1 Interactions of Phidippus regius jumping spiders with lifelike 3D printed models. Abstract Dummies (lifelike models of real animals) are valuable tools in the st udy of animal behavior. T hey allow researchers to stage interactions and observe live animal s responses They also allow researchers to conduct controlled manipulative experiments to test hypotheses about animal communication. In this study, we created dummies of the jumping spider, Phidippus regius, by microCT scanning real spiders and p rinting life s ized 3D models of t hese scans. The models were hand painted, manipulated to move like real spiders (using magnets), and then presented to live spiders. T o understand if a spider would interact with the model similarly to a live conspecific, w e set up tests that compared spider behavioral responses to a live conspecific of the opposite sex, a model made to look like a live conspecific of the opposite sex, and a cricket. We found that both male and female spiders responded differently to the mod el than they did to pre y but did not interact with the model exactly how they would with a live conspecific. Further testing of the models in the presence of other cues could provide better understanding of a live spider's perception of the mo del and how t hese models might be improved for use in future research. Introduction Basic r esearch on animal behavior, communication, and evolution ary traits can lead to a variety of valuable yet unexpected insights and discoveries (Brennan, Irschick, Johnson, & Craig Albertson, 2014; Hsiung et al., 2017) Behavioral studies focused on animals' evolutionary changes in body colors and str uctures have led to innovations in technologies used in medicine and the military (Brennan et al., 2014) In a particularly interesting example, basic research on
2 the behavior of peacock jumping spiders ( Maratus robinsoni ) brough t attention to the striking and colorful courtship displays of males (Girard, Kasumovic, & Elias, 2011) Follow up studies examining the physical structures of their rainbow iridescent scales has inspired new miniature optics (Hsiung et al., 2017) Innovations like these may have never come about without basic research focused on animal communication. Lifelike models of real animals, referred to as dummies, have been a valuable research tool for animal communication studies over the past hundred years (Woo & Rieucau, 2011). Dummies allow researchers to control and stage specific scenarios in experiments so that they may answer questions about animal response and communication. Scenarios can include inter action between conspecifics of the same or opposite sex, common prey, or common predators. The use of dummies in behavioral studies has occurred across various animal groups such as fish, birds, and wasps (Eriksson & Wallin, 1986; Karlson & Butenandt, 1959 ; Rowland, 1975). Tests with dummies have been administered in field experiments, aquariums, and lab settings (Eriksson & Wallin, 1986; Karlson & Butenandt, 1959; Rowland, 1975). Types of dummies used in experiments include robots, acrylic or silicon casti ngs, or even dead specimens (Eriksson & Wallin, 1986; Girard, Kasumovic, & Elias, 2011; Rowland, 1975). Dummies provide a variety of benefits for behavioral research that aims to understand the interaction of a test specimen when presented with a specific type of behavior from a conspecific. To begin, models allow for control of the desired behavior three dimensionally. This type of movement manipulation in turn allows a live test specimen to display behavior from any perspective in regard to the model. In addition to movement manipulation, patterns and coloration are also easily manipulated on a model. Manipulating the color of a live specimen causes concern for potential harm such as toxins, suffocation, or even squishing a specimen during manipulations.
