Organisms may respond in different ways to the risk posed by conspecifics, but the cause of such variation remains elusive. Here, we use a half-sib/full-sib design to evaluate the contribution of (indirect) genetic or environmental effects to the behavioral response of the cannibalistic wolf spider Lycosa fasciiventris (Dufour, 1835) towards conspecific cues. Spiders showed variation in relative occupancy time, activity, and velocity on patches with or without conspecific cues, but direct genetic variance was only found for occupancy time. These three traits were correlated and could be lumped in a principal component: spiders spending more time in patches with conspecific cues moved less and at a lower rate in those areas. Genetic and/or environmental components of carapace width and weight loss in the social partner were significantly correlated with the principal component of focal individuals. Individual Variation in these traits may reflect the quality and/or quantity of cues produced by social partners, hence focal individuals were likely behaving along a continuum of strategies in response to the risk posed by social partners. Therefore, environmental and genetic trait variation in the social partners may be key to maintain trait diversity in focal individuals, even in the absence of direct genetic variation.

Spider collection and rearing conditions

Lycosa fasciiventris individuals (adult males and subadult females) were collected between June and July 2015 in four different localities within the Almeria province (South-East Spain), in dry temporal washes (“ramblas”): near Boca de los Frailes (36.8036°N, 2.1386°O), near Carboneras town (36.9667°N, 2.1019°O), at Almanzora river (37.3414°N, 2.0078°O) and near Paraje las Palmerillas, Estación Experimental Cajamar (36.7917°N, 2.6891°O). They were reared in individual tanks (22 cm x 18 cm x 18 cm) with the bottom filled with 2–3 cm of soil collected from one of the sites. All individuals were fed once a week with size-matched crickets (Gryllus assimilis Fabricius, 1775) purchased from the pet supply virtual store Exofauna (available at: https://exofauna.com). Spiders had access to water ad libitum through a 40 ml vial filled with water and covered with cotton. Vials were checked and refilled every 2-3 days. Holding tanks were placed in a climate chamber that simulated outdoor climatic conditions of the preceding weekly average conditions in the Almeria province (night:day temperature cycles between 18.7–34.3 ºC; photoperiod of 17:7 h - 16:8 h light-dark with light bulbs of 54W and a relative humidity of 50–65 %).

Individuals used in the experiments were removed from their mothers’ back 42 ± 8 days after hatching and placed in separate cylindrical containers (15 cm height and 6 cm of diameter) inside a growth chamber with controlled temperature (25°C ± 1ºC), humidity (70% ± 5%) and photoperiod (16:8 hours light:dark). The bottom of each container was covered with a filter paper replaced weekly. Water was provided ad libitum from a cotton string submerged in a reservoir, below each container and providing water by capillarity (Moskalik and Uetz 2011). Each week, spiderlings were fed with fruit flies (Drosophila melanogaster Meigen, 1830), reared in a nitrogen-rich medium supplemented with high-quality dog food to ensure survival and growth of spiderlings (Jensen et al. 2011). A portion of the offspring within each dam family (3 out of 12) was reared in a richer environment by providing them three times the amount of food given to the remaining spiders, to examine how this affected both cue emission and the response to such cues. For logistic reasons, spiders were not reared until maturation and thus their sex was unknown. We assumed that sex differences were not very pronounced at the early stages, as is the case in a syntopic and co-generic wolf spider (e.g., Fernández‐Montraveta and Moya‐Laraño 2007).

Experiment

We aimed at testing the role of genetic and indirect genetic effects in shaping the variability in the interaction between cue-emitting spiders and spiders that perceive such cues. Our experimental design is thus highly asymmetrical, as one individual is only present via the cues it emits, whereas behaviour is measured in the other individual. Note also that each individual served first as a social partner, emitting cues, and was subsequently used to measure its behaviour in another arena. All trials were recorded between April 21^st and June 8^th 2016 in a total of 50 blocks and 17 recording days. Blocks consisted of 3 rows x 5 columns = 15 Petri dishes (individuals) recorded under a single camera. Usually, three blocks were recorded in a single day with three different cameras. More details are provided below. 

