1. INTRODUCTION

Spiders are among the most common and abundant predators in terrestrial ecosystems (Coddington & Levi 1991; Nyffeler & Sunderland 2003). With >49,000 described species, these animals exhibit an enormous diversity of lifestyles and foraging strategies (Nyffeler & Birkhofer 2017; World Spider Catalog 2020). While spiders had been believed for a long time to depend almost exclusively on live insects or other small arthropods for food, it has been shown in more recent years that the spiders' feeding habits are much more diverse than previously thought. The natural diets of the spiders include “odd foods” such as earthworms (Annelida), velvet worms (Onychophora), bristle worms (Polychaetes), slugs (shell-less Gastropoda), snails (shelled Gastropoda), amphipods (Talitridae), shrimps (Palaemonidae), crayfish (Cambaridae), freshwater crabs (Trichodactylidae), and a large variety of different types of plant materials (Nyffeler & Symondson 2001; Nyffeler et al. 2001, 2016, 2017b). Furthermore, spiders capture a variety of small vertebrates including birds (Aves), bats (Chiroptera), mice (Muridae), deer mice (Cricetidae), voles (Cricetidae), shrews (Soricidae), rats (Muridae), mouse lemurs (Cheirogaleidae), mouse opossums (Didelphidae), pygmy possums (Burramyidae), fish (Osteichthyes), frogs (Anura), toads (Anura), snakes (Serpentes), lizards (Squamata), newts (Salamandridae), lungless salamanders (Plethodontidae), mole salamanders (Ambystomatida), and caecilians (Caeciliidae) (see Nyffeler & Altig 2020).

Snakes are occasionally overpowered and eaten by a limited number of spider taxa (i.e., ophiophagy; McCormick & Polis 1982). Although ophiophagy in spiders is well documented (e.g., Emerton 1926; Pinkus 1932; McCormick & Polis 1982; Raven & Gallon 1987; Punzo & Henderson 1999), a synthesis focusing on this specific topic has not yet been published. Here, we compile and review all that is currently known on snake predation by spiders.

2. METHODS

2.1 Data collection.—We searched published reports on snake predation by spiders with Thomson-Reuters database, Elsevier's Scopus database, Google Search, Google Scholar, Google Books, and Google Pictures as well as ProQuest Dissertations and Theses (compare Nyffeler et al. 2017c). Social media sites were searched as well. We made a library search of books and scientific journals not included in the large databases. A total of 319 reports of predation (or predation attempts) on snakes by spiders were found, about 1/3 of which had previously been published in the scientific literature (see Supplementary  Table S1 (arac-49-01-06_s01.pdf), online at  https://doi.org/10.1636/JoA-S-20-050.s1). The remaining 2/3 were found on social media sites (e.g., Scientific American, National Geographic, Australian Geographic, BBC, YouTube, Project Noah, and Panama Birds & Wildlife; see Supplementary  Table S1 (arac-49-01-06_s01.pdf)) or provided to us as personal communication by scientists/photographers. Of the 319 records, 297 (93%) refer to naturally occurring incidents witnessed in the field or in buildings; the remaining 22 records (7%) are based on laboratory feeding trials or staged field experiments. In one case where it was stated that a certain type of predation event had been witnessed several dozen times (without providing an exact number; Peters 2016), we hypothetically estimated that this type of predation event had occurred two dozen times (as a proxy to the unknown true number). In this paper, the Neotropic realm is understood as the combined area of Mexico, Central America, South America, and the Caribbean, whereas all information referring to Florida is included in the USA.

2.2 Identification of unidentified spiders and snakes depicted in photos.—Almost 80% of the 319 reported predation events were documented with photos or videos. Some photos or videos depicted unidentified spiders and/or snakes which were identified to the lowest taxon possible on our behalf by established spider and snake taxonomists. Spiders were identified by R. Bennett, A. Dippenaar-Schoeman, G.B. Edwards, H. Höfer, P. Jäger, K. Kissane, Y. Lubin, R. Raven, B. Thaler-Knoflach, and R. West. Snake identifications were carried out by W. Gibbons, W. Lamar, R. Shine, S. Spawls, G. Vogel, and R. Whitaker. Some snakes unable to be unambiguously identified (based on only photos) are indicated in Appendix 2 by a question mark behind the snake's name, but the cumulative data provide an estimate of the number of snake morphospecies preyed on by spiders.

Several reports from the northeastern part of USA (Vermont, Connecticut, Maryland, Pennsylvania, and New York) or the southeastern part of Canada (Ontario) of unidentified theridiids killing snakes in a manner typical of black widows are from a geographic area that coincides with the geographic range of Latrodectus variolus Walckenaer, 1837 (see  https://usaspiders.com/latrodectus-variolus-northern-black-widow/) suggesting that in some of these cases, the predator in question was most likely the northern black widow Latrodectus variolus.

Snake and spider nomenclatures are based on Uetz et al.'s “The Reptile Database – update from 2020” and the “World Spider Catalog 2020”, respectively. In 89% cases, at least a family-level identification of the reported snake victims could be conducted, whereas 11% remain unidentified. Contrary to the “World Spider Catalog 2020”, we placed the genera Nephila Leach, 1815 and Trichonephila Dahl, 1911 in the family Nephilidae (sensu Kuntner et al. 2019).

Body length of spiders is understood as the length of cephalothorax + abdomen (without legs). Snake body length is defined as total length (= snout-vent length + tail length). In cases where information of spider body length was missing in published reports, average spider body length values representative of those species, extracted from arachnology books, were added to the data set in the Supplementary  Table S1 (arac-49-01-06_s01.pdf). So, for instance, an average spider body length value of 1.0 cm was assumed for adult female black widows (Latrodectus spp.). Spider body length data used in this paper are biased towards the araneomorph spiders (i.e., in particular the small-sized theridiids) due to the fact that in many reports referring to the much larger-sized mygalomorphs body length data were unavailable. Furthermore, because mygalomorph species show wide intraspecific variation in body size, it was usually not possible for this spider group to operate with an assumed species-specific body length value based on arachnology books. In very few cases, where exclusively snout-vent length data for snakes were available, those were converted to total length data, using conversion factors taken from the literature (Boada et al. 2005; Brust 2013; Mirtschin et al. 2017; Murphy et al. 2019).

