Variation in Predator Recognition Across Three Species of Jumping Spiders ( Salticidae )

– The ability to recognize threats and to respond in a timely and appropriate manner carries significant benefits. Depending on the recognition task, this can be cognitively demanding. The zebra jumping spider ( Salticus scenicus ) is capable of visually recognizing static predator stimuli and reacts via a robust “freeze and retreat,” a potentially innate response in this species. Here, we extend this finding, asking whether the ability of spiderlings to recognize a static predator and to initiate an escape response is common across juvenile salticids, and if so, whether there is species-specific variation of anti-predator responses. We found that captive-reared spiderlings of three European salticid species from different genera ( Heliophanus cf. cupreus, Evarcha arcuata, Marpissa muscosa ) were able to robustly recognize and retreat from a stationary predator stimulus. Additionally, we found differences in the reaction times between the species as well as different behavioral repertoires associated with the escape response which may reflect species-specific predator avoidance strategies


_____________________________________________________________________________________
For species with similar life histories, a strong selective pressure to recognize and avoid predators may generate similar behavioral responses that transcend taxonomic boundaries.Early life stages are typically the most vulnerable, due to the small size of the animals, which likely makes them prey for a wide range of predators.This vulnerability could be exacerbated by a lack of experience.Newly emerged jumping spiders frequently fall prey to other jumping spiders (Okuyama, 2007;pers. obs. DCR), thus, it is reasonable to assume that newly emerged spiderlings are already able to recognize threats such as larger salticids without requiring experience or learning.Being precocial, once emerged, jumping spiders are equipped with the visual acuity of adults and readily hunt and navigate through their environment (Goté et al., 2019).While predator recognition is indeed linked to learning in some animals (Griffin et al., 2001;Ferrari et al., 2008;Polo-Cavia & Gomez-Mestre 2014;Mezrai et al., 2020), a recent study has demonstrated that newly emerged spiderlings of the zebra jumping spider (Salticus scenicus) readily recognize and flee from static 3D-printed salticid models (Rößler et al., 2022).To test whether this recognition ability in Salticus scenicus is also present in other jumping spiders, we experimentally tested static predator recognition in early life stages of three species of European jumping spiders of different genera.Furthermore, since predator recognition and responses likely are the result of species-specific selection pressures, we also tested whether species differed in reaction time or overall predator avoidance response.

Experimental Setup for and Procedure of Trials
The experimental setup was based on a previous study on Salticus scenicus (Rößler et al., 2022) but was adjusted to accommodate the smaller size of the spiderlings as well as to include a jump.This jump served as an integrated test to guarantee visual and physical fitness of spiders and to control for an overall explorative behavioral state.Spiders who failed to successfully jump in any trials were excluded from the experiment (three inidividuals).The setup consisted of a start platform and an object platform (Figure 1).The static model (spider or control model, described below) was placed one centimeter from the edge of the object platform.Spiderlings were gently placed onto the start platform by manually transferring them from a plastic vial (3 × 7 cm).This included either spiders voluntarily jumping out of the vial and onto the lower half of the start platform or the experimenter gently tapping onto the vial and placing the spider in the lower half of the start platform.From the start platform spiderlings could not see the model because the start platform was placed at an angle and below the object platform with a 0.5 cm gap between the platforms (Figure 1).Only once the spiderling jumped across the gap, it could see the model, after which the spiderling had to decide whether to retreat from it or move towards it (i.e., passing the model).Both the start and object platform were covered with filter paper which was replaced after each trial to avoid potential effects of chemical cues via silk trails.Spiders were filmed from above using a Nikon D7200 with a 40mm DX Micro Nikkor lens at 30 fps (Nikon, Tokyo, Japan).
Two models were used in this experiment: A black sphere (length = 6 mm) made from plasticine clay (Noris Club, Staedtler) and a black 3D-printed model based on a micro-CT scan of a Phidippus audax specimen (length = 6 mm) (Formlabs, Cambridge, the stl file of the model is publicly available from Zenodo open science repository, see Rößler & Shamble, 2022).The 3D-printed model's frontal eyes (anterior median and anterior lateral eyes) were painted with shiny black enamel paint (Item-Nr.REV-32107, Revell GmbH, Bünde, Germany) to ensure a more realistic reflection or "shininess" of the eyes, since eye features have been shown to be crucial for recognition (Harland andJackson, 2000, Rößler et al., 2022).Phidippus audax is a large salticid, known to prey on other salticids, and is found in North America (Okuyama, 2007), meaning that tested spiderlings were both ontogenetically and evolutionarily naive to the model.Each spiderling completed six trials, three with each model, all in randomized order.

