Comparative analysis of passive defences in spiders (Araneae)

Authors


Summary

  1. Being frequent prey of many predators, including especially wasps and birds, spiders have evolved a variety of defence mechanisms. Here I studied patterns of passive defences, namely anachoresis, crypsis, masquerade, aposematism and Batesian mimicry, in spiders.
  2. Using published information pertaining more than 1000 spider species, the phylogenetic pattern of different passive defences (i.e. defences that decrease the risk of an encounter with the predator) was investigated. Furthermore, I studied the effect of foraging guild, geographical distribution and diel activity on the frequency of defences as these determine the predators diversity, presence and perception.
  3. I found that crypsis (background matching) combined with anachoresis (hiding) was the most frequent defence confined mainly to families/genera at the base of the tree. Aposematism (warning coloration) and Batesian mimicry (imitation of noxious/dangerous model) were found in taxa that branched later in the tree, and masquerade (imitation of inedible objects) was confined to families at intermediate positions of the tree. Aposematism and Batesian mimicry were restricted to a few lineages.
  4. Masquerade was used particularly by web-building species with nocturnal activity. Aposematism was rare but mainly used by web-building diurnal species. Batesian mimicry was frequently observed in cursorial species with diurnal activity. Cryptic species were more common in temperate zones, whereas aposematic and mimetic species were more common in the tropics.
  5. Here I show that the evolution of passive defences in spiders was influenced by the ecology of species. Then, I discuss the evolutionary significance of the particularly defences.

Introduction

Animals have evolved a great variety of defences against predators. The defences are classified into two major types (Ruxton, Sherratt & Speed 2004). Primary or passive defences, such as crypsis and mimicry (Batesian or Müllerian), decrease the risk of an encounter with the predator, while secondary or active defences, such as fleeing, decrease the probability of capture once the prey is discovered. Both types of defences can be employed by prey species, and there might be a trade-off in investment into one or the other depending on the strength of selection pressures imposed, for example, by different predatory guilds (Pekár et al. 2011).

If considering only visually oriented predators, the passive defences of prey are based upon morphological adaptations, namely body coloration and/or shape, and behavioural adaptations, such as choosing a background to avoid being detected by predators or exposing themselves to predators and advertising defences (Endler 1984). A variety of passive defences have been described. For example, crypsis is a resemblance of background perceived by a predator in the microhabitat where the prey is most often sought (Endler 1981). Similarly, masquerade is the imitation of an inedible, generally inanimate but abundant object, such as a plant part. Aposematic (and Müllerian mimics) produce honest signalling to the predator by means of a contrast in coloration (i.e. warning coloration), Batesian mimics imitate unpalatable prey though they themselves are palatable (e.g. Edmunds 1974).

All defences are associated with certain costs, and none is universally the best one, that is, none protects prey from all enemies. For example, crypsis is most efficient when the animal is not moving on the particular background, which may constrain its search for food or mates as well as limit the ability to thermoregulate. It is considered as a primitive category of defence (Ruxton, Sherratt & Speed 2004). It has been proposed that mimicry is costly due to the production of the coloration, shape and behavioural resemblance. The evolution of a certain defence category depends on the balance between antagonistic selection pressures – natural history activities (such as intraspecific communication) and predator avoidance (Gomez & Théry 2007). These are dependent on the environment and the characteristics of the predatory guilds.

Spiders are abundant arthropods present in all terrestrial habitats. The majority of spider species are tiny, palatable and therefore they frequently prey a number of predators. Predators of spiders are found among both invertebrates and vertebrates, such as arachnids (Jackson 1992), dragonflies (Young & Lockely 1988), mantids (Brushwein, Hoffman & Culin 1992), dipterans (Bristowe 1945), reptiles (Schoener & Spiller 1996) and mammals (Edmunds & Edmunds 1983). The two major groups of spider predators, however, are wasps and birds, which has been documented by abundant evidence (e.g. Fincke, Higgins & Rojas 1990; Vander Haegen 1990; Edmunds 1993; Gajdoš & Krištín 1997).