Based on snake body length information, corresponding snake body mass values have been roughly estimated based on published snake length/body mass ratios (see Nyffeler et al. 2017a) if snake body mass was not provided in the published predation accounts.

2.3 Definition of the term “venomous spiders”.—All spider species except for the uloborids and the holarchaeids possess a pair of venom glands and are therefore capable of producing venoms used for prey capture and in some cases also for defense (Coddington & Levi 1991; Foelix 2011). A limited number of notoriously dangerous spiders are referred to as “venomous spiders” (Bücherl & Buckley 1971). This group of spiders includes in particular widow spiders (Latrodectus spp.), wandering spiders (Phoneutria spp.), funnel-web-spiders (Atrax spp.), and recluse spiders (Loxosceles spp.). The very potent venoms produced by spiders from this group are potentially dangerous not only to humans, they can also be lethal to other mammals, birds, snakes, lizards, anurans, etc. (Nyffeler & Vetter 2018; Nyffeler & Altig 2020).

2.4 Definition of the terms “venomous snakes” vs. “nonvenomous snakes”.—Venomous snakes include species of snakes that use their fangs to inject toxins into their victims (Encyclopedia Britannica 2020a). In this paper, the term venomous snakes is used for species known to be potentially lethal to humans; included in this category are the families Elapidae and Viperidae. The toxicity of various snake species is compared based on LD₅₀ values of snake venom injected in mice. In contrast, nonvenomous snakes are species that do not produce a toxin that is clinically significant to humans. Nonvenomous snakes are represented by the families Colubridae, Dipsadidae, Lamprophiidae, Leptotyphlopidae, and Typhlopidae.

2.5 Definition of the term “poisonous snakes”.—By definition the term “venomous snakes” implies that these snakes inject toxins by biting their prey or enemy, while the term “poisonous snake” is applied when a snake is toxic for a natural enemy after being eaten by it (Encyclopedia Britannica 2020b). A snake species potentially poisonous to spiders is the common garter snake (Thamnophis sirtalis). This snake can become poisonous to its natural enemies after preying on toxic newts (Williams et al. 2004).

2.6 Statistical methods.—In this paper, neither spider body length data nor snake body length data are normally distributed (tested with the Shapiro-Wilk test software, online at  http://www.statskingdom.com/320ShapiroWilk.html). Likewise, LD₅₀ data for spiders and snakes are not normally distributed (Shapiro-Wilk test). Accordingly, the median must be used as a measure of central tendency. Despite the fact that data from field collections and observations are often not normally distributed, many authors still use the mean as central tendency. In order to make our data comparable to those from the literature, we calculated a mean (± SE) in addition to the median (while at the same time acknowledging that this is not the ideal measure). The two-tailed Mann-Whitney U test was used to compare mean snake lengths in USA vs. Australia, and the same test was also used to compare mean LD₅₀ values (mg/kg mouse) for venomous spiders vs. venomous snakes. These analyses were performed at  https://www.socscistatistics.com/tests/mannwhitney/default2.aspx. To test whether the percentages of rescued snakes vs. rescued birds differed statistically significantly, a chi-square calculator test (without Yates's correction) was performed; this was accomplished using MedCalc statistical software (online at  https://www.medcalc.org/calc/comparison_of_proportions.php). This same statistical test was also performed to test whether there was a significant difference between the percentages of venomous snakes killed on different continents. Based on 86 available data pairs, snake body lengths and spider body lengths were tested for a relationship using the Pearson Correlation Coefficient Calculator (online at  https://www.socscistatistics.com/tests/pearson/).

3. RESULTS

3.1 Which spider species are engaged in snake predation?—More than 30 spider species have been reported to prey on snakes under natural conditions, and 11 additional species have been documented to capture and eat snakes in captivity or in staged field experiments (Table 1 and Appendix 1). The following ten families have been documented to be engaged in snake predation under natural conditions: Agelenidae, Araneidae, Ctenidae, Idiopidae, Nephilidae, Pholcidae, Pisauridae, Sparassidae, Theraphosidae, and Theridiidae (Table 1 and Appendix 1). Additionally, a large-sized species of the wolf spider family (Lycosidae) has been witnessed attacking a snake in captivity (Anonymous 2014). Based on available data, snake-eating spiders had an average body length of 1.69 ± 0.19 cm (median = 1.00 cm, n = 86).

Table 1.

Spider families engaged in snake predation or predation attempts (based on 319 incidents reported in the scientific literature or in the social media). Evidence based on field observations except for Lycosidae.

Approximately 60% of the 319 incidents of snake predation were attributable to theridiids (≈0.6–1.1 cm body length), a spider family that uses strong tangle webs to capture prey (Table 1; Figs. 1–5). Roughly half of all incidents were attributable to widow spiders (Latrodectus spp.; Table 1). Several species of widow spiders (Latrodectus spp.) have proven to be expert snake catchers (Figs. 1A, 2B, 3–5), including four species of North American widow spiders (Latrodectus geometricus C.L. Koch, 1841, Latrodectus hesperus Chamberlin & Ivie, 1935, Latrodectus mactans (Fabricius, 1775), Latrodectus variolus Walckenaer, 1837), the Australian redback spider (Latrodectus hasselti Thorell, 1870), the African button spider (Latrodectus indistinctus O. Pickard-Cambridge, 1904), an Israeli widow spider (Latrodectus revivensis Shulov, 1948), and a white colored, not uniquely identifiable Iranian widow spider (possibly Latrodectus pallidus O. Pickard-Cambridge, 1872). The seven widow species reported in our study are of similar size (usually weighing ≈0.35–0.50 g as adult females) with maximum weights of up to ≈1 g (Canadian Geographic 2006; Nel et al. 2014; Nyffeler & Vetter 2018). Snakes are killed exclusively by adult female widow spiders (see Nyffeler & Vetter (2018) for an explanation).

Figure 1.

Blind snakes killed in webs of theridiid spiders. A. Brown widow spider, Latrodectus geometricus, feeding on a Brahminy blindsnake, Indotyphlops braminus (Typhlopidae), in a garden house in Zaachila, Oaxaca, Mexico (Photo by Matias Martinez). B. Texas Blind Snake, Rena dulcis (Leptotyphlopidae), trapped in web of the theridiid spider Steatoda triangulosa in a building near San Antonio, Texas, USA (Photo by George Popescu).

