The decision to leave or join a group is important as group size influences many aspects of organisms’ lives and their fitness. This tendency to socialise with others, sociability, should be influenced by genes carried by focal individuals (direct genetic effects) and by genes in partner individuals (indirect genetic effects), indicating the trait’s evolution could be slower or faster than expected. However, estimating these genetic parameters is difficult. Here, in a laboratory population of the cockroach Blaptica dubia, I estimate phenotypic parameters for sociability: repeatability (R) and repeatable influence (RI), which indicate whether direct and indirect genetic effects respectively are likely. I also estimate the interaction coefficient (Ψ), which quantifies how strongly a partner’s trait influences the phenotype of the focal individual and is key in models for the evolution of interacting phenotypes. Focal individuals were somewhat repeatable for sociability across a three-week period (R = 0.080), and partners also had marginally consistent effects on focal sociability (RI = 0.053). The interaction coefficient was non-zero, although in the opposite sign for the sexes; males preferred to associate with larger individuals (Ψ_male = -0.129) while females preferred to associate with smaller individuals (Ψ_female = 0.071). Individual sociability was consistent between dyadic trials and in social networks of groups. These results provide phenotypic evidence that direct and indirect genetic effects have limited influence on sociability, with perhaps the most evolutionary potential stemming from heritable effects of the body mass of partners. Sex-specific interaction coefficients may produce sexual conflict and the evolution of sexual dimorphism in social behaviour.

Experimental animals: Blaptica dubia is a quite large (up to 45 mm in length) sexually dimorphic blaberid cockroach (Wu, 2013). They typically live in aggregations at high temperature and humidity in central and south America (Alamer & Hoffmann, 2014), consume vegetative matter, and are ovoviviparous. They are described as “gregarious” (Grandcolas, 1998) or “communal” as individuals of the same generation cohabit (without shared parental care; Bewick et al., 2017). I purchased an initial colony of B. dubia online in March 2021. I maintained them at the University of XXX [removed to preserve anonymity] at 28°C, 50% humidity, with a 50:50 light:dark cycle. I provided them with cardboard egg trays for shelter, carrots for hydration, and Sainsbury’s Complete Nutrition Adult Small Dog Dry Dog Food (approx. nutritional composition = 1527 kJ energy, 24g protein, 12g fat per 100g) for nutrition. Mortality was very low at all life stages (0.31% of the adults in the colony died per half week) indicating the colony was healthy. I moved newly born nymphs every few days to a container of dimensions 610 x 402 x 315 mm of similarly aged individuals (density ranged from a few hundred of the earliest instars to 10-80 of later instars) and maintained them in mixed groups until adulthood (seven instars which take approx. 250 days at this temperature; Wu, 2013). Upon reaching adulthood I moved them to either single-sex groups (again in containers of 610 x 402 x 315 mm) or in small groups of two males and four to eight females in a container of dimensions 340 x 200 x 125 mm for breeding to maintain the stock population. For this experiment, I selected 48 unmated males and 48 unmated females from the single-sex adult groups. Individuals were all more than five days old, and so presumed to be sexually mature (Hintze-Podufal & Vetter, 1996). I transferred each individual to a clear plastic box (79 x 47 x 22 mm) labeled with its unique ID to allow individual recognition. I gave individuals a small piece of carrot for hydration which was replaced weekly.

Data collection: I tested individuals in two blocks of 48, treating all individuals in each block once as a focal individual and once as a partner for a member of the same sex over two days. This means that in the first two days 24 males and 24 females were each assayed for sociability once and each acted as a partner individual once. On days three and four I repeated this with a second block of 24 males and 24 females. In this way individuals only ever acted as focal or partner individuals with members of the same sex in the same block (either first or second) and were each assayed for sociability and acted as a partner once per week. I repeated this for three weeks, so each individual was assayed up to three times as a focal individual and acted as a partner up to three times. Some individuals received fewer than three trials if they died (n = 5 males and 0 females), in which case I replaced them with a member of the same sex from the stock population (who did not inherit the same ID and was therefore another unique individual). Individuals might also record fewer than three measures for sociability if the mesh was breached by either the partner or the focal before the trial began (11 females and eight males recorded one measure, 48 females and 42 males recorded two measures, 30 females and 54 males recorded three measures).