How do spiders avoid predation by wasps and birds? As the wasps and birds are diurnal and using vision to detect prey, spiders have evolved defences that are efficient for visual modality. Spiders have achieved an enormous diversity of phenotypes with more than 40 000 species described so far (Platnick 2011). They use almost all categories of defences known from animals. Spiders are thus an ideal group for the analysis of the evolution of defences. Such an analysis is now feasible given the many natural history reports on the defences of a wide range of spider species and the considerable progress made in the phylogeny of spiders (Agnarsson, Coddington & Kuntner 2013).

My aims in this paper were to summarize the use of proposed passive defences in spiders, not strong evidence-based conclusions about defences, and to determine which categories predominate by employing information from published sources. I focused on defences that protect spiders mainly from diurnal visually oriented predators. Using this information, I investigated the phylogenetic pattern of different categories of defences. I predicted that some defences, aposematism and Batesian mimicry, are derived categories and restricted to relatively few lineages because unlike crypsis, these are genetically complex adaptations (Kunte 2009). Then, I investigated the effect of selected ecological variables. Specifically, I studied whether the category of defence is influenced by geographical distribution, spider foraging guild and diel activity. I expected (i) the geographical distribution to affect the frequency of different defences because the selection pressure exerted by variety of predatory guilds differs along latitude (Schemske et al. 2009); (ii) the foraging guilds to affect the frequency of different defences as the guilds differ in overall movement activity and affinity to microhabitats, which both determine the exposure to predators (Gomez & Théry 2007); and (iii) the diel activity to influence the frequency of defence categories because the light constrain predator visual perception (Merilaita & Tullberg 2005).

Materials and methods

Data collection

Data on the passive defences of spiders were taken from more than 200 papers published between 1870 and 2012 (Appendix S1). Although majority of taxonomic papers include information on body colour, the category of defence was reported only in some papers because it requires also information on the life history, background, common objects, etc., to designate a defence. It must be stressed that the category of defence attributed to each species in the literature in great majority of species is only a reasonable hypothesis, not evidence-based information. In fact, these hypotheses await to be confirmed experimentally, except for very few cases where evidence is already available (Palmgren, Ahlqvist & Langenskiöld 1937; Théry & Casas 2002; Heiling, Herberstein & Chittka 2003; Nelson & Jackson 2009; Durkee, Weiss & Uma 2011). Thus, in the great majority of cases, I am referring to proposals about the passive defences of different spider species. For example, when I simply say that a species is aposematic, then it means that it has been proposed that it is aposematic and not that it is.

Altogether, I found such hypotheses for 1017 species of spiders belonging to 546 genera and 83 families. I distinguished five categories of passive defence: (i) anachoresis – is defined as a species – that is completely hidden in a retreat, burrow, cave, etc. during the day and comes out only at night; (ii) crypsis – if a species possesses coloration matching the background (including homochromy, disruptive coloration and countershading); (iii) masquerade – if a species imitates any inedible object by positioning its body parts; this also includes active camouflage, when the spider makes a an effort to imitate an inedible model (e.g. by stretching along stalk); (iv) aposematism – if a spider has a contrasting colour pattern (black/white, black/red, black/yellow) and obvious defences; (v) Batesian mimicry – if a spider bears a marked resemblance to another animal (see e.g. Ruxton, Sherratt & Speed (2004) for a complete definition of the defence categories). Some defensive traits, such as crypsis, are known to have also offensive function (i.e. aggressive mimicry), though evidence is still very rare (e.g. Jackson & Blest 1982; Théry 2007). From such limited information, it is not possible to estimate which of the selective forces have been driving the evolution of crypsis, and therefore, offensive function is not treated here.