3 P ainting models instead of a live specimen rids a study of such concerns. A model also projects coloration that does not fade and a firm exterior that cannot shrivel or be squished as could occur with a live or dead specimen. Models allow for easy manipulat ion of test environment as they can be used under any form of lighting and mitigate the introduction of unnatural items to a testing environment. Finally, models also eliminate chemical cues that have been shown to entice courtship behavior (R. J. Clark & Jackson, 1995). Overall, models allow for behavioral studies that are completely dependent on visual cues alone. Currently, the only known form of dummy usage in spider research is manipulation of a dead specimen (Girard et al., 2011; Jackson & Ha rding, 1982; Sullivan Beckers & Hebets, 2011). The use of dummy spiders has provided opportunities for research on courtship preferences and displays of males and females (Girard et al., 2011; Jackson & Harding, 1982; Sullivan Beckers & Hebets, 2011). Whil e a dead specimen presents a potentially viable visual cue directly after death, it has the potential to maintain chemical cues that might influence a receiver's behavior (Kaston, 1936). Moreover, the abdomens of dead spider specimens dry out, shrivel, and their legs curl after death, meaning that only carefully positioned and freshly killed specimens maintain a lifelike appearance. These types of specimens must be discarded immediately after use. A solution to the concerns of dead specimen dummies has bee n the use of animations presented to receivers on video screens (D. L. Clark & Uetz, 1992). Animations allow greater manipulation of courtship display such as angle of leg raising, speed of a display, and color manipulation. However, there are potential dr awbacks of animations. For example, computer screens used to display animations to spiders have been developed for human vision, not spider vision, which may affect a spider's perception of them (Fleishman & Endler, 2000).
4 Species from the family Salticidae are ideal test specimens for testing the interactions of spiders with lifelike models. Spiders of this family are known for having highly sophisticated vision (Harland, Jackson, & Macnab, 1999). Members of Salticidae have been shown to be able t o distinguish between different colors and there is substantial variation in color vision across the family (Nakamura & Yamashita, 2000). In addition, species within this family are able to distinguish between different types and speeds of movement (Komiya Yamashita, & Tateda, 1988). Though vision is a prominent sense used by jumping spiders, other cues such as chemical and vibratory cues are used in tandem with movement and color in courtship displays (R. J. Clark & Jackson, 1995; Girard et al., 2011). E ven with multiple cues being involved in courtship, the species Phidippus johnsoni is shown to be most dependent on visual cues when a male encounters a female outside of her nest (Jackson, 1978). The closely related species, Phidippus regius has an estab lished population in Gainesville, Florida making it an ideal species for collection and testing. P. regius is also one of the largest species of the family Salticidae making it easy to replicate the minute details of the species in hand painted 3D models. The goal of this project was twofold. First, we used microCT scanning and 3D printing technology to create realistic hand painted 3D models of male and female P. regius Second, we went on to determine whether live P. regius test spiders would treat these models the same as they would a live conspecific (and different from a prey item). Because spiders of closely related species have been shown to display to both dead specimen dummies and animations, it is expected that test specimen would also treat a lif elike model as if it were a live spider. If spiders interact with models as if they are real, this would provide numerous opportunities for future experiments on jumping spider behavior. A lifelike model, unlike a dead model or an animation, can allow test s that manipulate color pattern while also manipulating 3 dimensional movement. In addition,
5 chemical and vibratory cues could be used alongside a painted model to better understand how the presence of multiple manipulated cues can entice different behavio rs. To our knowledge, this is the first study to use 3D scanning and printing technology to create spider dummies. Methods Males and females (n=18, 9 males, 9 females) of the species Phidippus regius were collected in natural areas of Gainesville, Florida (Table 1). Each spider was housed individually in the lab for at least two weeks before scanning or testing. Each spider had reached sexual maturity before undergoing testing. Specimens were fed cric kets approximately equivalent to their own body size three times per week and provided water daily. Model Preparation An adult male and a female of the species P. regius were knocked out using gaseous carbon dioxide. Once knocked out, the specimen was ma nipulated into a natural position on a piece of polyethylene foam to ensure the legs remained spread from the body. The pedipalps were also pulled down away from the chelicerae to distinguish the separation of the two body parts and to ensure minimal movem ent during scanning. The piece of foam was then placed in the bottom of a deli cup and a piece of cotton was secured over the top of the specimen to help maintain the positioning of the legs. The specimen was then placed in a freezer for one to two days. T he deli cups were transferred to the University of Florida 's Nanoscale Research Facility, where specimens were removed and prepared for scanning. The spiders were again manipulated into position on a piece of polyethylene foam that was fixed on the top end of a metal rod. Each specimen was given