I assayed sociability in medium-sized plastic boxes (200 x 100 x 70 mm) where I glued a fine polypropylene mesh (mesh size 0.6 x 0.6 mm, Micromesh, Haxnicks) across the interior 50 mm from one end. This creates an arena with a small compartment (50 x 100 x 70 mm) and a large compartment (150 x 100 x 70 mm) separated by the mesh (Fig. 1A). Separating by mesh was necessary to prevent a partner individual from imposing close proximity on the focal individual by constantly following or attempting to dominate it (Clark et al., 1995), and therefore my assay captures the focal individual’s willingness to socialise, rather than the partner’s (Gartland et al., 2022). For the first block, I randomly placed 12 females and 12 males alone, each into their own plastic box, in the large compartment. These were the focal individuals. I then randomly placed an individual of the same sex into the small compartment; these were the partner individuals. I used individuals of the same sex to ensure I was measuring sociability rather than a willingness to mate. I then placed these 24 arenas into four large plastic boxes (six in each) which I placed underneath a video camera (ABUS IP video surveillance 8MPx mini tube camera), so that each video camera recorded six arenas simultaneously. I maintained the room the video recordings occurred in at 20-22°C using portable heaters, while I used a thermometer to record the temperature at the start and end of each trial. I was not able to control or monitor humidity during the trials. Once all arenas were in position and cameras focused, I started the recording and left the room. The lights automatically switched off after 40 minutes, and so the trial began 40 minutes after I left the room, in darkness, which is when B. dubia is active (Bouchebti et al., 2022). I returned two hours after leaving to end the trial, meaning the trials lasted 80 minutes. In the darkness, the cameras automatically switch to infra-red filming using infra-red LEDs.

For each trial, every ten minutes I recorded the proximity of the focal individual (in the larger compartment) to the mesh that separated it from the partner individual, giving a maximum of eight measures per trial. The distance of an individual to a conspecific in this manner is often used to measure sociability (reviewed in: Gartland et al., 2022) and by using the distance to the mesh rather than the partner I have a measure solely under the influence of the focal individual; the partner cannot directly influence the distance by moving closer or further away. If the focal was sat directly on the mesh (perpendicular to the floor) I recorded a distance of zero, otherwise, I used the hexagons on the bottom of the box to record how far the focal individual’s head was from the mesh. Smaller values mean a focal individual closer to the partner individual which indicates higher sociability. Individuals were in some cases able to bypass the mesh (this occurred 95 times before the lights went out and 33 times after they did out of 288 trials, the 33 breaches after lights out are still included in the analyses with only the measurements before the breach used, see Data analysis). I used the video recordings to determine when this happened and stopped recording data from the video as soon as either individual bypassed the mesh. If either individual bypassed the mesh in the 40 minutes the lights were on before the trial started then I recorded no data from that trial. To avoid mixing individuals up at the end of the trial I dotted either the partner or the focal with white paint (Edding Extra-fine paint markers). At the end of the trial, I returned all individuals to their unique boxes. I then weighed all partner individuals to the nearest 0.01 g (Fisherbrand Analytical Balances, readability 0.0001 g). In the first two days I also weighed the focal individual, but since the correlation between an individual’s mass as a focal and its mass as a partner either the following or preceding day was 0.994 (Pearson correlation, t₃₈ = 56.894, p < 0.001) I stopped doing this to save time. Instead, I entered the mass of the focal individual as its mass as a partner individual recorded the same week (always within one day). As described above, each individual in the two blocks was assayed once as a focal and acted once as a partner for another individual of the same sex in that block per week, and this was repeated for three weeks.

After the third trial, I aggregated individuals into four groups of 21-24; all the individuals of the same sex from the same block were together, with groups having fewer than 24 individuals if any died (I did not replace individuals that died with stock individuals as I was only interested in the social network position of those with known sociability from the dyadic trials). I gave each individual a unique combination of two colours (red, green, blue, white, gold) on their wing cases using paint pens (Edding Extra-fine paint markers) which allowed me to track them individually (combinations were repeated between groups i.e., red-blue featured in each of the four groups). I then placed each group into new plastic boxes (340 x 200 x 125 mm) along with four shelters made from cardboard egg trays (approx. 100 x 120 mm), each placed vertically at each corner on a long side. Shelters were taped to the walls of the box, creating clear space between both the shelter opposite it (on the opposite long side) and next to it (on the same long side). I placed 2 g of dog food and 10 g of carrot in the centre of each box. Each shelter was large enough to accommodate many but not all of the individuals, and the number of shelters was considerably less than the number of individuals. Therefore, the formation of aggregations in shelters was enforced, but individuals could move between shelters and therefore could exert some influence on who they co-habited with. Regularly after placing the individuals into these groups (after 3, 10, 14, 18, and 21 days) I recorded which individuals were using the same shelter. Individuals who could not be identified were recorded as such but they were not used to build the networks. Collecting data in this way gives a group-by-individual matrix analogous to those generated by observing flocks of birds or herds of ungulates in the wild, and further is similar to methods than have been used to generate social networks in forked fungus beetles (Bolitotherus cornutus; Formica et al., 2012, 2016, 2020) and maritime earwigs (Anisolabis maritima; Vipperman, 2021). While a single incidence of sharing a shelter could be due to chance, by aggregating these observations I can infer consistent social associations. When recording these data, I also updated any paint markings that were starting to wear, maintaining individually-recognisable marks for the duration of the experiment, and replaced carrot and dog food as necessary.

R is required to run the analysis code. The script relies on various packages which will need to be installed before using.