The geographical areas of distribution for each species, classified as tropical (latitudes between 0 and 25°), subtropical (latitudes 25° and 40°) and temperate (latitudes above 40°), were taken from Platnick (2011). Of all data, 24·5% (n = 1017) were for temperate spider species, 25·3% were for subtropical and 50·2% were for tropical. The classification of foraging strategies (i.e. guilds) was species specific and was taken from the various literature sources (Appendix S1). I distinguished only three guilds (cursorial, burrowing and web-building) because this was the basic division when several ecological characteristics were used to distinguish guilds and functional groups (see Cardoso et al. 2011). The burrowing guild (including webs on the ground) (Cardoso et al. 2011), that is, species foraging like cursorial species but from burrows, contained 6·8% (n = 1017) of the species. The web-building guild, that is, species building webs off the ground for prey capture, contained 35·2% of the species. And the cursorial guild, that is, species that do not build capture webs and most often capture prey without the use of a web (or a burrow), contained 58% of the species. The diel activity pattern (diurnal or nocturnal) was also species specific and was taken from the various literature sources (Appendix S1). Diurnal is used for species that is visible to potential predators using vision to detect prey disregarding whether the spider moves or not. Nocturnal then is used for species that hide during the day and is thus not visible to predators. Altogether I found data for 417 species, of which 63·3% were diurnal and 36·7% were nocturnal.

Phylogenies

Phylogenetic trees used to map defences were constructed from available information from published data. At the family level, I used the most recent and the most complete phylogeny for Araneae proposed by Coddington (2005) because more recent ones (e.g. Dimitrov et al. 2012) produced phylogeny conflicting with the consensus view (Agnarsson, Coddington & Kuntner 2013). This phylogeny lacked a few newly designated families, namely Cybaeidae, Hahniidae, Homalonychidae and Nephilidae. Their positions were resolved using additional sources (Jocqué & Dippenaar-Schoeman 2006; TOL 2011). Generic-level phylogenies were constructed for each family (i) that included at least two categories of passive defence, (ii) that included more than 10 species with known defences and (iii) for which a phylogenetic analysis was available. These conditions were met for nine families. The genus-level phylogenies were constructed using data from the following sources: Araneidae (Scharff & Coddington 1997; Agnarsson & Blackledge 2009; T. Blackledge, pers. comm.), Corinnidae (Bosselaers & Jocqué 2002; J. Bosselaers, pers. com.), Eresidae (Miller et al. 2010), Linyphiidae (Miller & Hormiga 2004; Arnedo, Hormiga & Scharff 2009), Salticidae (Maddison & Hedin 2003; Maddison, Bodner & Needham 2008), Theridiidae (Arnedo et al. 2004; Arnedo, Agnarsson & Gillespie 2007), Thomisidae (Benjamin et al. 2008), Tetragnathidae (Álvarez-Padilla et al. 2009) and Zodariidae (Jocqué 1991).

Statistical analyses

Analyses of comparative data were performed within R (R Development Core Team 2010) using methods that can handle phylogenetic information. I adapted the linear model concept to test for the directionality of defence evolution at the family level by studying the relationship between the category of defence and the tip distance (i.e. number of nodes from the root to the tip (Stireman (2005)). As the response variable was binary, generalized estimating equations with the binomial error structure (GEE-b) from the ape package (Paradis 2006) were used. The correlation structure among observations was created from the trees reported above assuming a Brownian motion model of character evolution (Hansen & Martins 1996). For generic-level phylogenies, branch lengths were taken from published phylogenies. For family-level phylogeny, the branch lengths were not reported. So these were estimated using the Grafen's (1989) method which has been shown to provide good estimates of evolutionary relationships (Martins & Hansen 1996). I modelled the relationship between the category of defence and tip distance by means of a logistic model. The relationship between the category of defence and tip distance would decrease linearly on a logit scale if the defence was a basal condition, increase linearly if the defence was a derived condition or be nonlinear (hump-shaped) if defence was at an intermediate position within a tree. Thus, I first fitted a quadratic logit model to the data, and if the quadratic coefficient was not significantly different from zero, it was removed from the model. A corrected number of degrees of freedom for the Wald test of parameters were used. Bonferroni correction was applied to the significance level for the multiple tests within each family.