6 an hour to settle on the end of the foam rod to minimize movement during scanning. Specimens were scanned using the Phoenix v|tome|xs Industrial High Resolution and X Ray System. Scans were prepared for 3D printing using Meshlab. Individual, life sized models of females and males were printed with resin by an Eden 260V 3D printer. Nine models (4 males, 5 females) were hand painted with Liquitex basics acrylic paint to match the color patterns observed within natural population where these spiders were collected (Figure 1). Arena Preparation A test arena was constructed using two large pieces of glass separated and surrounded by foam (Figure 2) Glass and foam were both spray painted with a tannish brown color to mimic colors of P. regius natural habitat. A strip of foam 1 inch wide was glued on level with two pieces of glass on opposing sides of the foam to make the arena platform (Elias, Sival inghem, Mason, Andrade, & Kasumovic, 2010). More foam was then used to create external boundaries of the arena platform. The boundaries were 21 inches by 13 inches with each of the two glass areas allowing for 10 inches by 13 inches of movement area. The a rena was raised 5 inches off of a table and surrounded by a cardboard barrier that rose 1 f oo t off of the table in all areas except the horizontal camera. In front of the horizontal camera the cardboard barrier rose 8 inches. Testing Each test spider (9 mature males and 9 mature females) underwent three separate tests: (1) with a live conspecific of the opposite sex, (2) with a live cricket, and (3) with a model conspecific
7 of the opposite sex. The goal of these tests was to determine whether t he test spider would respond to the model in the same way as it would respond to a live conspecific, and whether the test spider would differentiate between the model and a live cricket (a common prey item of P. regius ). Each male was randomly assigned to a female. Each female male pair was then randomly assigned the order in which they would go through testing. Each test was separated by at least one week. Before a test began, the arena and any other equipment used was cleaned with ethanol in order to elim inate any chemical cues that may have been left from a previous test. In addition, boundary foam was covered with petroleum jelly before each test to deter escape from the arena platform. Each specimen, model, and cricket used for a test was placed in the arena three to four centimeters away from the dividing foam. They would then be left under the petri dish for a five minute acclimation period. Each test involved a five minute acclimation period followed by a 20 minute testing period. All tests were video taped for subsequent analysis. Live Conspecific Testing After the acclimation period, the petri dishes were removed, and spiders were allowed to freely roam in the arena for twenty minutes. Sometimes the spiders would crawl upside down on the pe tri dish. In these circumstances, the petri dish would be flipped over and removed as soon as the spider was no longer touching the dish. Because male and female jumping spiders have both been shown to alter behavior based on the female's mating status, a male was displaced from a female before copulation could occur but after the male had touched the female (Jackson, 1980, 1981). Cricket Testing
8 Crickets were also placed under a petri dish during the five minute acclimation period and the spider and cricket were then a llowed to roam freely for 20 minutes. If the spider attacked and consumed the cricket, the trial ended. Model Testing Models were manipulated for testing by the use of magnets. Female models had a metal washer painted to match the arena and males had small magnets glued beneath the abdomen. The models were placed on top of the arena platform and a large magnet was manipulated beneath the arena platform in order to create movement with the model. Both male and female models were man ipulated to circle the petri dish during the five minute acclimation period. Once the acclimation period was completed, each model began a preplanned movement pattern based on previously observed behavior of each sex (Edwards, 1975). Female models moved in a figure eight pattern using the entirety of their half of the arena. Every time the model reached the separation foam, it would orient to the live male. If the live male were to cross the foam, the model would cease movement, other than orienting toward the live male, in order to mitigate any vibrations between the magnet and glass. Male models would also begin with a figure eight pattern for the first five minutes following the acclimation period. After the figure eight pattern, the male model would prov ide a courtship display by first orienting to the female and then by moving 1.5 cm to the left, then to the right, to the left again, and to the right once more in a zig zag pattern (Edwards, 1975). After the zig zag pattern was complete, the model would t hen lunge three times toward the female (Edwards, 1975). After this display was complete, the model would begin moving in a figure eight again. The model would repeat this pattern of courtship display and figure eight movement every 90 seconds until the en d of the allotted testing time. Male
