Statistical Analysis
To analyze these data, I used binomial generalized linear models with a with a complementary log-log link function in R (ver. 1.1.463). For all models, I set avian predation as the binomial response variable (0 = not attacked, 1 = attacked) and included days exposed as an offset term. For the primary model, I included the following independent variables: trial, location, leaf roll treatment (rolled, open), color treatment (eyespotted, non-eyespotted), canopy openness (a proportion, from 0 = sky fully obscured to 1 = sky not obscured by anything), and plant height (in cm). I also tested for an interaction between the roll treatment and color treatment. To determine the statistical significance of each independent variable across the model, I compared the full model to models lacking the term of interest using likelihood ratio tests (package lmtest). To determine the simple effects of eyespots and leaf rolls, without the interaction effect, I constructed 4 additional models using data from (1) only eyespotted prey, (2) only non-eyespotted prey, (3) only leaf-rolled prey, and (4) only open-leaf prey. For models (1) and (2), I compared full models to models without the leaf roll treatment; for models (3) and (4), I compared full models to models without the color treatment. These comparisons were also made using likelihood ratio tests.
RESULTS
The overall avian predation rate was 13.4% of artificial caterpillars over a 5-day period. Predation did not vary significantly by trial (χ2 = 0.79, df = 1, p = 0.38), location (χ2 = 0.55, df = 1, p = 0.46), canopy openness (χ2 < 0.01, df = 1, p = 0.95), or plant size (χ2 = 0.71, df = 1, p = 0.40). Leaf rolls significantly reduced predation relative to prey on open leaves (12.9% reduction, χ2 = 24.43, df = 1, p < 0.001). This was true of both eyespotted (17.3% reduction, χ2 = 25.77, df = 1, p < 0.001) and non-eyespotted (8.4% reduction, χ2 = 4.54, df= 1, p < 0.05) prey. Eyespots alone had no significant effect on predation (χ2 = 0.91, df = 1, p = 0.34), though this was only true for prey on open leaves (χ2= 0.13, df = 1, p = 0.72). There was a significant interaction between leaf rolls and eyespots (χ2 = 5.96, df= 1, p < 0.05): in leaf rolls, eyespots reduced the probability of predation (7.1% reduction, χ2 = 6.98,df = 1, p < 0.01).
DISCUSSION
In this study, leaf rolls not only protected prey in general, but also increased the effectiveness of eyespots at deterring visual predators (Fig. 2). This result supports the environmental deimatism hypothesis, i.e., the active use of objects in the environment to create a deimatic display. Similar to the mounds built by bowerbirds for sexual communication (Endler et al. 2010), leaf rolls are not just passive “backgrounds,” but an integral part of the visual signal’s success. While environmental modification has been shown to effectively complement (or serve as) visual camouflage in several systems (Canfield 2009; Hultgren & Stachowicz 2011), these behaviors are not well-documented in other types of visual defense (e.g. aposematism, mimicry, masquerade, deimatism). Environmental deimatism may be a common strategy among other leaf-sheltering organisms, such as eyespotted hesperiid caterpillars (Janzen 2010) or spiders that abruptly jump out of rolls when disturbed (Postema pers. observations, Fig. 3). Beyond leaf rolls, environmental deimatism may exist as a more general strategy for shelter-using or -building species.
The overall protective effect of leaf rolls supports past experimental work on these structures’ role in predator defense (Murakami 1999; Tvardikova & Novotny 2012). However, the mechanism of protection is not entirely clear. One possibility is that leaf rolls physically hide the organism from detection. However, folded leaves are relatively noticeable against non-folded foliage; in some systems, leaf rolls even seem to act as a target for visually-oriented predators (Kobayashiet al. 2020). Naturally occurring leaf rolls in the study area were common, and often occupied by living organisms: over half (56%) of surveyed plants had at least one leaf roll, and over a third (34%) of rolls were occupied. Given their frequency and apparent profitability, it seems advantageous for avian predators to learn to search for prey in leaf rolls. However, the organisms inhabiting leaf rolls were not necessarily desirable prey items (Fig. 3). Surveyed rolls most commonly contained small, fast-moving spiders (49% of occupied rolls). Spiders often jumped from the roll immediately when disturbed, potentially making prey startling, hard to catch, or both. Other common prey items included very small organisms (e.g. springtails) and small weevils that often dropped to the ground when disturbed. Larger, less mobile, and more profitable prey – e.g. caterpillars – were rare (found in only ~4% of occupied rolls). The main defensive function of the leaf roll, then, may be to conceal prey identity. The added ambiguity and handling time of leaf rolls may make them relatively low-value foraging microhabitats, though this likely depends on the predator community’s degree of specialization, and perhaps temporal shifts in leaf roll abundance/occupancy. Predator uncertainty could further enhance the effectiveness of unexpected or startling visual signals as well.