Generalized linear models with the quasi-binomial setting (GLM-qb) were used to test the effect of distribution range, guild type and diel activity on the frequency of defence category because the GEE did not converge and because Moran's I autocorrelation (Gittleman & Kot 1990), ran on the proportion of all categories of defences at family-level phylogenies, revealed independence. To correct for non-independency and soft polytomy (sensu Purvis & Garland 1993) at the genus-level (similar trait values for all species within a genus), the number of degrees of freedom was lowered to the number of genera used in the analysis.

The ancestral states for each node were estimated within the nine families using a maximum likelihood method assuming Brownian motion model using ace function (Schluter et al. 1997). The model used was continuous instead of discrete because the tip values for each genus were not discrete categories but relative frequency of species using a certain defence category. Genera whose position in the genus-level phylogenies was unclear were excluded from these analyses.

Results

Anachoresis and crypsis

Anachoresis and crypsis were proposed for 51·1% of species (n = 1017). Anachoretic species can hide themselves in burrows in the ground (e.g. some representatives of Lycosidae, Migidae, Theraphosidae), under bark (e.g. some representatives of Clubionidae, Sparassidae, Trochanteridae), in a cave (e.g. some representatives of Desidae, Scytodidae, Telemidae), in a crevice (e.g. some representatives of Agelenidae, Desidae, Oonopidae), in a snail shell (e.g. some representatives of Anyphaenidae), in a rolled leaf (e.g. some representatives of Araneidae, Clubionidae, Miturgidae), under a rock (e.g. some representatives of Caponiidae, Cithaeronidae, Gnaphosidae), or are buried in sand (e.g. some representatives of Microstigmatidae, Pisauridae, Zodariidae) and thus become unavailable to diurnal visually oriented predators.

Cryptic species blend with the appearance of bark (e.g. some representatives of Hersiliidae, Tetragnathidae, Thomisidae), flowers (e.g. some representatives of Thomisidae), grass (e.g. some representatives of Dictynidae, Lycosidae, Pisauridae), soil (e.g. some representatives of Agelenidae, Phyxelididae), leaves (e.g. some representatives of Araneidae, Oxyopidae, Sparassidae), lichens (e.g. some representatives of Araneidae, Philodromidae), sand (e.g. some representatives of Lycosidae, Salticidae) or stone/walls (e.g. some representatives of Nesticidae, Scytodidae). Anachoresis and crypsis, however, often overlap. For example, some lycosids hide in a burrow during the day, yet possess coloration that is cryptic.

Family-level analyses

Anachoresis and crypsis are the only defences employed by Liphistiidae, Mygalomorpha (Actinopodidae, Antrodiaetidae, Barychelidae, Ctenizidae, Cyrtaucheniidae, Dipluridae, Hexathelidae, Idiopidae, Microstigmatidae, Migidae, Nemesiidae, Paratropidae, Theraphosidae), Lycosoidea (Ctenidae, Lycosidae, Oxyopidae, Pisauridae, Psechridae, Senoculidae, Trechaleidae, Zoropsidae), most Haplogyne (Caponiidae, Diguetidae, Drymusidae, Filistatidae, Leptonetidae, Orychoceratidae, Plectreuridae, Scytodidae, Segestriidae, Sicariidae, Telemidae, Terablemmidae), Orbiculariae (Anapidae, Cyatholipidae, Mysmenidae, Nesticidae, Synotaxidae), Dionycha (Anyphaenidae, Clubionidae, Cybaeidae, Dictynidae, Nicodamidae, Philodromidae, Phyxelididae, Selenopidae, Titanoecidae, Zoridae), Gnaphosoidae (Ammoxenidae, Cithaeronidae, Lamponidae, Prodidomidae), Agelenidae, Amaurobiidae, Desiidae, Hersiliidae, Oecobiidae and Palpimanidae (Fig. 1).

Figure 1.

Topology of families within Araneae with mapping of their passive defence strategies. The coloured pies at the terminals indicate the proportion of the number of species using particular defence category within each family. Only families for which hypotheses on defences were available are shown.