Figure 2.

Theridiid spiders capturing colubrid snakes. A. Southern Ringneck Snake, Diadophis punctatus punctatus (Colubridae), caught in a spider web in a garage in Virginia, USA (Photo by Erika Packard). The spider in question was a non-widow theridiid (most likely Parasteatoda tepidariorum (C. L. Koch, 1841)). B. Juvenile Eastern garter snake, Thamnophis sirtalis, trapped in a brown widow web (Latrodectus geometricus) observed in Douglas, Georgia, USA (Photo by Julia Safer).

Figure 3.

Black widow spiders killing colubrid snakes. A. Lined snake, Tropidoclonion lineatum (Colubridae), killed in web of black widow spider (Latrodectus sp.) in a building in Tulsa, Oklahoma, USA (Photo by Jeff Lewis). B. Juvenile scarlet snake (Cemophora coccinea, Colubridae) entrapped on web of Latrodectus geometricus, observed in a private residence in Georgia, USA (Photo by Daniel R. Crook).

Figure 4.

Black widow spiders killing snakes. A. Scarlet snake (Cemophora coccinea) entrapped and killed in a black widow web (Latrodectus sp.) in the corner of the front porch of a house in Gulf Breeze, Florida, USA (Photo by Trisha Haas). B. Adult female black widow (Latrodectus hesperus) feeding on a subadult coral snake Micruroides euryxanthus (Elapidae) near the Boyce Thompson Arboretum, Superior, Arizona, USA (Photo by Lawrence L. C. Jones).

Figure 5.

Black widow spiders killing snakes. A. Juvenile Eastern brown snake, Pseudonaja textilis (Elapidae), killed by a redback spider (Latrodectus hasselti) in Australia (Photo provided by the Australian Museum, © Kevin Moore). B. Widow spider, Latrodectus revivensis, was feeding for several days on a dead sand viper Cerastes vipera (Viperidae) in the Negev desert (Photo by Ori Segev).

The second most important snake catchers among the spiders were members of the tarantula family (Theraphosidae) which made up 10% of all instances of snake predation by spiders (Table 1; Figs. 6, 7A). These are large-sized vagrant spiders which do not spin a catching web. Some members of this family hunt prey in trees, while others catch prey on the ground. Tarantulas are among the largest and most powerful spiders, whose largest species reach a leg-span of >20 cm and a weight of up to >100 g (Nyffeler & Knörnschild 2013). Other nonweb-building spiders (e.g., Pisauridae) were less significant snake catchers (Table 1; Fig. 7B).

Figure 6.

Tarantulas killing and eating snakes. A. Fer-de-lance, Bothrops atrox, captured, killed and consumed by a Theraphosa blondi in Venezuela (staged incident photographed by Rick West). B. Theraphosa blondi capturing a juvenile false coralsnake, Oxyrhopus sp. (Colubridae), ≈30 cm in length, on forest floor at Senador José Porfírio municipality, Pará state, Brazil. Subsequently, the spider transported the snake to its burrow (Photo by Karll Cavalcante Pinto).

Figure 7.

Nonweb-building spiders killing and eating snakes. A. Adult male tarantula (Brachypelma sp.) preying on a juvenile fer-de-lance, Bothrops asper (Viperidae), in the Yaxchilan Archaeological Zone, Chiapas, Mexico (Photo by Silvano Lopez & Danniella Sherwood; © Silvano Lopez). B. Fishing spider Dolomedes tenebrosus Hentz, 1844 preying on a Middle American burrowing snake, Adelphicos quadrivirgatum (Colubridae), in the La Estanzuela Natural Park, Nuevo Leon State, Mexico (Photo by Salvador Contreras-Arquieta).

Large orb-weavers of the families Araneidae and Nephilidae reached third place in terms of importance as snake catchers (8.5% of all incidents, Table 1; Fig. 8). Especially species in the genera Argiope Audouin, 1826, Nephila, and Trichonephila have been witnessed preying on snakes that got trapped in their strong orb-webs (Wilder 1865; Gudger 1931; Pinkus 1932; Burt 1949; Zippel & Kirkland 1998; Tanaka & Mori 2000; Jantos 2012). The spiders in the genera Nephila and Trichonephila spin orb-webs with a diameter of 0.5–1.5 m (Nyffeler & Knörnschild 2013). The largest snake reported as a spider's prey, an adult green snake (Opheodrys aestivus) of 75 cm snout-vent length (equaling ≈1 m total length) had been caught in a web of Trichonephila clavipes (Linnaeus, 1767) in Florida (Zippel & Kirkland 1998).

Figure 8.

Snakes killed by large orb-weaving spiders. A. Argiope aurantia Lucas, 1833 eats red-bellied snake, Storeria occipitomaculata (Colubridae), in North Carolina (Photo by Eleanor Spicer Rice). B. Eyelash Viper Bothriechis schlegelii (Viperidae), ≈40 cm in length, has been captured in web of Trichonephila clavipes (Nephilidae) in Costa Rica rain forest (Photo by Jonathan Sequeira).

These three spider groups (Theridiidae, Theraphosidae, and Araneidae/Nephilidae) combined were responsible for roughly 80% of all reported incidents of snake predation (Table 1).

3.2 Which snake species are captured by spiders?—The snakes known as prey of spiders had an average body length of 25.9 ± 1.3 cm (median = 27 cm; range: 5.8–100 cm; n = 104). The vast majority of snake victims were either neonates or juveniles (e.g., Burt 1949; Owens 1949; Klemens 1993; Durigo 2010; Peters 2016) with a rather low body weight (see snake body weights in Table 2 for a comparison). The smallest snakes captured by spiders (i.e., juvenile typhlopids) are known to weigh only 0.2–0.7 g (Table 2). Younger snake individuals are probably more vulnerable to predators than adults due to their small size (Jorge et al. 2016). But on the other hand, in rare cases larger colubrids and vipers captured by spiders were fairly heavy, reaching an estimated body weight of up to 40–60 g (see subsection 3.6). The vast majority of snakes killed by spiders are terrestrial, with some exceptions. For example, there are several reports of juveniles of the northern water snake Nerodia sipedon having been killed by web-building spiders (Anonymous 1946; online at  https://www.youtube.com/watch?v=l8v8LPgzD_I).

