walking behaviour described as statically stable ‘grounded running’ (Szczecinski et al., 2018). Moreover, caste and subcaste characteristics mainly affect speed and, in this context, relevant walking parameters, while the general locomotion strategy is just slightly affected. Similar small intraspecific differences in locomotion were observed in male and female stag beetles (Goyens et al., 2015) and in workers of Cataglyphis bicolor, a species with strong variation in body size (ML 2.1–5.0 mm) but no subcaste differentiation. Cataglyphis bicolor show little intraspecific difference in walking performance with speed, in comparison to allometrically similar C. albicans ants (Tross et al., 2021).

Investment in different castes and especially worker subcastes influences colony survival, fitness and reproduction of ants (Bershers and Traniello, 1994; Powell, 2009; Wills et al., 2018), as matching of worker size and shape with task promotes organisational efficiency (Oster and Wilson, 1978; Powell and Franks, 2006). Such caste and subcaste characteristics are obviously reflected in the walking behaviour of C. fellah. On the one hand, our results support the idea of the division into small-sized and fast-walking foragers (minors and medias) and the considerably slower majors with large heads and mandibles for nest defence (Hölldobler and Wilson, 1990; Laciny et al., 2019). However, it seems that such a strict division may not accurately reflect additional factors such as size-dependent maximum load capacity (Mersch et al., 2013; Merienne et al., 2020). On the other hand, our study reveals that C. fellah ants use a common locomotor strategy across all body sizes, castes and subcastes, but with minors reaching the highest speeds. Thus, body size and shape mainly affect speed, and to a comparably small extent basic walking parameters, footprint positions, phase relationships and COM movements. The walking behaviour of the allometrically special males was particularly surprising here, but supports the assumption that differences in locomotion are more species specific than size or caste specific, when comparing our data with those from other ant species (e.g. Reinhardt and Blickhan, 2014; Pfeffer et al., 2019; Merienne et al., 2021).

We thank Marius Köhle for his support with laboratory experiments and high-speed video analysis. Alexander Lindt and Andrea Kubitz deserve sincere thanks for their support in animal care, and Wolfgang Mader for statistical advice. Angela Soiné-Wolf and Ursula Seifert helped with editing and correcting the manuscript text.