At the family level, the relative frequency of anachoresis and crypsis was significantly related to the tip distance in a U-shaped (quadratic) shape (GEE-b, F2,17 = 6·9, P = 0·0017, Fig. 2a), suggesting that these defences are typical for basal and very derived taxa.

Figure 2.

Relationship between the relative frequency (proportion of number of species) of anachoresis and crypsis (a), masquerade (b), aposematism (c) and Batesian mimicry (d), and the tip distance. The size of points is scaled to the number of species used in the analysis. The size of points shows their influence on the estimation of the model. Estimated logit models are shown only for significant relationships.

There was a significant difference in the number of species using anachoresis or crypsis across geographical zones (GLM-qb, F3,375 = 22·4, P < 0·0001): anachoresis and crypsis were more frequent in temperate and subtropical species than in tropical species (Fig. 3a). There was also a significant difference among guilds (GLM-qb, F2,375 = 82, P < 0·0001): anachoresis and crypsis were the most frequent in burrowing species (which by definition hide in burrows) (Fig. 3b). In addition, there was a significant effect on diel activity (GLM-qb, F1,142 = 142·6, P < 0·0001): anachoretic and cryptic species were particularly nocturnal (Fig. 3c).

Figure 3.

Comparison of the proportion of the number of species using crypsis (+anachoresis), masquerade, aposematism and Batesian mimicry across three geographical zones (a), three guilds (b), and between two diel activity patterns.

Generic-level analyses

Within families, anachoresis and crypsis have evolved in few lineages within Eresidae, Linyphiidae, Tetragnathidae, Theridiidae, Zodariidae, Thomisidae, Araneidae and Salticidae (Figs 4 and 5). The estimation of ancestral states within families revealed that anachoresis and crypsis were global root condition in Linyphiidae, Theridiidae, Zodariidae, Thomisidae, Salticidae and Araneidae (Figs 4 and 5).

Figure 4.

Topology of eresid (a), linyphiid (b), corinnid (c), tetragnathid (d), theridiid (e) and zodariid (f) genera with the number of species used in the analysis in round brackets. The pie charts at the terminals show the proportion of the number of species using certain defence category by each genus. Ancestral conditions displayed by pies at nodes show the estimated proportion of anachoresis + crypsis (white) and masquerade (Tetragnathidae, Theridiidae), aposematism (Theridiidae) and Batesian mimicry (Eresidae, Linyphiidae, Corinnidae, Zodariidae) (all grey). Ancestral states were estimated using maximum likelihood.

Figure 5.

Topology of araneid (a), thomisid (b) and salticid (c) genera with the number of species used in the analysis in round brackets. The pie charts at the terminals show the proportion of the number of species using certain defence category by each genus. Ancestral conditions displayed by pies at nodes show the estimated proportion of anachoresis + crypsis (white) and masquerade (Araneidae, Thomisidae, Salticidae), aposematism (Araneidae) and Batesian mimicry (Araneidae, Thomisidae, Salticidae) (all grey). Ancestral states were estimated using maximum likelihood.

Masquerade

Masquerade was proposed for 10·6% of species (n = 1017). Most species resemble twigs (e.g. some representatives of Araneidae, Deinopidae, Tetragnathidae), debris in a web or on the ground (e.g. some representatives of Araneidae, Salticidae), bird droppings (e.g. some representatives of Araneidae, Thomisidae) and fruits (e.g. some representatives of Araneidae) (Fig. 6a).

Figure 6.

Comparison of the percentage of models of masquerade (a) and Batesian mimicry (b) in spiders. Hymenoptera stands for ants in majority of cases (93%).

Family-level analyses

Masquerade was found only in Orbiculariae and a few families belonging to Dionycha. At the family level, the relative frequency of masquerade was significantly related to the tip distance in a humped quadratic shape (GEE-b, F1,17 = 25·6, P = 0·0002, Fig. 2b), showing that masquerade had evolved in species at intermediate positions of the family tree.