Table 2.

Total lengths and body weights of selected snake species based on literature data.

The group of snakes killed and eaten by spiders in the wild encompasses 86 species, representing seven families (i.e., Colubridae, Dipsadidae, Elapidae, Lamprophiidae, Leptotyphlopidae, Typhlopidae, and Viperidae; Table 3, Appendix 2). In addition, 5 more snake species (i.e., Chironius carinatus, Drymobius margaritiferus, Imantodes cenchoa, Bothrops jararaca, and Crotalus durissus) were killed and consumed by spiders in captivity or during staged field experiments. Two families (Colubridae and Elapidae) accounted for 84% of all 284 identified snake victims, with colubrids being the most representative family (162 incidents, Table 3). The predominance of Colubridae in the total of captured snakes reflects the fact that colubrids are by far the largest and most abundant snake family on all continents except Australia (Degenhardt et al. 2005). The list of North American colubrid snakes known as prey of spiders includes the following species: black-headed and flat-headed snakes (Tantilla gracilis and Tantilla hobartsmithi), brown snakes (Storeria dekayi and Storeria occipitomaculata), crayfish snakes (Regina septemvittata), earth snakes (Virginia valeriae), garter snakes (Thamnophis cyrtopsis and Thamnophis sirtalis), glossy snakes (Arizona elegans), green snakes (Opheodrys aestivus and Opheodrys vernalis), hog-nosed snakes (Heterodon platirhinos), king snakes (Lampropeltis calligaster and Lampropeltis triangulum), lined snakes (Tropidoclonion lineatum), night snakes (Hypsiglena chlorophaea), racers (Coluber constrictor), rat snakes (Pantherophis guttatus), ringneck snakes (Diadophis punctatus), sand snakes (Sonora straminea), scarlet snakes (Cemophora coccinea), and watersnakes (Nerodia sipedon and Nerodia rhombifer) (Appendix 2). Among these North American colubrids, ringneck snakes and garter snakes have been reported to be trapped in spider webs particularly frequently (e.g., Burt 1949; Groves & Groves 1978; Klemens 1993).

Table 3.

Snake families reported to be victims of spider predation (based on 319 incidents reported in the scientific literature or social media).

That Elapidae are the snake group second most frequently killed by spiders can be explained by the numerical predominance of this snake family in Australia (Shine 1977). Brown snakes (Pseudonaja spp.) are particularly frequently killed by Australian spiders (Fig. 5A; also see De Rebeira 1981; Bush 1989; Peters 2016; Australian Museum 2020).

3.3 Global distribution of snake predation by spiders.—Predation on snakes by spiders has been witnessed so far in 25 countries (see Supplementary  Table S1 (arac-49-01-06_s01.pdf)). Predation on snakes by spiders is a global pattern occurring on all continents except Antarctica. The vast majority of reports of snake predation by spiders originates from the USA (51% of all naturally occurring incidents) and Australia (29%). In USA, snake predation by spiders is known from 29 states and all parts of the country except Alaska (Supplementary  Table S1 (arac-49-01-06_s01.pdf)).

To a lower extent, incidents of snake predation were reported from the Neotropics (8%), Asia (6%), Africa (3%), Canada (1%), and Europe (<1%). The two European reports both refer to incidents of predation upon tiny blind snakes (Indotyphlops braminus and Xerotyphlops vermicularis: Petrov & Lazarov 2000; Faraone et al. 2019), whereas the Canadian reports deal with neonate ringneck snakes (Diadophis punctatus) and presumably a juvenile wandering gartersnake (Thamnophis elegans) trapped in spider webs (The Spider Club of SA 2018; Algonquin Wildlife Research Station 2020; Malgonquin Photography 2020).

3.4 How spiders kill snakes.—Eighty-seven percent of the snakes captured by spiders got killed, 1.5% could escape on their own, and 11% were rescued by humans (see Supplementary  Table S1 (arac-49-01-06_s01.pdf)). The different groups of snake-eating spiders employ different methods to capture and kill their prey. In the following, we explain the killing behavior of the three dominant groups of ophiophagous spiders (black widows, tarantulas, and large orb-weavers).

Black widows build an irregular tangle of threads at a height of ≈10–100 cm above the substrate. From there, vertical sticky gumfooted threads extend to the substrate (Blackledge et al. 2005; Blackledge & Zevenbergen 2007). The webs are very strong and tough, enabling the spiders to capture prey many times larger and heavier than themselves (see Shao & Vollrath 1999; Blackledge et al. 2005; Swanson et al. 2006; Nyffeler & Vetter 2018). When a small snake slides into such a web, it sticks to the vertical viscid threads. The spider approaches the snake, throws sticky silk masses over it, and bites it one or more times (Vollrath 2000). The neurotoxin thereby injected is a very potent, vertebrate-specific toxin (α-latrotoxin) that has proven to be highly lethal to small vertebrates (Gendreau et al. 2017; Nyffeler & Vetter 2018). Subsequently the spider pulls its victim off the ground, raising it between 10 and 120 cm above the floor, a process which may last several hours (Figs. 1A, 2B, 3–5; Anonymous 1932a; Ervin & Carroll 2007; Jones et al. 2011; Stevenson & Crook 2018).

The time it took a theridiid to kill a snake varied greatly. In some instances, the snake was paralyzed a few minutes after the venomous bite had been administered by the spider (e.g., Krumm-Heller 1910). In other instances, it took a spider several days to kill a snake (Gudger 1925; Anonymous 1932b; Anonymous 1946). Unfortunately, in quite a number of reports of theridiid predation upon snakes, the spiders remained unidentified. Nevertheless, based on the specific behavior of lifting the snake off the ground, it can be assumed that in such cases the spiders in question must have been theridiids (see Nyffeler & Vetter 2018); but it remained open whether the predator in question was a species in the genus Latrodectus Walckenaer, 1805, Parasteatoda Archer, 1946 or Steatoda Sundevall, 1833. Reports according to which snakes were killed fairly rapidly seem to refer exclusively to widow spiders (Latrodectus spp.) (Krumm-Heller 1910; Bush 1961; Sweeney 2009; Anonymous 2016). This makes sense when taking into account that the very potent α-latrotoxin present in the Latrodectus venom is expected to reduce the time it takes to kill a snake. On the contrary, it seems likely that several instances in which it took a long time to kill a snake can be attributed to non-widow theridiids (e.g., Parasteatoda or Steatoda) which lack the potent α-latrotoxin in their venom and which produce a lower quantity of toxin due to their smaller body size (≈0.02–0.07 g) compared to the Latrodectus spp.. However, instances in which it took widow spiders several hours to kill a snake have been reported as well (e.g., De Rebeira 1981).