9 models, like the female models, would also stop movement and orient to the female if she were to pass over the foam. Video Analysis Videos of each test were reviewed, and several behaviors were recorded (Tab le 2). The amount of time spent engaged in a behavior for both males and females was recorded for the following when directed toward the live conspecific, model, or cricket: orientation, leg waving, and stalking. Orientation was considered as a sudden, dra matic movement intended to make the primary eyes of the specimen face the test subject or any prolonged movement that allowed the specimen to remain with their primary eyes directed at the test subject. Leg waves were considered as any noncourtship movemen t of the front pair of legs. Stalking was considered as a spider lowering their abdomen and slowly advancing directly toward the test subject, as they typically do just before attacking prey. The number of times that a male or female attacked the test subj ect was also recorded. This was distinguished by a sudden advance toward the test subject that did not follow a courtship display and in which the specimen of interest is displaying their fangs. In addition, males had the amount of courtship lunges and the amount of time spent displaying courtship recorded. A courtship display was distinguished by a zig zag pattern in which the male also extended his front pair of legs out to the side. A lunge that followed this movement or in which the male advanced while holding his front pair of legs directly above its carapace was considered a courtship lunge. Finally, the first touch of a live specimen to a model was recorded and behavior exhibited before and after this touch were recorded separately. Statistical Anal ysis
10 We compared the continuous variables described above among the three treatment groups using individual ANOVA models and Tukey HSD post hoc comparisons. Because each test spider was exposed to all three treatments (dummy, live conspecific, cricket), individ ual ID was included as a random factor in the ANOVA models. We used logistic regression to determine if behavior differed in females before and after they had touched a model. Results Across all tests, the focal spiders oriented at least once to the treatment being presented (dummy, live conspecific, or cricket), even if just for a few seconds. Across the three treatment groups, males showed no difference in the amount of orientations, total time spent oriented, time spent stalking, or number of leg w aves (Table 3). Though stalking behavior of males across all treatments was similar, males were more likely to attack crickets than either the m odel or the live female (Table 3 Figure 3). In regard to courtship behavior, males showed no difference in the amount of displays init iated across treatments (Table 3 ). Our ANOVA model suggested that there were significant differences in courtship lunges between treatments, but none were significant in our Tukey post hoc tests. Males did, however, show a significan t difference in the amount of time spent engaged in courtship, with higher rates of courtship directed at live females compared with either the cricket or female model (Table 3 Figure 4). The females also showed no significant difference in amou nt of orientations or total time oriented between all three treatments (Table 4). Specifically, in regard to tests with a model, females showed no significant difference in orientation (p>0.60) or stalking (p>0.84) before and after the female touched the m od el for the first time Additionally, there was no significant
11 difference in leg waving behavior of females across tests (Table 4 ). Females did however show significant differences in total stalking time, amount of stalking initiations, and amou nt of att acks performed (Table 4 ). In all three cases the female had a higher rate of the respective aggressive behavior directed toward a cricket compared with either the model or the live male (Figures 5, 6, and 7). Discussion In our experiments, live m ales treated the dummy differently than the cricket ( suggesting that they recognized the model as something different than prey), but contrary to our expectations they did not treat the models exactly the same as the live female. This suggests that the app roach of using a 3D printed spider dummy has some potential, but additional techniques will need to be developed to increase the likelihood of a test spider interpreting the model as a live conspecific. Though certain behaviors were similar between all tes ts, males were shown to have differences in attacks and courtship displays. A male had a much higher likelihood of attacking a cricket. The live female and model had a lower, but similar likelihood, of enticing an attack