There was no significant difference in the number of species using masquerade across geographical zones (GLM-qb, F3,52 = 2·3, P = 0·09, Fig. 3a). There was a significant difference among guilds (GLM-qb, F2,52 = 60·6, P < 0·0001): masquerade was the most frequent defence in web-building species (Fig. 3b). In addition, there was significant effect of diel activity (GLM-qb, F1,14 = 7·5, P = 0·016): nocturnal species used masquerade during the daytime when they are quiescent more frequently than diurnal species (Fig. 3c).

Generic-level analyses

Masquerade had evolved in one lineage in Tetragnathidae and Salticidae, in two lineages in Theridiidae and in a number of lineages in Araneidae (Figs 4 and 5). The estimation of ancestral states in the nine families revealed that masquerade was the global root condition in Tetragnathidae (Fig. 4).

Aposematism

Aposematism was proposed for 2·7% of species (n = 1017). Conspicuous colour patterns of aposematic species included combinations of black and red, black and yellow, white and red, and red and yellow (some representatives of Araneidae, Ctenidae, Nephilidae, Theridiidae).

Family-level analyses

Aposematism was only rarely found in Mygalomorphae and Orbiculariae. There was no significant difference in the number of aposematic species across geographical zones (GLM-qb, F3,14 = 2·0, P = 0·16, Fig. 3a). However, there was a significant difference in the number of aposematic species among guilds (GLM-qb, F2,14 = 17·9, P = 0·0001): aposematism was most frequent in web-building species (Fig. 3b). In addition, there was a significant effect of diel activity (GLM-qb, F1,6 = 4·9, P = 0·04): aposematic species were particularly diurnal (Fig. 3c).

Generic-level analyses

Aposematism had evolved in one lineage in Theridiidae and in two lineages within Araneidae (Figs 4 and 5) particularly in derived genera. The estimation of ancestral states in the nine families revealed that aposematism was not the global root condition in any family.

Batesian mimicry

Batesian mimicry was proposed for 35·6% of species (n = 1017). Mimetic species imitate 11 animal groups, with the highest proportion for Hymenoptera (89%, n = 362) (Fig. 6b). Within Hymenoptera, 93% (n = 322) of species imitate ants; the rest imitate Mutillidae (6%) and Vespidae (1%). The highest number of mimics was found in Salticidae, followed by Corinnidae, Araneidae and Theridiidae.

Family-level analyses

Batesian mimicry was found only rarely in Mygalomorphae and Haplogyne, and mainly in Orbiculariae and Dionycha. At the family level, the relative frequency of Batesian mimicry increased significantly with the tip distance (GEE-b, F1,17 = 5·7, P = 0·03, Fig. 2d), suggesting that this defence is characteristic for derived families.

There was a significant difference across geographical zones (GLM-qb, F3,151 = 15·2, P < 0·0001): Batesian mimicry was more frequent in tropical species and least frequent in temperate species (Fig. 3a). There was also a significant difference among guilds (GLM-qb, F2,151 = 94·5, P < 0·0001): Batesian mimicry was most frequent in cursorial species and least frequent in burrowing species (Fig. 3b). In addition, there was significant effect of diel activity (GLM-qb, F1,36 = 179·8, P < 0·0001): Batesian mimicry was used almost only by diurnal species (Fig. 3c).

Generic-level analyses

Batesian mimicry had evolved in two lineages within Eresidae, Linyphiidae and Araneidae, in three lineages in Thomisidae, in four lineages within Theridiidae, in five lineages in Zodariidae, and in many lineages in Salticidae (Figs 4 and 5). The estimation of ancestral states in the nine families revealed that Batesian mimicry was the global root condition in Corinnidae (Fig. 4).

Discussion

Analysis of the available data shows that spiders seem to use several categories of passive defences described in animals. Crypsis (including countershading and disruptive coloration) was the most frequently proposed category of antipredator defence utilised by spiders coming from a variety of families. These species are mostly active during the night when predators with good eyesight are not active. Some cryptic species are, however, diurnal – actively moving during the day. Such species can change their body colour (e.g. Théry & Casas 2002; Heiling, Herberstein & Chittka 2003) or adopt a special stealthy gait (Bristowe 1941; Cloudsley-Thompson 1995).