Cues and traits in social partners

Behavioral responses to conspecific cues were measured in small Petri dishes (5.5 cm diameter) with the bottom covered with filter paper divided in two halves: one containing intact filter paper (control) and the other impregnated with conspecific cues (e.g., excreta and silk). These cues were produced by juvenile conspecifics (hereafter social partners) enclosed in a small Petri dish for 10 days with filter paper on the bottom and fed with 10 fruit flies each during the first 36 hours. The nature of the cues released included the excreta, silk, odour and chemo-tactile cues of dead prey. The spatial position of the two filter paper halves was randomized to eliminate any potential side bias.

As a proxy for the amount of cues produced, we used (a) weight loss of spiders confined within a petri dish for 36h and (b) carapace width, reflecting body size. In spiders, weight loss may be an indication of the quantity or quality of conspecific cues released.  Indeed, since a great proportion of the cues likely correspond to excreta, animals losing more weight were likely those that also released more cues. Alternatively, higher weight loss may be related to animals that have higher voracities, as recently found in another wolf spider (Rádai et al. 2017),  which could be associated to their willingness to attack conspecifics (Arnqvist and Henriksson 1997). Moreover, carapace width is known to be correlated with the amount of cues produced in spiders (Persons et al. 2001).

Weight loss was calculated by feeding spiders with 10 flies each, then weighting them after 36 hours and after 10 days. Proportional weight loss was then estimated as the difference in weight between these two measures divided by the initial weight. During this period, spiders were provided water ad libitum by means of a soaked piece of cotton of about 1 cm in diameter, which was re-soaked every two days. Body mass was measured to the nearest 0.1 mg using a high-precision scale (Mettler Toledo XP26). Body size was assessed by measuring the maximum carapace width (i.e., carapace width at its maximum span). Measurements were performed with a stereomicroscope (Leica MZ125) with a precision of 0.1 mm.

Behavioral measurements

Spiders in which the behavioral response to social cues was measured (hereafter focal individuals) were randomly assigned to a Petri dish, but care was taken that they were not genetically related to and came from the same feeding treatment as their social partner. Note that each individual was used both as social partner and focal individual.

During the behavioral observations, light was made homogeneous by introducing the camera and the Petri dishes inside a 40x40x30 styrofoam box. Locomotor behavior was measured by monitoring spiders, in blocks of 15, through recordings retrieved from a video camera (Sony® HDR CX-150) placed overhead. A video-tracking software was implemented, allowing estimating movement at 25 frames/second. We obtained information for 563 spiders, with a mode recording time of 6 hours for each spider (min = 0.38 h, max = 6.04 h, median = 5.5 h). This large variation in recording time was due to an ant or fly walking on top of the Petri dish, or to a lack of space in the camera hard drive. However, we included recording time in the statistical analyses (see below), and it did not affect our results.

Differences in behavioral patterns between sides were estimated using the relative interaction intensity (RII) index described in Armas et al. (2004):

R_trait = (trait_cues - trait_control ) /  (trait_cues + trait_control )

where trait_cues is the mean value of the trait in the patch with cues and trait_control is the mean value of the trait in the patch without cues. Using this general formula, we calculated: (a) the relative activity (RA) index as the time spent moving, (b) the relative mean velocity (RMV) index as the difference in speed and (c) the relative occupancy time (ROT) index as the difference in time spent in patches with or without conspecific cues (fig. S1).

In addition to these three traits, we performed a principal component analysis (PCA) to estimate a composite behavioural score based on the three traits, thus accounting for potential correlations among them. This was done in R v.4.0.2 (R Core Team 2020), using the function “principal” in library “psych” without axes rotation (fig. S2). We sought a single PC that explained > 50% of the variance and in which the three traits had substantial loadings (, ≥ |0.7|).

In these focal individuals, we also used the measurement of weight loss and carapace width, as described above (cf. Cues and traits in social partners). Moreover, we added a measure of body condition, i.e., the maximum abdomen width (where nutrients and body fats are stored). Body condition has been shown to be negatively associated with movement in these spiders (Moya-Larano 2002), hence we used them as covariates in the statistical models measuring DGE, such as to account for this potential source of variation (see below).