Tarantulas are equipped with powerful orthognath chelicerae and produce neurotoxins effectively targeting the snake nervous system (Figs. 6, 7A; Emerton 1926; Borges et al. 2016; Almeida et al. 2019). Based on laboratory feeding trials, the toxicologists Brazil & Vellard investigated the snake-catching behavior of Grammostola actaeon (Pocock, 1903), a large Brazilian tarantula (Emerton 1926). Emerton (1926) summarizes their observations as follows: “. . ..the spider tries to catch the snake by the head and will hold on in spite of all efforts of the snake to shake him off. After a minute or two the spider's poison takes effect, and the snake becomes quiet. Beginning at the head, the spider crushes the snake with its chelicerae and feeds upon its soft parts, sometimes taking 24 hours or more to suck the whole animal, leaving the remains in a shapeless mass. . ..” Theraphosa blondi (Latreille, 1804), another huge Neotropical tarantula, was seen feeding on a snake prey for 18 hours (Rick West, pers. comm.).

Tree snakes and other colubrids occasionally get trapped in the sticky webs of large araneid and nephilid orb-weaving spiders (genera Argiope, Nephila, and Trichonephila) while moving across vegetation. Once a snake is entangled in a web, the spider wraps it tightly in silk, followed by one or more venomous bites (Fig. 8A; Anonymous 1911; Pinkus 1932). After the snake's death, the spider extracts the dissolved tissue from its victim by the process of extraintestinal digestion.

3.5 Clash of the venomous: black widow spiders vs. venomous snakes.—About half of all spiders engaged in snake predation belong to spider taxa that produce neurotoxins of high toxicity. Included among these highly toxic spider species are at least seven species of widows (Latrodectus spp.; Figs. 1A, 2B, 3–5) in addition to wandering spiders (Phoneutria sp.), whereas the neurotoxins of most theraphosids are somewhat less toxic and usually harmless to humans (Foelix 2011).

By comparison, ≈30% of all snakes captured by spiders belong to species known as venomous snakes (according to the definition presented in the Methods section). Among those, a large number of species are considered highly toxic, such as New World coral snakes, Australian brown snakes (Pseudonaja spp.), rattlesnakes (Crotalus spp.), and Neotropical lanceheads (Bothrops spp.)(Figs. 4B, 5, 6A, 7A, 8B). One Australian species killed by spiders – the Eastern brown snake Pseudonaja textilis – is even considered to be one of the world's most toxic snakes to humans (Hoy 2012; Fig. 5A).

Toxicity values (expressed as LD₅₀ in mice) of venomous snakes and spiders are compared in Table 4. It follows that the LD₅₀ values of venomous spiders do not differ statistically significantly from those of venomous snakes (Mann-Whitney U test, n₁ = 13 [for spiders], n₂ = 16 [for snakes], Z = 0.3947, P > 0.05). In other words, spiders equipped with very potent venoms are at times attacking, killing, and consuming some highly lethal snakes, whereby the high toxicity of these snakes has so far only been proven in primates, some other mammals, birds, fish, lizards, and snakes. We know of no documented record in which a venomous snake captured by a spider has envenomated the spider. Thus, we don't know whether spiders might be vulnerable or immune to snake venoms. We know, however, that the venomous snakes from the families Elapidae and Viperidae (Table 4) feed more or less exclusively on vertebrate prey (e.g., Mao 1970; Klauber 1997; Goodyear & Pianka 2008). This indicates that it is less likely that spiders are vulnerable to snake venom because there was no need for this type of snake to evolve a specific toxin aimed to target the spider nervous system to facilitate the capture of spider prey. In spite of that, a snake bite could become lethal to a spider if the snake would succeed to puncture the spider's body with its fangs (Rick Vetter, pers. comm.).

Table 4.

Toxicity of spider and snake venoms injected in mice (LD₅₀ values based on literature). Route of injection: IM = intramuscular injections; IP = intraperitoneal injection; IV = intravenous injection; SI = subcutaneous injection. Low LD₅₀ values indicate high toxicity. Available online at  http://snakedatabase.org/pages/ld50.php Accessed 5 June 2020.

Apart from feeding on venomous snakes, spiders at times also feed on poisonous snakes (i.e., snakes which became poisonous after feeding on toxic prey). As pointed out in the Methods section, the common garter snake (Thamnophis sirtalis) can be poisonous to a natural enemy after being consumed by it (Williams et al. 2004). The common garter snake is among the snakes most frequently captured by spiders in USA (Fig. 2B; Supplementary  Table S1 (arac-49-01-06_s01.pdf)). But as far as known to us, spiders which consumed garter snakes survived such incidents without apparent harm.

3.6 Predator-prey size ratio.—Black widow spiders can overcome snakes 15–30 times their own size (Sutton 2016; Rocha et al. 2017; Fig. 12). For comparison, it is known that black widows can overcome mice of up to 14.4 times the spider's body mass (Nyffeler & Vetter 2018). An even more extreme example was reported in 1933 in the National Geographic Magazine (Anonymous 1933). In this particular case, a 15 cm long garter snake (weighing 8 g) was found trapped in the web of a triangulate cobweb spider, Steatoda triangulosa (Walckenaer, 1802), of only 0.0225 g body mass. Hence, this snake was 355 times heavier than the spider.

Figure 12.

Relationship between spider body length and snake body length based on 86 available data pairs. Snake body length refers to total length (snout-vent length + tail length), whereas spider body length refers to cephalothorax plus abdomen. A low positive correlation (r = 0.4136; P < 0.01) between snake body length and spider body length was found.