A number of species were proposed to use masquerade. These are often nocturnal. Certain types of masquerade allow the species to stay active during the day when combined with a special behavioural strategy. For example, species imitating debris, such as Portia, avoid detection by adopting stealthy movement when moving on the litter (Cloudsley-Thompson 1995).

Few spider species were proposed to be aposematic but evidence to support this hypothesis is still needed. Araneid spiders, such as Micrathena, Macrathena or Gasteracntha, were suggested to advertise the sharp and hard spines on their abdomen, which could harm the soft mouth parts of predatory vertebrates (Edmunds & Edmunds 1983). These spiders are, unlike other relatives, diurnal advertising the conspicuous body colour from their web. It is very likely that some aposematic species, such as Micrathena, are Müllerian mimics – if few species occur sympatrically (Moya-Laraño et al. 2013). In the theridiid black widow species and the theraphosid species, the aposematic pattern on the abdomen may advertises either their potent venom or other forms of effective defence (Vetter 1980). This awaits to be confirmed.

Finally, Batesian mimicry was proposed almost as frequently as crypsis. Such a high frequency is probably a scientific bias due to the conspicuous appearance of mimetic species. Visual mimics are active during the day and occur sympatrically with their models (Edmunds 1974). The most frequently proposed models of mimicking spiders are ants. Ants are wingless, have a rather similar body shape and size, and occur in all types of terrestrial habitats, and are therefore abundant models for spiders (Cushing 1997). Yet, the high frequency of ant-mimics might also be a result of bias. Ant-mimicry, or myrmecomorphy, is often reported without experimental evidence and comparison with the putative model. Therefore, probably for this reason, more frequently reported than other types of defences.

The estimation of ancestral state revealed that crypsis combined with anachoresis is the most basic defence category for several spider families. Similar pattern was found in other animal groups: butterflies (Valin et al. 2006; Kunte 2009), carnivorous mammals (Stankowich, Caro & Cox 2011) and cicindellid beetles (Vogler & Kelley 1998). However, could crypsis be derived, that is, could it evolve from aposematism or Batesian mimicry? As some mimetic and aposematic defences in spiders – in particular in Theridiidae – were found at intermediate positions of trees, there is room to expect reversed evolution – from mimicry to crypsis. Using phylogenetic analysis of nymphalid butterflies, Prudic & Oliver (2008) found that in locations where a model is absent, nymphalids reverted back to a cryptic form. Such could be the case for theridiid spiders. However, Kunte (2009) questioned the reversed evolution in butterflies. According to his arguments, the loss of aposematism or Batesian mimicry cannot happen because Batesian mimicry is a complex adaptation involving supergenes and multiple alleles.

I expected that the occurrence of masquerade, aposematism and Batesian mimicry, which include alteration of body shape (besides coloration), would occur in derived genera mainly. The evolution of these defences probably requires a series of mutations with large effects, since in butterflies, the mimetic wing pattern was regulated by a single gene with a major effect (Turner 1977). Thus, once aposematism or Batesian mimicry evolved, such species would be at the peak in the adaptive landscape and subsequent diversification should produce species with a similar phenotype. It seems that the particular defences have evolved several times independently. At the family level, certain defences were restricted to certain clades: aposematism to Orbiculariae, and Batesian mimicry to Dionycha. Specifically, in Zodariidae, all genera at derived positions were mimetic. Similarly, in Salticidae and Corinnidae, several clades of many genera were mimetic. This is in agreement with other studies. Aposematism, for example, evolved independently several times in mammals in the middle of a tree (Stankowich, Caro & Cox 2011), namely in herpestids, mustelids and viverrids. Similarly, Batesian mimicry in tropical Ithomia (Jiggins et al. 2006) and Papilio butterflies (Prudic, Oliver & Sperling 2007; Kunte 2009) evolved several times independently.