Eyespots did not significantly increase predation risk on open leaves relative to non-eyespotted prey on open leaves (Fig. 2). This could suggest that prey with large eyespots are not more detectable to visual predators than prey without eyespots. Given that birds are highly attuned to eye-like stimuli, this seems unlikely (De Bona et al.2015). Alternatively, readily visible eyespots may be more detectable to predators, but simultaneously function to deter predators at a distance. In this scenario, the combined effects of eyespots (increased detection and predator deterrence) may be counterbalanced. This hypothesis is supported by the fact that the majority of Papilio species that possess eyespots rest on open leaves, as well as the general positive association between body size and presence of eyespots in lepidopteran larvae (Wagner 2005; Hossie et al. 2015; Gaitonde et al.2018). The fear of paired, eye-like patterns appears to be relatively innate for avian predators (Merilaita et al. 2011). This contrasts with other conspicuous color strategies, such as aposematism, where learning is more central to the pattern’s antipredator effect (Hämäläinen et al. 2020). If eye mimicry does not require predators to have prior negative experience with the “model” organism, then costs of being conspicuous due to encounters with naïve predators may be minimal.
Habitat characteristics may have also played a role in the perception and effectiveness of eyespots in this study. Both field sites were densely vegetated, with an average canopy openness of 14.3% (SD: 8.3%) – i.e, ~86% of the area above each artificial caterpillar was covered by vegetation. In complex, highly vegetated, and low-light environments, it may be difficult for predators to distinguish between real and fake eyes, or it may be too risky to spend a long time investigating (Janzen et al. 2010). This may also help to explain why eyespots did not significantly increase predation, despite presumably higher predator detection, compared to non-eyespotted prey on open leaves (Fig. 2). While there was no direct support for the influence of background conditions (such as canopy openness and plant height) on avian predation in this study, I did not experimentally manipulate these characteristics. In other studies of visual signalling, habitat heterogeneity, vegetation density, and lighting conditions have had effects on the perception of animal color patterns (Gotceitas & Colgan 1989; Endler 1993; Coker et al. 2009; Seymoure et al. 2018). To better understand the effect of environmental context on the perception of eyespots, it would be useful to directly observe predator responses to eyespotted and non-eyespotted prey across various habitat types.
It may be useful to consider P. troilus leaf rolls as an example of Dawkin’s “extended phenotype” (1999). There are clear consequences of the leaf roll on caterpillar fitness, as well as synergistic interactions between leaf rolling and color traits (Fig. 2). In this system, selection is acting on multiple interacting levels: on the structure of the roll, the expression of leaf-rolling behavior, and the organism’s color patterns (Laland 2004; Hunter 2018). This makes the evolution of environmental deimatism a question of both morphology and behavior. Umbers et al (2017) suggest two potential pathways for how deimatic displays evolve: the “defense-first” and “startle-first” hypotheses. In the former, initially cryptic prey gain constitutive defenses (e.g., toxins), which then selects for conspicuous color patterns to advertise toxicity, and finally a concealing mechanism to create the “startle” effect. In the latter, initially cryptic prey develop a sudden movement that deters predators, which is later enhanced by a conspicuous visual component (and additional chemical defenses, in some species). Given that P. troilus larvae are generally considered non-toxic (Wagner 2005), the “startle-first” hypothesis may be more likely. Via this pathway, we would expect larvae to have evolved the leaf-rolling behavior (a proxy for the “sudden movement”) before the development of large, conspicuous eyespots. It is less likely that leaf-rolling developed simply as a way to conceal conspicuous eyespots, as there were no obvious detectability costs of eyespots for prey on open leaves (Fig. 2). This aligns with Schaedlin and Taborsky’s (2009) observation that external structures involved in signalling often provide an initial, direct fitness benefit to the signaler, that then selects for a progressively stronger signal. A phylogenetic comparative study, tracking both color traits and deimatic behaviors across the evolutionary history of swallowtails and/or other relevant lepidopteran groups, could potentially clarify when and how the behavior-morphology pairing arose (Janzen et al. 2010; Vidal-García et al.2020).