Video Analysis
Videos were scored manually using the software BORIS (Friard & Gamba, 2016).A trial started as soon as the spiderling was on the start platform.The following behaviors were scored: "jump" (defined as a leap across the gap onto the object platform); "freeze" (defined as when the spiderling came to a noticeable, complete stop while oriented towards the model), "retreat" (defined as a sudden increase in distance to the model after a freeze), and "pass" (defined as when the spiderling walked past the model or climbed on top of it).Any time the frontal eyes were directed at the tested object (approximated by a perpendicular line from the pedicel through the midline of the frontal eyes meeting the object), we scored these phases as "oriented towards the object."A trial ended after a retreat, a pass, or if the spiderling was oriented towards the object but showed no reaction.Reaction time was calculated from the beginning of the last freeze to the beginning of the retreat.Due to this definition, reaction time was only scored for trials with a retreat reaction and is therefore not available for control trials.Additionally, we scored distinct behaviors that were associated with the retreat such as "backward walking", "leg waving", "instant jump/drop", or "pedipalp spreading".Multiple behaviors could co-occur during a retreat.

Statistical Analysis
We only used trials in which spiders were oriented towards the object presented to them (see definition above).Statistical analyses were conducted using R version 4.2.2 (R Core Team, 2019).Generalized linear mixed models (GLMMs) were performed using the package glmmTMB (Brooks et al., 2017).Subject ID was included as a random factor.We then used an analysis of deviance on the resulting model with the package car to test which factors had significant effects on the dependent variable (Fox & Weisberg, 2019).The emmeans package was used to carry out post-hoc analyses with Bonferroni-correction (Lenth et al., 2019) and model fit was checked using the DHARMa package (Hartig, 2017).Binomial distribution was used to model the probability of a spiderling passing the object in both tested conditions.The ggplot2 package was used to generate all plots (Wickham, 2016).All data and R scripts for analyses are available in the supplementary information of this article.

Results
In total, we conducted 265 trials.In 217 trials, spiders oriented towards the presented object.In trials using the 3D-printed spider model, we could observe the same freeze and retreat behavior reported for Salticus scenicus (in Rößler et al., 2022) in all three salticid species (Video S1).
A second measure of trial outcome, the probability of retreating from the model, was greater when the object was a 3D-printed spider, compared to the control object (GLMM analysis of deviance, ꭓ 2 = 17.26, p < .001,n = 217).

Plate & Rößler 54
Likely due to the small sample size, no GLMM could be converged to analyze the differences in reaction time towards the spider model.However, Evarcha took the longest time to react (median (IQR); 4.63 s (7.63 s); nobs = 28), followed by Heliophanus (1.25 s (2.50 s); nobs = 35), while Marpissa had the fastest reaction time (0.75 s (0.75 s); nobs = 25) (Figure 3).Although we detected no overall effect of eggsac/mother, we additionally visualized reaction time per eggsac/mother because of the limited number of eggsacs the spiderlings derived from.Spiderlings with the longest reaction times in Evarcha all stemmed from the same mother (Figure 4).Lastly, we observed differences in retreat-associated behaviors across the three tested species (Figure 5, Video S2).Heliophanus showed a broader behavioral repertoire associated with the reaction towards the 3D model than the other two species.It is particularly noteworthy that Heliophanus always spread their pedipalps and often bobbed their abdomen when reacting towards the predator model, exhibiting behaviors which were completely absent in the other two species.