The significant effects of selected ecological variables found in this study suggest that the evolution of defences in spiders is related to the ecology of the species. The examples of anachoresis, crypsis and masquerade came from nocturnal species, while the examples of aposematism and mimicry came from diurnal species. Similar patterns were reported for butterflies (Merilaita & Tullberg 2005). The type of a foraging guild as defined in this study (see Material and methods) is strongly related to the category of defence due to the affinity to different microhabitats. The guild of web-building species was found to use mainly masquerade probably because debris or stalks, which they imitate, often get stuck in the web. Aposematism was also found particularly in the web-building species, such as araneids, that produce aerial webs in the open. Batesian mimicry was, instead, found to be mainly restricted to the cursorial guild in this study because the most frequent model (ants) are non-stationary and forage on a solid substrate, either the ground, bark or leaves.

I found that Batesian mimicry is particularly more frequent in the species occurring in the tropics and crypsis is, in turn, most frequent in the species occurring in the temperate zone. Such a disparity can be explained, for example, by the ‘ecological adaptation’ model, which assumes different selection forces produced by predators in the tropics and the temperate zone, or the ‘Pleistocene refuge’ model, in which the higher diversification rate of models (and Batesian mimics) is caused by the absence of ice glaciers in the tropics (Mallet, Jiggins & McMillan 1996). There is higher predation pressure (Schemske et al. 2009) as well as a greater number of ant species models in the tropics than in the temperate zone (Hölldobler & Wilson 1990). Thus, the higher diversity of Batesian mimics in the tropics could be the result of both effects. Additionally, the lower diversity of spiders in temperate zones may decrease interspecific competition and selection for diurnal activity, which is most associated with aposematism and Batesian mimicry.

The evolution of primary defences in spiders can be read from the fossils. If predators were exerting the selection pressure driving the evolution of mimetic resemblance in spiders, then we could predict that we would not find fossil spiders that resembled, for example, bird droppings before birds had evolved. So the bird-dropping masquerade must have evolved no sooner than in the upper Cretaceous, when modern birds radiated (Ericson et al. 2006). Similarly, the masquerade of fruits or flowers could only have evolved once angiosperms diversified in the early Cretaceous (Taylor et al. 2006). Frog mimics must have evolved later than the Jurassic; wasp and ant-mimics not earlier than in the Cretaceous (Grimaldi & Engel 2005). Indeed, Orbiculariae, being rich in masquerade, first appeared in the Jurassic, and families from the Dionycha clade (Zodariidae, Salticidae, Corinnidae, Thomisidae) that imitate ants appeared in the Cretaceous (Vollrath & Selden 2007). Unfortunately, colour is rarely preserved in fossils, and thus, it is difficult to judge some categories of defence (such as crypsis) from preserved material. In Baltic amber, that is, from the upper Cretaceous, crypsis was found in several species of Salticidae, Pisauridae, Oxyopidae, Uloboridae and Thomisidae – virtually in the same families as among the extant taxa (Wunderlich 2004). In Baltic or Dominican amber from the Tertiary, ant-mimicry was found in Zodariidae, Corinnidae, Dysderidae, Segestriidae, Spatiatoridae, Archaeidae and Salticidae (Wunderlich 1986, 2004). This suggests that Batesian mimicry evolved rather soon after the origin of ants.

Although I collected hypotheses for more than thousand species of spiders, it is a small number when considering total spider diversity that is amounting to more than 40 000 species (Platnick 2011). The scarcity of hypotheses on non-mimetic species clearly biased the estimate of some ancestral conditions, particularly in Salticidae and Corinnidae. The current results are also constrained by the current state of the art in spider phylogenetics (e.g. polytomy in the Dionycha clade). In spite of these shortcomings, this study revealed that the evolution of passive defences in spiders was influenced by species geographical distribution, foraging guild, diel activity and to some extent by phylogeny. Effect of other important drivers, such as habitat or composition of enemies, remains to be investigated.

Acknowledgements

I want to thank T. Blackledge and J. Bosselaers for valuable advice on the phylogeny of particular families. Furthermore, I wish to thank Robert Jackson and one anonymous reviewer for constructive comments that markedly improved the earlier version of the manuscript.

Ancillary