Given that leaf-rolling is an effective antipredator strategy forP. troilus larvae, and appears to work synergistically with the species’ defensive color strategy (Fig. 2), why is leaf-rolling not observed more generally across swallowtails? One possible constraint is the time and energy investment involved in constructing multiple leaf rolls over the course of larval development. After larvae lay down layers of silk, leaves may take over an hour to fully fold into a roll (supplemental video 1). These periods of high activity and potential exposure to predators are not accounted for in this study, but may temper the antipredator benefit of leaf rolls. Secondly, some host plants may not be conducive to the formation of leaf rolls. The leaves of common P. troilus host plants are relatively thin, wide, and flexible compared to common host plants of other eyespotted swallowtail species (e.g., Populus spp., Salix spp.; Wagner 2005). While many Papilio larvae form Velcro-like silk pads to rest on, the leaves of their host plants may be too stiff, thick, or narrow to easily fold into full leaf rolls. Larvae in the swallowtail family (Papilionidae) use a diverse array of host plants, and their later-instar color defenses correspond closely to evolutionary shifts in host plant usage – e.g., aposematism has mainly evolved in larvae that use narrow-leafed, toxic plants, while cryptic or mimetic strategies are associated with more dense, nontoxic plants (Gaitonde et al. 2018). It would be worth investigating how other aspects of host plant morphology (particularly leaf width and thickness) may have shaped the evolution of leaf-rolling, deimatism, and color traits among insects (Janzen et al. 2010).
The results of this study provide support for the environmental deimatism hypothesis, and, more generally, the key role of behavior in defensive visual signals (Ruxton et al. 2009; Cuthill et al. 2017; Stevens & Ruxton 2018). They also suggest that deimatic displays can arise without strong costs to conspicuousness, though this likely depends on the mechanism of predator deterrence (learned vs. reflexive avoidance). To better understand the ecology and evolution of defensive visual signals, it is essential to consider color patterns less as static characters, and more as “multivariate optima”; i.e., complex strategies that may involve selection on morphology, behavior, and/or extended phenotypes beyond the body of the organism (Dawkins 1999, Laland 2004, Cuthill et al. 2017, Stuart-Fox 2022, Postema et al. 2022).
ACKNOWLEDGEMENTS
Thank you to Louie Yang, Andy Sih, and Rick Karban for their assistance in developing the experimental design. Additional thanks to Hee Jin Chung, Gwen Erdosh, Kirsten Sheehy, and Lohit Garikpati for help with constructing clay caterpillars; to Louie Yang and Tracie Hayes for comments on the manuscript; and to Pavel G. for his assistance with fieldwork. Caterpillar molds were designed and printed with the help of Tez Stair and Steven Lucero. Caterpillar icons were created by Mia Lippey. Spectral measurements were taken thanks to Alison Rabosky and Hayley Crowell. All study sites were used with permission from the Ann Arbor Parks and Recreation Department and the University of Michigan.
FUNDING INFORMATION
This project was funded by the Phi Beta Kappa Northern California Association graduate scholarship, UC Davis Jastro-Shields graduate research award, and NSF IntBIO Collaborative Research grant (IOS-2128245).
REFERENCES
Aronsson, M. & Gamberale-Stille, G. (2009). Importance of internal pattern contrast and contrast against the background in aposematic signals. Behav Ecol , 20, 1356–1362.
Badiane, A., Carazo, P., Price-Rees, S.J., Ferrando-Bernal, M. & Whiting, M.J. (2018). Why blue tongue? A potential UV-based deimatic display in a lizard. Behav Ecol Sociobiol , 72, 1-11.
Barnett, J.B., Cuthill, I.C. & Scott-Samuel, N.E. (2018). Distance-dependent aposematism and camouflage in the cinnabar moth caterpillar (Tyria jacobaeae , Erebidae). R Soc Open Sci , 5, 171396.
Barnett, J.B., Redfern, A.S., Bhattacharyya-Dickson, R., Clifton, O., Courty, T., Ho, T., et al. (2017). Stripes for warning and stripes for hiding: spatial frequency and detection distance.Behav Ecol , 28, 373–381.