out of a male. This suggests that t he male does not view either the female nor the model as a prey item. Though the male does not view the model and female as a prey item, this does not necessarily mean the male will view them both as potential mates. A male was just as likely to initiate a display to a cricket, model, and live female. Though it was equally as likely to initiate a display, a male had a much higher rate of continued courtship display toward a live female when compared to the cricket and model. This demonstrates a noticeable d ifference in the male behavior elicited from a live female and a model female. A movement based cue from a spider or cricket may be enough to entice the beginning of a courtship display from a male. However, it is understood that males are capable of
12 inter preting multiple cues when presenting a display, therefore the presentation of a secondary cue such as a chemical or a vibratory sequence may be necessary for a male to proceed with a display (R. J. Clark & Jackson, 1995; Girard et al., 2011). Though there is some similarity in the behaviors elicited from a male by a model and a live female, it is not conclusive enough to state that the male is treating a model and a female as if they were the same. Further testing using a female model in tandem with chemic al and vibratory cues may provide more evidence toward the model being a viable mimic of a live female. In tests with live females, the male dummy also showed some promise, but our study revealed limitations here as well. The most significant diffe rences noted in female behavior is demonstrated in aggressive behaviors. A female was more interested in initiating and continuing stalking behavior toward a cricket. In addition, a female was significantly more likely to attack at a cricket. Live males an d model males elicited a similarly low level of stalking and lunging from the live female. This suggests that the female does not see the live male or the model as a prey item. Though this is supportive of a model enticing similar behavior when compared to a live male, it is difficult to determine that a female will treat a model similarly in all situations. For instance, P. regius females do not have a deliberate and specialized receptivity display for a male that is attempting to court (Edwards, 1975). Re maining still and oriented is the only visible source of receptivity, but this may not even signify the female ha s interest in a male. Therefore, it is difficult to determine whether or not a female is observing a male model display similarly to how she vi ews a live male display. We can conclusively determine that both male models and female models are not viewed as a prey source by a live conspecific however, we cannot yet state whether or not a live spider would view these models as if they were a live conspecific. There could be multiple reasons as to
13 why a spider did not respond to the models similarly to a live spider. For instance, the live spider specimens we used in our tests were housed in the lab over a long period of time. This time peri od in the lab could have negatively affected the likelihood that these specimens engage in behavior typical of a specimen in the field (Carducci & Jakob, 2000). Additionally, though models were capable of having notable color patterns painted on them, it w as difficult to make them look exactly like the fine hairs of a jumping spider. Also, the models were manipulated to move three dimensionally through the space provided but did not have their front legs manipulated to move exactly like a live spider. These smaller scale visual aspects of a spider could be more noticeable to a spider than we had originally believed. Moreover, aside from appearance, the models do not provide prominent non visual cues that we know are important in jumping spider communication. The need for a chemical or vibratory cue could play an important role in notifying a conspecific of another spider's species and sex. These cues may be a required component in eliciting mating behavior and their absence may deter a spider from treating th e model similar to a live conspecific. Further testing in which a model is presented in tandem with other known cues could provide more evidence toward model viability. Acknowledgements 3D spider files were created by a t eam of students as part of UF's Insects Alive project (Ediel Dominguez, Annie Gormaley, Ja son Cochran, Nate Duerr, Shimul Chowdhury www.insectsalive.com ), in collaboration with the staff at UF' s Nanoscale Research Facility, and funded by a gran t from UF's Creative Campus Initiative (to Andrea Lucky and Lisa Taylor). Th ank you to the National Science Foundation for fu nding undergraduate research in the Taylor Lab. Thank you to members of the Taylor Lab for their help in car ing for spiders and
14 hel ping me to learn more throughout this process. Thank you to Gary Scheiffele at UF's Nanoscale Research Facility for helping acquire scans of the spiders. Thank you to Ediel Dominguez for creating high quality, printable files of the scans. Thank you to Ma nny Rodriguez of the FabLab for working to make the 3D printed models a reality. Thank you to Brett Taylor for his work painting the models so that they may look as lifelike as possible.
