Data from: Resting networks and personality predict attack speed in social spiders

<b>Animal collection and maintenance </b><br>Colonies of <i>S. dumicola</i> were collected from roadside Acacia trees in the Northern Cape of South Africa in November 2015 (subadults) and March 2016 (adults), transported to the laboratory and fed crickets ad lib. The size of collected colonies ranged between 70 and 300 individuals and contained only females—males are short-lived and rare (12%) in natural colonies (Henschel et al. 1995). We created 15 groups of 26–30 subadult female spiders, from 4 source colonies of subadults, and 24 groups of 10 adult female spiders each, from 3 other source colonies. Individuals from different source colonies were not mixed. Group sizes were larger for subadults because of the small size of those individuals, and because it potentially requires more small individuals to execute a successful attack on large prey<br>Groups were housed in large round containers (18 cm diameter, 8 cm depth for subadults and 11 cm diameter, 10 cm depth for adults) with vertical wire meshes (two 9 × 6cm sheets positioned 10 cm apart for subadults and a 5 × 5cm sheet for adults) to allow the spiders to build both a retreat and a capture web. Trials were conducted from January until August, 2016.<br><br><b>Boldness </b><br>To determine individuals’ boldness, each spider was tested once a week using an established assay that measured the recovery of a spider from exposure to air puffs, which mimic the approach of an avian predator (Riechert and Hedrick 1993). Spiders react to the air puffs by huddling and remaining still. The faster the spiders resume movement after this simulated threat, i.e. move one body length away from where they were huddled, the bolder they are considered. <br><br>To test boldness, spiders were placed individually in a plastic container (15 × 15 cm) and after 30 s of acclimation, two puffs of air were administered to the anterior prosoma using an infant nose-cleaning bulb. Boldness was measured as the latency to resume movement and move one body length. Because bolder individuals resume movement faster, we subtracted the time to resume movement from the maximum duration of the procedure (600 s) to create a metric that increases with boldness. We designated as ‘shy’ individuals those with a latency to resume movement of 400–600 s, while ‘bold’ individuals were those with a latency to resume movement of 0–200 s. The abdomen of each spider was marked uniquely with acrylic paint to track their behaviour over time.<br><br><b>Group composition </b><br>To examine the effect of group composition on collective behaviour, we assigned spiders to one of three group compositions: all shy, all bold and ‘keystone’ (all shy individuals plus one bold individual). For subadults, we established five groups of each composition and for adults we established five groups of all bold, nine of all shy and ten keystone groups. Individuals that were not assigned to experimental groups, including those with a boldness score of 200–400, were returned to their source colony<br><br><b>Social interactions</b> <br>The physical contacts among spiders were manually recorded three times a week: immediately (1–2 h) before the prey capture assay, 2 days prior to the prey capture assay and 4 days before the prey capture assay. Resting interactions were defined as a physical contact between any body parts of two spiders. Interactions were observed during the day, when spiders are resting and inactive, which is their condition most of the time, unless disturbed by a prey in their web or by a destruction of their web that requires maintenance. Care was taken to note each spider in the colony so that all interactions are recorded. These interactions were used to construct social networks and calculate network variables that indicate individual, subgroup and group level dynamics.<br><br><b>Prey response </b><br>To determine the speed at which groups attacked prey collectively, we examined the groups’ latency to respond to vibrations on their capture web (Grinsted et al. 2013). We used a custom-made vibratory device assembled from an Arduino Uno board, a vibratory motor, and a metal wire, directed at a 1 × 1 cm piece of paper placed in the capture web (PinterWollman et al. 2017b). The stimulus was always placed on the capture web at the same distance (4 cm) from the nest retreat, where most spiders were gathered, to control for any effects the distance of the stimulus might have on the response of the group. The Arduino board was programmed to vibrate the piece of paper in pulses that varied randomly between 0.5– 1.5 s in both the duration of the vibration and the pauses between vibrations, to simulate the irregular vibrations that a prey makes when captured in the web (Hedrick and Riechert 1989). The paper was vibrated until a spider touched it, to avoid habituation to our stimulus, or until 10 min elapsed, in which case the trial was stopped (Pinter-Wollman et al. 2017b). As the first individual left the retreat, others followed, creating a collective response. The first individual to leave the retreat was not necessarily the one closest to the simulated prey (personal observations). When no attack took place, we set the latency to attack to 10 min. We noted the identity of the first individual(s) to touch the stimulus, as well as the identity of all the individuals that left the nest during the attack as participants, so that we could assess whether the keystone (boldest) individual participated in prey attack. Both adult and subadult groups responded to the simulated prey in a similar manner.<br>