Based on our survey, the largest snakes caught by spiders under natural conditions were those that became entangled in webs of Nephila pilipes (Fabricius, 1793) and Trichonephila clavipes. These are spiders with a weight of ≈1–7 g that spin orb-webs with a diameter of 0.5–1.5 m (Nyffeler & Knörnschild 2013). Snakes of ≈40–100 cm total length are captured in the strong webs of Trichonephila clavipes (Zippel & Kirkland 1998; Sequeira 2011), which would correspond to an estimated maximum snake weight of ≈10–45 g (Table 2). Spiders in the genera Nephila and Trichonephila have previously been reported trapping birds and bats many times heavier than themselves in their exceedingly strong webs (Brooks 2012; Nyffeler & Knörnschild 2013). In Taiwan, a Kikuchi Habu, Trimeresurus gracilis (Viperidae), weighing an estimated ≈60 g (Table 2), became entangled in a Nephila pilipes web. After the snake was fighting for its life for many hours, the web finally broke and the snake fell to the floor in a near dead state (Shanghaiist 2014). Another interesting incident was witnessed in India. There, after a bronzeback tree snake (Dendrelaphis sp.) of almost 100 cm total length got temporarily entangled in a Nephila pilipes web, the spider intentionally broke off some of the capture threads, thereby enabling the snake to escape from the web (Ashwini Kumar Bhat, pers. comm., online at  http://sumasuta.blogspot.com/2008/09/tale-of-hanging-unfortunate.html).

Not only giant orb-weaving spiders, but also large tarantulas are capable of killing larger snakes. Tarantulas have been reported killing snakes of 25–56 cm total length (Gudger 1931; Borges et al. 2016; Pinto et al. 2017; Almeida et al. 2019; Rick West, pers. comm.). The largest tarantula species are those in the genera Grammostola Simon, 1892, Lasiodora C. L. Koch, 1850, and Theraphosa Thorell, 1870 occurring in South America. Such spiders reach body weights of ≥ 50 g and have been observed killing pit vipers with a total length of 30–53 cm, which corresponds to a prey weight of 9–40 g (Table 2).

A comparison between USA and Australia shows that the snakes captured by Australian spiders were, on average, larger (mean = 26.9 ± 0.8 cm; median = 27.0 cm; range = 14–50 cm) compared to those captured by North American spiders (mean = 21.5 ± 2.4 cm; median = 17.9 cm; range = 6.5–100 cm). The difference is statistically significant (Mann-Whitney U test, n₁ = 38, n₂ = 41, Z = 4.0622, P < 0.0001). This difference might be explained by the fact that predominantly small, lightweight colubrids such as ringneck snakes (Diadophis punctatus), garter snakes (Thamnophis spp.), Dekay's brownsnake (Storeria dekayi), green snakes (Opheodrys spp.), etc., with a neonate weight of 0.3–2 g are caught by U.S. spiders, whereas Australian spiders more often prey on heavier built elapids from the Pseudonaja-complex weighing as neonates ≈3.5–10 g (Table 2).

Based on pooled data from all over the world, snake total body length was weakly positively correlated with spider body length (r = 0.4136; P < 0.01; Fig. 12). An apparent accumulation of dots around 1.0 on the x-axis can be explained by the fact that snake-eating spiders are made up to a large extent by adult female black widows with an estimated average body length around 1 cm.

4. DISCUSSION

4.1 Are the reported incidents real predation events?—It is arguable whether all incidents reported in this paper were real predation events or whether some were just cases of scavenging. Predation requires that a prey item has been killed and eaten by the predator (Begon et al. 2005). Under natural conditions web-building spiders (i.e., theridiids, araneids, nephilids, agelenids, and pholcids) feed more or less exclusively on prey killed in their webs. Multiple YouTube videos show snakes suspended in spider webs, often fighting for extended periods of time for their life prior to their death. In addition, numerous anecdotal reports on this have been published (e.g., Jantos 2012; Anonymous 2016; Brown 2017). Most cases of this type end with the spider feeding on the dead snake.

In vagrant hunters (theraphosids, pisaurids, and ctenids) the situation is somewhat different. These are voracious, highly opportunistic feeders which feed not only on freshly killed prey, but occasionally also on carcasses of animals that had not been killed by themselves (“scavenging”: Schmidt 1957; Nyffeler & Altig 2020). Notwithstanding this, many well documented cases of predation on snakes by vagrant hunters observed in the field and in captivity have been reported in the literature (e.g., Rick West, pers. comm.; Emerton 1926; Pinto et al. 2017).

A very special case of scavenging has been observed in Singapore. There, a tiny, red colored theridiid kleptoparasite (possibly genus Argyrodes) was observed feeding on tissue of a painted bronzeback snake carcass which had apparently been killed in a Nephila pilipes web (Lim 2009).

4.2 What percentage of captured snakes was rescued by humans?—Of the captured snakes, only 11% were rescued by humans and 1.5% could escape on their own, whereas 87% were killed by spiders (see Results section). By comparison, Brooks (2012) analyzing 69 incidents of bird captures by web-building spiders found that >50% were rescued by humans, a little over 10% could escape on their own, and a little over 30% were killed by spiders. The percentages of rescued snakes vs. rescued birds differed statistically significantly (Chi-square test, χ² = 50.03, df = 1, P < 0.001). In conclusion, the willingness to rescue a snake is lower than the willingness to free a bird. Considering that some of the snakes caught in spider webs are highly toxic to humans and that in addition to this a high percentage of the snake-catching spiders are highly toxic, it is understandable that only a small percentage of observers is willing to get involved in a snake rescue operation. This might be especially true in countries like Australia dominated by venomous elapids (e.g., brown snakes) where even a juvenile snake can administer a deadly bite to a human (Sutherland & Tibballs 2001).

4.3 Nonpredation causes by which snakes are seriously harmed by spiders.—Predation is not the only way snakes can be harmed by spiders, as the following examples demonstrate. Murphy (2014) reports tail amputation in a reticulated python (Malayopython reticulatus) at the Dallas Zoo due to necrosis from a brown recluse spider bite (Loxosceles reclusa Gertsch & Mulaik, 1940) (Sicariidae). Although no evidence of spider involvement is provided in the aforementioned case, it cannot be ruled out that the necrosis in the python was indeed caused by a recluse spider bite, since sicariid venom has been reported to produce dermonecrotic effects on reptile tissue and reptile fatality (Ramires & Fraguas 2004; Taucare-Rios & Piel 2020).