Discussion
Our study indicates that the recognition of stationary salticid predators is potentially common in European jumping spiders.Testing such abilities across different species holds valuable information about the origin and evolution of cognitive traits and how life history or ecology may play a role in shaping these traits (Aguilar-Arguello & Nelson, 2021).
The robust static visual recognition of potential threats and the subsequent behavioral response carry a selective benefit; therefore, we predicted it to be a common trait across jumping spiders, which are highly visual and apt predators.Given the morphologically conserved features of jumping spiders such as their characteristic large anterior eyes and their unique eye arrangement (Morehouse, 2020), it is not surprising that spiderlings could detect the models and were triggered to retreat, even when the model was based on a spider that spiderlings were naïve to.
We detected variation in the reaction time, raising questions around the underlying factors causing this.We consider several possibilities that warrant follow-up studies and further inquiry.First, there could be an overall difference of cognitive abilities between species (Aguilar-Arguello & Nelson, 2021; Gómez, 2005), which might lead to differences in processing information.For example, jumping spiders in the genus Portia are recognized for their high levels of cognitive abilities that include the ability to detour (Cross et al., 2020;Jackson & Cross, 2011;Tarsitano & Andrew, 1999).Other species of jumping spiders have demonstrated associative learning and reversal learning (Liedtke & Schneider, 2014), while others are seemingly lacking associative learning abilities, even for ecologically relevant cues (Vickers et al., 2021).An alternative, but not mutually exclusive explanation, could be species-specific differences in life history and habitat use (Carducci & Jakob, 2000;Steinhoff et al., 2018).While Heliophanus and Evarcha are mainly found amongst highly structured, three-dimensional vegetation in meadows (Sanders et al., 2015;Scheidler, 1990), Marpissa resides predominantly on two-dimensional surfaces such as bark, fences or walls (Steinhoff et al., 2020), offering fewer structural opportunities to hide or retreat from predators during close-range encounters, such as the one simulated in our experiment.Lastly, the response differences could indicate species-specific anti-predator strategies.It is especially noteworthy that spiders sometimes froze for extended periods before initiating an escape response, particularly so in E. arcuata.Here, it is particularly noteworthy that the longest freezing times were consistently observed in spiderlings from one specific eggsac/mother, including one freeze that lasted over 100 sec before a reaction was initiated (Figure 4).Due to the relatively small sample size, we cannot be certain whether the overall longer freezing can generally be attributed to Evarcha or whether we are observing a parental effect, as is known, for example, from damselfish (Atherton & McCormick, 2020).Similarly, in Heliophanus, spiderlings from one eggsac/mother have seemingly longer reaction times.
Freezing likely entails scanning and information gathering, but it seems unlikely that extended freezing periods (> 10 sec) would result from ongoing stimulus processing, since predators are likely to strike more rapidly in a natural setting.Freezing most likely already indicates a successful recognition; thus, freezing may be a first line of defense and an anti-predator mechanism in itself.Rather than swiftly investing in a potentially costly escape response, staying motionless could prevent an attack.Freezing is a common response to threats in animals (Eliam, 2005;De Franceschi et al., 2016).This may be particularly effective in close range encounters and when the predator is likely to react to movement by the prey.This explanation could also be connected to differences in habitat.The highly complex structure of dense vegetation would allow species like Heliophanus and Evarcha to simply drop or jump into a highly structured surrounding when a potential attack is imminent, potentially driving this initial and extensive "freezing" response.To fully investigate the natural differences in anti-predator strategies, an experiment testing the reaction to predator models in more natural settings is crucial.This could also include an experimental setup better reflecting the three-dimensionality of natural habitat as described earlier.
During the trials, we observed differences in retreat-associated behaviors between the tested species (Figure 5, Video S2) that substantiate the presence of species-specific responses and may inspire follow-up studies.We regularly observed what appears to be signaling with the front legs in all species, but it seems to be less common in Evarcha.Similar, but mostly unilateral, front leg movements can sometimes be observed in neutral context locomotion in Marpissa (pers.obs.DCR).All three species, but more so Evarcha and Heliophanus, engage in jumping away or dropping as a response towards the 3D model.This is in line with our previous hypothesis that these responses may reflect a beneficial strategy in a more complex habitat setting.Additionally, Heliophanus frequently bobs its abdomen when faced with the predator model and always extensively spreads its pedipalps.While abdomen bobbing is generally observed in this species during locomotion stops, the frequency and extent of the behavior is visibly increased during predator encounters (pers.obs.DCR).These behaviors are completely absent in the other two species and warrant a deeper inquiry.Marpissa showed the highest proportion of walking away backwards, which may, again, reflect a strategy that is adapted to its less structurally complex habitat.Our study had some limitations that need to be addressed in future studies, such as the range in age and the resulting range in size of the tested spiderlings (from 2 to 5 mm), as well as a generally small sample size.Ideally, spiderlings would also stem from a higher number of different eggsacs/mothers to enable specifically testing parental effects.Explicitly testing the effect of previous parental predator exposure on offspring in jumping spiders could also be an intriguing future avenue.
Additionally, testing ontogenetic changes of the response and reaction time could be important in understanding the impact of experience and development or whether the reaction time and response is hardwired within a species.Preliminary analyses using the data from Rößler et al. (2022), which used adults and spiderlings of the same species (Salticus scenicus), indicate that reaction time might be relatively constant within a species (spiderlings: median reaction time = 4.22 s (IQR = 6.25, nobs = 62); adult median reaction time: 3.49 s (IQR = 5.62, nobs = 131).Beyond the species level, a deeper inquiry into individual variation and within-individual robustness of anti-predator behavior is equally possible and offers an intriguing opportunity to test questions of animal personality in this group.
The robustness of threat recognition in salticids when faced with a larger salticid (i.e., a potential predator) offers numerous opportunities to explicitly test aspects of anti-predator adaptations, behavioral strategies, and visual stimulus processing that can feasibly be examined across a range, if not all, salticids.Ultimately, larger comparative studies across the salticid tree of life could yield powerful insights into this intriguingly robust visual-cognitive behavior and its evolution.An extension of the paradigm into spider lineages that are also visually apt (e.g., lynx spiders, Oxyopidae, or wolf spiders, Lycosidae) would confirm if this cognitive ability has evolved several times independently in visual hunters among spiders.

Figure 1 Experimental
Figure 1Experimental Setup for a Predator Recognition Trial in Jumping Spider Spiderlings

Figure 2
Figure 2 Model-Based Plot for Post Hoc Analyses Showing Probability to Pass the Tested Object

Figure 3
Figure 3Boxplots Showing Reaction Time Towards the Spider Model for Tested Species

Figure 4
Figure 4 Boxplots Showing Reaction Time of Spiderlings Towards the Spider Model, Grouped by Egg Sac/Mother

Figure 5
Figure 5Percentage of Retreat-Associated Behaviors Towards the Spider Model Across Three Tested Species with Multiple Behaviors Possible per Retreat