15 References Cited !"#$$%$&'()'*)'+)&',"-./0.1&'2)'3)&'34/$-4$&'5)&'6'7"%08'9:;#"<-4$&'+)'=>?@AB)'CDD;%::'E.0#$.#F'G/H' E?@PB)' +%0$;4Z'[#%.4.1'-[0D#"-'0$-[0"#'N0$0%A]@ R ^ Jackson, R. R. (1978). The Mating Strategy of Phidippus johnsoni (Araneae, Salticidae): I. Pursuit Time and Persistence. Behavioral Ecology and Sociobiology, 4 (2), 123 132. Jackson, R. R. (1980). The Mating Strategy of Phidippus johnsoni (Araneae, Salticidae): II. Sperm Competition and t he Function of Copulation. The Journal of Arachnology, 8 (3), 217 240. Jackson, R. R. (1981). Relationship between reproductive security and intersexual selection in a jumping spider Phidippus johnsoni (Araneae: Salticidae). Evolution, 35 (3), 601 604. doi: 10.1111/j.1558 5646.1981.tb04922.x Jackson, R. R., & Harding, D. P. (1982). Intraspecific interactions of Holoplatys sp. indet., a New Zealand jumping spider (Araneae: Salticidae). New Zealand Journal of Zoology, 9 (4), 487 510. doi:10.1080/03014223.1982.10 423881
16 Karlson, P., & Butenandt, A. (1959). Pheromones (ectohormones) in insects. Annual review of entomology, 4 (1), 39 58. Kaston, B. J. (1936). The senses involved in the courtship of some vagabond spiders. Entomologica Americana, N.S. 16 97 166. Komi ya, M., Yamashita, S., & Tateda, H. (1988). Turning reactions to real and apparent motion stimuli in the posterolateral eyes of jumping spiders. Journal of Comparative Physiology A, 163 (5), 585 592. doi:10.1007/BF00603842 Nakamura, T., & Yamashita, S. (200 0). Learning and discrimination of colored papers in jumping spiders (Araneae, Salticidae). Journal of Comparative Physiology A, 186 (9), 897 901. doi:10.1007/s003590000143 Rowland, W. J. (1975). The Effects of Dummy Size and Color On Behavioral Interaction in the Jewel Cichlid, Hemichromis Bimaculatus Gill. Behaviour, 53 (1), 109 125. doi:doi: https://doi.org/10.1163/1568 53975X00560 Sullivan Beckers, L., & Hebets, E. A. (2011). Modality specific experience with female feedback increases the efficacy of courtship signalling in male wolf spiders. Animal Behaviour, 82 (5), 1051 1057. doi: https://doi.org/10.1016/j.anbehav.2011.07.040 Woo, K. L., & Rieucau, G. (2011). From dummies to animations: a review of computer animated stimuli used in animal behavior studies. Behavioral Ecology and Sociobiology, 65 (9), 1671 1685. doi:10.1007/s00265 011 1226 y
17 Figure 1. (a) Image of a live Phidippus regius female. (b) Image of one of the painted 3D printed models. Live female and model photo taken by Brett Taylor.
18 Table 1. Collection information of each individual test spider. Table 2. Description of behaviors recorded during each test. Behaviors in groups a, b, and c were recorded in both males and females. Behavior group d was recorded only in males.
19 Figure 2. Test arena. The spider was placed on one side of the middle dividing foam and the subject of interest was placed on the opposite side. Both the spider and the subject of interest remained on their respective sides, under a petri dish for five minutes durin g the acclimation period before being allowed to roam freely for 20 minutes. The exterior foam rectangle was a half inch tall barrier and was covered in petroleum jelly to deter escape from test subjects.
20 Table 3. Results of ANOVA examining differences in live male behavior when presented with a live female, model, and cricket. Figure 3 A comparison of amount of attacks by live males on models, live females, and crickets.
21 Figure 4. A comparison of time spent displaying by a live male to a model, a live female, and a cricket. Table 4. Results of ANOVA examining differences in live female behavior when presented with a live male, model, and cricket.
22 Figure 5. Comparison of the amount of time a live female spent stalking a model, live male, and cricket.
23 Figure 6. Comparison of amount of stalking initiations by a live female toward a model, live male, and cricket.
24 Figure 7. Comparison of amount of attacks by a live female on a model, live male, and cricket.