A second case where a snake was put in danger by a nonpredation cause was witnessed in a residential area in Oregon. In that case, a juvenile Sonoran gopher snake (Pituophis catenifer) about 20–25 cm in length, whose head and much of its body was wrapped up in spider webbing with other detritus (i.e., gum, small bits of paper, and pebbles) attached to it, was found in a building (Tim Akimoff, pers. comm.; Fig. 9). This snake's view was strongly obstructed by the amount of webbing engulfing its head. The person who reported this incident does not believe the snake would have been able to free itself from the web. He states with regard to the webbing “. . ..It was extremely sticky when I removed it, and I had to hold the snake's head firmly in order to get all of the webbing pulled off of it” (Tim Akimoff, pers. comm.). There are two possibilities how this snake's head got engulfed by the webbing: 1. The snake broke loose from a spider web after having gotten temporarily entangled in it; 2. The snake picked up the pocket of webbing along with the detritus while passing under a door from the building to inside (Tim Akimoff, pers. comm.). Be that as it may, due to the impaired vision this snake's chance of survival would have been severely diminished if the bag of webbing on its head had not been removed by the human who found the snake.

Figure 9.

Juvenile Sonoran gopher snake (Pituophis catenifer) found in a building in Salem, Oregon. Its head and much of its body was wrapped up in spider webbing with other detritus (i.e., gum, small bits of paper, and pebbles) attached to it, and it is assumed that the snake either was wrapped up in spider webbing after having escaped from a spider web, or that it picked up the pocket of webbing along with the detritus while passing under a door from the building to inside. The snake's view was strongly obstructed by the amount of webbing engulfing its head indicating that its chance of survival would have been severely diminished if the bag of webbing on its head had not been removed by the human who found it in the building (Photo: Tim Akimoff).

4.4 How important are reptiles as spider diet?—Most spiders reported in this review (i.e., Agelenidae, Araneidae, Ctenidae, Idiopidae, Lycosidae, Nephilidae, Pholcidae, Pisauridae, Sparassidae, Theraphosidae, and Theridiidae) feed to a large extent on arthropod prey, and under most circumstances snakes are probably only marginal food for them. However, there are a few exceptions to this.

In laboratory feeding experiments conducted in Brazil, the large tarantula Grammostola actaeon refused to eat insects offered to it (Emerton 1926). Thereafter, small snakes and frogs were offered to the spiders, and those were readily consumed (Emerton 1926). As a result of Emerton's report, this tarantula species has been implied in several books to be a specialized predator of snakes and frogs (Berland 1932; Millot 1949; Gertsch 1979; Hillyard 1994). In light of newer information, the characterization of this tarantula species as a specialized snake- and frog-eater must be somewhat relativized. Contrary to Emerton's experiences, Ibler et al. (2013) were able to rear Grammostola actaeon on a diet of crickets, which shows that this species apparently has a broader diet than previously assumed. Apart from Grammostola actaeon, still other large tarantula species (e.g., Theraphosa blondi) kill snakes (Punzo & Henderson 1999; Aguilar-Lopez et al. 2014; Jorge et al. 2016; Pinto et al. 2017). It took large tarantulas 18–24 hours to devour snakes of 30–55 cm total length (Emerton 1926; Rick West, pers. comm.). The long feeding times reported in these cases indicate that a snake of that size is a “big, profitable meal” for a large tarantula. Tarantulas are capable of ingesting large amounts of food (thereby storing surplus energy in their body's interstitial tissue as lipid or glycogen) and in time periods of starvation the spiders can access these stored energy reserves (Foelix 2011). Viewed this way, feeding on snakes might have long-term nutritional value, – especially in environments with irregular food availability where food can become scarce at times. Most reports of large tarantulas feeding on snakes originate from the Neotropics (Emerton 1926; Nunes et al. 2010; Aguilar-Lopez et al. 2014; Jorge et al. 2016; Pinto et al. 2017).

Not only for large tarantulas, but also for smaller sized web-building spiders, a snake prey might be a big catch. After a neonate red-bellied snake (Storeria occipitomaculata) fell prey to an unspecified, small web-building spider in a building in Pennsylvania, USA, the spider ate at the snake carcass for several days (Swanson 1952). A similar account was reported by Yael Lubin from the Negev desert, where a widow spider, Latrodectus revivensis, was seen feeding on a dead sand viper (Cerastes vipera) for several days (Yael Lubin & Ori Segev, pers. comm.).

The most compelling example of spiders feasting on snake prey has been reported from Western Australia. In a shed in Dongara, a scene was filmed where dozens of neonate brown snakes (Elapidae) were trapped and killed in the webs of a colony of redback spiders, Latrodectus hasselti (Peters 2016). This episode of redback spiders feeding heavily on reptiles does not seem to be so unusual. For instance, in a building in South Australia a redback spider had been observed catching eight lizards in succession (Poposki 2019). Altogether, a total of 27 spider families have been reported consuming reptile prey (Table 5).

Table 5.

Spider families engaged in predation on reptiles based on literature data.

4.5 Do spiders capture predominantly venomous or nonvenomous snakes?—In USA, 6 of 161 snake specimens (3.7 %) captured by spiders belong to the venomous snakes (i.e., Elapidae and Viperidae). By contrast, in Australia, 77 of 87 snakes (88.5 %) captured by spiders were venomous snakes (i.e., Elapidae, but also including the colubrid Boiga irregularis). This difference between the percentages is statistically highly significant (Chi-square test, χ² = 181.72, df = 1; P < 0.0001) and can be explained by the fact that the vast majority of the snakes occurring in USA are nonvenomous colubrids (Nellis 1997), whereas the Australian snake fauna is dominated by venomous elapids (Fig. 5A; Mirtschin et al. 2002). The few North American venomous snake taxa preyed upon by spiders include coral snakes (Micrurus, Micruroides), pygmy rattlesnakes (Sistrurus), and rattlesnakes (Crotalus) (Fig. 4B; Krumm-Heller 1910; Punzo & Henderson 1999; Jones et al. 2011). The situation in the Neotropics may be somewhere in between; here 13 of 33 snake victims of spiders (39.4 %) were venomous (i.e., Elapidae and Viperidae; Figs. 6A, 7A, 8B). The percentage of venomous snakes is significantly higher in the Neotropics than in USA (Chi-square test, χ² = 39.38, df = 1, P < 0.0001) and significantly lower than in Australia (Chi-square test, χ² = 30.50, df = 1, P < 0.0001). The Neotropical venomous snake taxa preyed upon by spiders include coral snakes, rattlesnakes, palm-pitvipers (Bothriechis), and lanceheads (Bothrops; Emerton 1926; Avila & Porfirio 2008; Sasa et al. 2009; Nunes et al. 2010; Sequeira 2011).