Blake, J.G., & Hoppes, W.G. (1986). Influence of resource abundance on use of tree-fall gaps by
birds in an isolated woodlot. Auk, 103, 328-340.
Canfield, M.R. (2009). The double cloak of invisibility: phenotypic plasticity and larval decoration in a geometrid moth, Synchlora frondaria , across three diet treatments. Ecol Entomol , 34, 412–414.
Caro, T. & Allen, W.L. (2017). Interspecific visual signalling in animals and plants: a functional classification. Philos Trans R Soc Lond B Biol Sci , 372, 20160344.
Coker, D.J., Pratchett, M.S. & Munday, P.L. (2009). Coral bleaching and habitat degradation increase susceptibility to predation for coral-dwelling fishes. Behav Ecol , 20, 1204–1210.
Cuthill, I.C., Allen, W.L., Arbuckle, K., Caspers, B., Chaplin, G., Hauber, M.E., et al. (2017). The biology of color.Science , 357, eaan0221.
Dawkins, R. (1999). The Extended Phenotype: The Long Reach of the Gene . Revised edition. Oxford University Press, Oxford, UK.
De Bona, S., Valkonen, J.K., López-Sepulcre, A. & Mappes, J. (2015). Predator mimicry, not conspicuousness, explains the efficacy of butterfly eyespots. Proc Royal Soc B , 282, 20150202.
Dookie, A.L., Young, C.A., Lamothe, G., Schoenle, L.A. & Yack, J.E. (2017). Why do caterpillars whistle at birds? Insect defence sounds startle avian predators. Behav Processes , 138, 58–66.
Drinkwater, E., Allen, W.L., Endler, J.A., Hanlon, R.T., Holmes, G., Homziak, N.T., et al. (2022). A synthesis of deimatic behaviour.Biol Rev , 97, 2237-2267.
Endler, J.A. (1983). Natural and sexual selection on color patterns in poeciliid fishes. Environ Biol Fish , 9, 173–190.
Endler, J.A. (1993). The color of light in gorests and its implications.Ecol Monogr , 63, 2–27.
Endler, J.A., Endler, L.C. & Doerr, N.R. (2010). Great bowerbirds create theaters with forced perspective when seen by their Audience.Curr Biol , 20, 1679–1684.
Fabricant, S.A. & Herberstein, M.E. (2015). Hidden in plain orange: aposematic coloration is cryptic to a colorblind insect predator.Behav Ecol , 26, 38–44.
Gaitonde, N., Joshi, J. & Kunte, K. (2018). Evolution of ontogenetic change in color defenses of swallowtail butterflies. Ecol Evol , 8, 9751–9763.
Gotceitas, V. & Colgan, P. (1989). Predator foraging success and habitat complexity: quantitative test of the threshold hypothesis.Oecologia , 80, 158–166.
Halfwerk, W., Jones, P.L., Taylor, R.C., Ryan, M.J. & Page, R.A. (2014). Risky ripples allow bats and frogs to eavesdrop on a multisensory sexual display. Science , 343, 413–416.
Hämäläinen, L., Mappes, J., Rowland, H.M., Teichmann, M. & Thorogood, R. (2020). Social learning within and across predator species reduces attacks on novel aposematic prey. J Anim Ecol , 89, 1153–1164.
Hossie, T.J. & Sherratt, T.N. (2012). Eyespots interact with body colour to protect caterpillar-like prey from avian predators. Anim Behav , 84, 167–173.
Hossie, T.J. & Sherratt, T.N. (2013). Defensive posture and eyespots deter avian predators from attacking caterpillar models. Anim Behav , 86, 383–389.
Hossie, T.J. & Sherratt, T.N. (2014). Does defensive posture increase mimetic fidelity of caterpillars with eyespots to their putative snake models? Curr Zool , 60, 76–89.
Hossie, T.J., Sherratt, T.N., Janzen, D.H. & Hallwachs, W. (2013). An eyespot that “blinks”: an open and shut case of eye mimicry inEumorpha caterpillars (Lepidoptera: Sphingidae). J Nat Hist , 47, 2915–2926.