4.6 How important are spiders as mortality agents of snakes?—Birds of prey, some carnivorous mammals (e.g., mongooses, opossums, and raccoons) as well as other snakes are generally considered to be the major predators of snakes (Fitch 1949; Greene 1997; Linzey & Clifford 2002; Gibbons 2017). Apart from carnivorous vertebrates, several groups of large invertebrate predators such as spiders, scorpions, praying mantises, giant water bugs, centipedes, and driver ants have been reported killing small snakes (McCormick & Polis 1982; Greene 1997; Nyffeler et al. 2017c). In their book “Snakes in question: The Smithsonian answer book” the herpetologists Zug & Ernst (2015) hypothesized that “. . ..spiders are probably the main invertebrate predators of snakes”. Similarly, Greene (1997) states in his comprehensive book on snakes “. . ..small, tropical forest snakes may be especially vulnerable to the centipedes, spiders, and other large invertebrates common in such habitats. . ..” Based on our review, it can be assumed that snake predation by certain spiders does indeed occur quite frequently. However, due to the fact that those spider species which are capable of killing snakes rely to a large extent on arthropods as a primary food source (Thierry Gasnier, pers. comm.; Lapinski & Tschapka 2013), spiders should be expected to play a subordinate role as mortality agents of snakes compared to the carnivorous vertebrates.

4.7 Spiders and snakes as intraguild predators.—Not only are spiders sometimes capturing and killing snakes, quite often the tables are turned – that is, a larger number of arthropod-eating snake species include spiders in their diets (Table 6; Figs. 10, 11). With few exceptions, spider-eating snakes belong to nonvenomous species in the family Colubridae (especially subfamily Colubrinae) and these snakes usually range between approx. 25 and 30 cm total length in most cases. Although spider-eating snakes feed for the most part on small, defenseless spiders (Plummer 1981, 1991; Degenhardt et al. 2005; Marques et al. 2006), there are exceptional cases in which large and/or highly toxic spiders considered dangerous prey are swallowed by snakes (Campbell 1999; Branch 2016; Holly Chapman, pers. comm.; Max Roberts, pers. comm.). In such instances, a captured spider may try to bite its captor in the mouth which might be a risky venture for the snake (Marques et al. 2006). In the following, we will present a few examples of risky snake/spider encounters.

Table 6.

Examples of colubrid snake species reported to feed on spiders, with percent spiders in the snakes' total diet (based on literature data).

Figure 10.

A., B. Orange-bellied ringneck snake (Gongylosoma baliodeirum) eating a wandering spider, Ctenus sp., observed in the Central Catchment Nature Reserve, Singapore (Photos by Holly C. Chapman).

Figure 11.

Snake attack on Trichonephila clavipes spider photographed in West Ashley, South Carolina, USA (Photos by Max Roberts/iNaturalist CC BY-NC). A. Rough green snake, Opheodrys aestivus (Colubridae), approaches webs of Trichonephila clavipes with the attempt to catch a spider. B. Rough green snake, Opheodrys aestivus, has captured a Trichonephila clavipes spider and is swallowing it.

Two snakes endemic to Mexico and Central America – the Degenhardt's Scorpion-eating Snake (Stenorrhina degenhardtii) and the Blood Snake (Stenorrhina freminvillei) are expert arachnid hunters that feed more or less exclusively on theraphosids and scorpions (Censky & McCoy 1988; Campbell 1999; Holm 2008; Köhler et al. 2017). These snakes appear to be immune to the stings of scorpions (Alvarez del Toro 1972).

In the Sonoran Desert (North America), numerous species of small, fossorial snakes occur that feed heavily on small spiders (Babb et al. 2005; Holm 2008; Smith et al. 2008). In stomach and fecal samples of the ground snake (Sonora semiannulata), the remains of black widow spiders and small scorpions are commonly detected, implying that these types of venomous arachnids are habitually consumed by these snakes, apparently without harming them (Funk 1967; Degenhardt et al. 2005). Black widow spiders were fed to the western shovel nose snake (Sonora occipitalis) in captivity – without any negative implications for the snakes (Funk 1967).

Last but not least, we wish to mention the example of the North American green snakes, the rough green snake (Opheodrys aestivus), which is frequently arboreal and commonly reaches a length of 80 cm, and the smooth green snake (Opheodrys vernalis), a ground dwelling species, which is smaller with an adult length of ≈50 cm. These snakes are known to feed heavily on spiders (Plummer 1981, 1991; Redder et al. 2006; Baldwin 2007). The vast majority of spiders captured by the green snakes are small and defenseless; however, occasionally rough green snakes snatch large orb-weaving spiders from large orb-webs (genera Argiope and Trichonephila) (Fig. 11; Anonymous 2020; Max Roberts, pers. comm.). To approach large orb-webs with the purpose of catching their inhabitants (Fig. 11A) can be a risky endeavor for these snakes which can get entangled irreversibly in the spiders‘strong sticky webs (Gudger 1931; Zippel & Kirkland 1998).

5. CONCLUDING REMARKS

The global spider community, that weighs an estimated 25 million tons, is assumed to consume about 400–800 million tons of prey per year (Nyffeler & Birkhofer 2017). To satisfy this enormous appetite, the spiders must acquire enough food from a broad variety of food sources (see Nyffeler & Symondson 2001; Nyffeler & Knörnschild 2013; Nyffeler & Pusey 2014; Nyffeler et al. 2016, 2017a,b; Nyffeler & Vetter 2018). To this effect, the use of vertebrates as a supplementary food source by spiders represents an opportunity to enlarge their food base, resulting in enhanced chance of survival. Interestingly the snakes captured by spiders also encompasses some species from the families Elapidae and Viperidae known to be highly toxic to humans and other vertebrates, however, there is so far no evidence in the literature that a venomous snake captured by a spider has envenomated the spider.