Hossie, T.J., Skelhorn, J., Breinholt, J.W., Kawahara, A.Y. & Sherratt, T.N. (2015). Body size affects the evolution of eyespots in caterpillars. PNAS , 112, 201415121.
Hultgren, K. & Stachowicz, J. (2011). Camouflage in decorator crabs: integrating ecological, behavioural and evolutionary approaches. In:Animal Camouflage: Mechanisms and Function , [eds. Stevens, M., & S. Merilaita], Cambridge University Press, Cambridge, UK, pp. 212–236.
Hunter, P. (2018). The revival of the extended phenotype. EMBO Rep , 19, e46477.
Janzen, D.H., Hallwachs, W. & Burns, J.M. (2010). A tropical horde of counterfeit predator eyes. PNAS , 107, 11659–11665.
Kobayashi, C., Matsuo, K. & Kawata, M. (2020). Contradictory effects of leaf rolls in a leaf-mining weevil. Sci Rep , 10, 12180.
Kobayashi, C., Matsuo, K., Watanabe, K., Nagata, N., Suzuki-Ohno, Y., Kawata, M., et al. (2015). Arms race between leaf rollers and parasitoids: diversification of plant-manipulation behavior and its consequences. Ecol Monogr , 85, 253–268.
Laland, K.N. (2004). Extending the extended phenotype. Biol Philos , 19, 313–325.
Leslie, A.J. & Fleming, N. (1990). Role of the osmeterial gland in swallowtail larvae (Papilionidae) in defense against an avian predator.J Lepid Soc , 44, 245-251.
de Lira, J.J.P.R., Perez-Jvostov, F., Gotanda, K.M., Kou-Giesbrecht, S., Pease, S.K., Jackson, M., et al. (2018). Testing for a whole-organism trade-off between natural and sexual selection: are the male guppies preferred by females more likely to be eaten by predators?Evol Ecol Res , 19, 441–453.
Mappes, J., Kokko, H., Ojala, K. & Lindström, L. (2014). Seasonal changes in predator community switch the direction of selection for prey defences. Nat Commun , 5, 5016.
Marshall, N.J. (2000). Communication and camouflage with the same ‘bright’ colours in reef fishes. Philos Trans R Soc Lond B Biol Sci , 355, 1243–1248.
Merilaita, S., Vallin, A., Kodandaramaiah, U., Dimitrova, M., Ruuskanen, S. & Laaksonen, T. (2011). Number of eyespots and their intimidating effect on naïve predators in the peacock butterfly. Behav Ecol , 22, 1326–1331.
Murakami, M. (1999). Effect of avian predation on survival of leaf-rolling lepidopterous larvae. Pop Ecol , 41, 135–138.
Nyffeler, M., Şekercioğlu, Ç.H. & Whelan, C.J. (2018). Insectivorous birds consume an estimated 400–500 million tons of prey annually.Sci Nat , 105, 47.
Postema, E.G. (2022). The effectiveness of eyespots and masquerade in protecting artificial prey across ontogenetic and seasonal shifts.Curr Zool , 68, 451-458.
Postema, E.G., Lippey, M.K. & Armstrong-Ingram, T. (2022). Color under pressure: how multiple factors shape defensive coloration. Behav Ecol , XX, 1-13.
Prudic, K.L., Skemp, A.K. & Papaj, D.R. (2007). Aposematic coloration, luminance contrast, and the benefits of conspicuousness. Behav Ecol , 18, 41–46.
Richards, L.A, & Coley, P.D. (2008). Combined effects of host plant quality and predation on a tropical lepidopteran: a comparison between treefall gaps and the understory in Panama. Biotropica , 40, 736-741.
Romero, G.Q., Gonçalves-Souza, T., Roslin, T., Marquis, R.J., Marino, N.A.C., Novotny, V., et al. (2022). Climate variability and aridity modulate the role of leaf shelters for arthropods: A global experiment. Glob Change Biol, 28, 3694–3710.
Rowe, C. (1999). Receiver psychology and the evolution of multicomponent signals. Anim Behav , 58, 921–931.
Rowland, H.M., Speed, M.P., Ruxton, G.D., Edmunds, M., Stevens, M. & Harvey, I.F. (2007). Countershading enhances cryptic protection: an experiment with wild birds and artificial prey. Anim Behav , 74, 1249–1258.
Ruxton, G.D., Speed, M.P. & Broom, M. (2009). Identifying the ecological conditions that select for intermediate levels of aposematic signalling. Evol Ecol , 23, 491–501.
Schaedelin, F.C. & Taborsky, M. (2009). Extended phenotypes as signals.Biol Rev , 84, 293–313.
Seymoure, B.M., Raymundo, A., McGraw, K.J., McMillan, W.O. & Rutowski, R.L. (2018). Environment-dependent attack rates of cryptic and aposematic butterflies. Curr Zool , 64, 663–669.
Skelhorn, J., Holmes, G.G., Hossie, T.J. & Sherratt, T.N. (2016a). Eyespots. Curr Biol , 26, 52–54.
Skelhorn, J., Holmes, G.G. & Rowe, C. (2016b). Deimatic or aposematic?Anim Behav , 113, 1–3.
Stevens, M. & Ruxton, G.D. (2018). The key role of behaviour in animal camouflage. Biol Rev , 94, 116-134.
Stuart-Fox, D. (2022). Defensive coloration as a multivariate optimum: a comment on Postema et al. Behav Ecol , arac065.
Stuart-Fox, D. & Moussalli, A. (2008). Selection for social signalling drives the evolution of chameleon colour change. PLoS Biol, 6, 22–29.
Tan, E., Reid, C., Symonds, M., Jurado-Rivera, J. & Elgar, M. (2017). The role of life-history and ecology in the evolution of color patterns in Australian chrysomeline beetles. Front Ecol Evol , 5, 1-15.
Tvardikova, K. & Novotny, V. (2012). Predation on exposed and leaf-rolling artificial caterpillars in tropical forests of Papua New Guinea. J Trop Ecol , 28, 331–341.
Umbers, K.D.L., De Bona, S., White, T.E., Lehtonen, J., Mappes, J. & Endler, J.A. (2017). Deimatism: a neglected component of antipredator defence. Biol Lett , 13, 20160936.
Umbers, K.D.L., Lehtonen, J. & Mappes, J. (2015). Deimatic displays.Curr Biol , 25, 58–59.
Umbers, K.D.L. & Mappes, J. (2015). Postattack deimatic display in the mountain katydid, Acripeza reticulata . Anim Behav , 100, 68–73.
Umbers, K.D.L. & Mappes, J. (2016). Towards a tractable working hypothesis for deimatic displays. Anim Behav , 113, 5–7.
Umbers, K.D.L., White, T.E., De Bona, S., Haff, T., Ryeland, J., Drinkwater, E., et al. (2019). The protective value of a defensive display varies with the experience of wild predators.Sci Rep , 9, 463.
Vidal-García, M., O’Hanlon, J.C., Svenson, G.J. & Umbers, K.D.L. (2020). The evolution of startle displays: a case study in praying mantises. Proc R Soc B: Biol Sci, 287, 20201016.
Wagner, D.L. (2005). Caterpillars of Eastern North America: A Guide to Identification and Natural History . Princeton University Press, Princeton, N.J.
Whiting, M.J., Noble, D.W.A. & Qi, Y. (2022). A potential deimatic display revealed in a lizard. Biol J Linn Soc, 136, 455–465.
Figure 1. (a) A live spicebush swallowtail (Papilio troilus ) larva on sassafras (Sassafras albidum ). Its leaf roll is held open, with strands of silk visible above the head. Eyespotted (b) and non-eyespotted (c) artificial larvae on open S. albidum leaves. (d) A true leaf roll with a live P. troiluslarva inside. (e) An artificial leaf roll with an artificial larva inside. (f) A live P. troilus larva in a leaf roll, its eyespots partially visible up-close. Photographs by EGP.
Figure 2. Mean proportions of artificial prey in each leaf roll treatment group (rolled versus open) attacked by avian predators, ± SE (n = 659). Yellow points represent eyespotted prey, while green points represent non-eyespotted prey. Illustrations by Mia Lippey.
Figure 3. (Left) Total counts of each organism type found in surveys of naturally occurring leaf rolls (n = 464). Within each organism category, counts of individuals that displayed escape behaviors in response to the leaf roll being disturbed (dropping, jumping, or no response) are represented in yellow, red, and brown, respectively. (Right) Examples of naturally occurring leaf rolls I observed in the field; leaf rolls varied in size, structure, and plant species. Photographs and illustrations by EGP.