Convergent evolution of eye ultrastructure and divergent evolution of vision-mediated predatory behaviour in jumping spiders


Daiqin Li, Department of Biological Sciences, National University of Singapore, 14 Science Drive 4, Singapore 117543.
Tel.: +65 65164372; fax: +65 67792486; e-mail:


All jumping spiders have unique, complex eyes with exceptional spatial acuity and some of the most elaborate vision-guided predatory strategies ever documented for any animal of their size. However, it is only recently that phylogenetic techniques have been used to reconstruct the relationships and key evolutionary events within the Salticidae. Here, we used data for 35 species and six genes (4.8 kb) for reconstructing the phylogenetic relationships between Spartaeinae, Lyssomaninae and Salticoida. We document a remarkable case of morphological convergence of eye ultrastructure in two clades with divergent predatory behaviour. We, furthermore, find evidence for a stepwise, gradual evolution of a complex predatory strategy. Divergent predatory behaviour ranges from cursorial hunting to building prey-catching webs and araneophagy with web invasion and aggressive mimicry. Web invasion and aggressive mimicry evolved once from an ancestral spartaeine that was already araneophagic and had no difficulty entering webs due to glue immunity. Web invasion and aggressive mimicry was lost once, in Paracyrba, which has replaced one highly specialized predation strategy with another (hunting mosquitoes). In contrast to the evolution of divergent behaviour, eyes with similarly high spatial acuity and ultrastructural design evolved convergently in the Salticoida and in Portia.


Jumping spiders (Family: Salticidae) have high spatial acuity eyes that can support a very rich and complex repertoire of vision-mediated predatory behaviour. With 5035 described species in 553 genera, the Salticidae is the most speciose spider family (Platnick, 2006). Yet the position of Salticidae within the Araneae and the relationships within Salticidae have only recently been addressed using numerical cladistic techniques (e.g. Maddison & Hedin, 2003; Maddison & Needham, 2006). Of the major groups within the Salticidae (Maddison & Hedin, 2003; Maddison & Needham, 2006), it is the Salticoida (higher salticids) that accounts for 90% of the species. Yet the relatively species-poor Lyssomaninae and Spartaeinae, which have both been variously described as ‘primitive’ or ‘basal’, are crucial for understanding the higher level systematics and early evolution of salticids. Whereas there is good evidence for the monophyly of the Salticoida, the monophyly of the Lyssomaninae and Spartaeinae and their position within the Salticidae remain controversial (Maddison & Hedin, 2003).

Salticid eye ultrastructure

Information on the retinal anatomy of salticids comes from TEM studies of 28 species, in 25 genera (Land, 1969; Eakin & Brandenburger, 1971; Wanless, 1980a, 1982; Williams & McIntyre, 1980; Blest & Price, 1984; Blest & Sigmund, 1984, 1985; Blest, 1985; Blest et al., 1988). It is the unique forward-facing anterior-median pair of eyes (principal eyes) that allow for high spatial acuity vision (Homann, 1928; Crane, 1949; Land, 1969) enabling some species to have a spatial resolution of ≈0.04°, whereas the highest spatial acuity known for insects of similar size is only ≈0.4° (Land & Nilsson, 2002; human eye: ≈0.007°).

In all salticids, the retina lies at the end of a long eye tube, and at the front end of the eye tube is a large corneal lens (Land, 1971, 1985). Immediately in front of the retina there is a second lens that turns the principal eye into a telescope (Williams & McIntyre, 1980). The retina is divided into four tiers and it is the rearmost layer (layer I) that takes primary responsibility for tasks requiring high spatial acuity (Land, 1969; Blest et al., 1988, 1990). The receptors in the central region of layer I are especially slender in transverse section and packed close together, providing the sampling mosaic needed for exceptional spatial acuity (Land & Nilsson, 2002). What makes layer I of particular interest is that across the Salticidae, and especially within the Lyssomaninae and Spartaeinae, there are numerous differences in receptor structure that are linked to eye performance. Although all salticids have higher spatial acuity than that found in other spiders, we here restrict the term ‘high acuity’ to salticid species that have retinas organized with minimal optical pooling and a particularly tight, regular sampling mosaic that provides spatial acuity of ≈0.04°. High spatial acuity is found in the Salticoida and some spartaeines (Portia). The price for having a finer grain receptor mosaic is reduced sensitivity to light, resulting from a lower photon catch per receptor. In salticids with high spatial acuity, this potential loss in sensitivity is compensated for by increases in the length of the photoreceptors and receptor structure changes that allow receptors to act as light guides (Blest et al., 1990).

However, in the absence of a phylogenetic tree, it remains unclear whether high spatial acuity was already a feature of the salticid stem species or evolved multiple times. Here, we address this question by tracing high spatial acuity on a phylogenetic tree.

Diverse predatory behaviours

It was suggested that acute vision might have been essential for the evolution of the complex predatory behaviours of Salticidae (Jackson, 1992; Jackson & Pollard, 1996). Most salticoid species that have been studied are cursorial hunters that see prey from a distance, stalk until close and then attack by leaping (Drees, 1952; Forster, 1977, 1982; Richman & Jackson, 1992). They prey primarily on insects and building webs is not normally part of the spiders' repertoire. However, every studied spartaeine departs from this repertoire. Spartaeines adopt different predatory strategies (Table 1). This includes various combinations of web building, araneophagy, web invasion and aggressive mimicry (Jackson & Pollard, 1996).

Table 1.   List of species, character scores and references.
SpeciesWeb-buildingGlue immunityAraneophagyReferences
  1. Web-building: 0 = absent, 1 = present. Glue immunity: 0 = absent, 1 = present. Araneophagy: 0 = absent, 1 = ambush predation, 2 = web invasion only, 3 = aggressive mimicry.

Cocalus murinus Simon112Jackson (2000), Cerveira et al. (2003), D. Li, unpublished data
Cyrba algerina (Lucas)113Jackson & Hallas (1986b), Jackson (1990a, 2000), Jackson & Li (1998)
Cyrba ocellata (Kroneberg)113Wanless (1984), D. Li, personal observation
Cyrba sp.113New species; D. Li, unpublished data
Gelotia springopalpis Wanless113Jackson (1990c), D. Li, unpublished data
Holcolaetis vellerea Simon101Wanless (1985), R. R. Jackson, unpublished data
Mintonia ramipalpis (Thorell)101Wanless (1984), D. S. H. Tay & D. Li, unpublished data
Neobrettus tibialis (Proszynski)113T. M. Wong & D. Li, unpublished data
Paracyrba wanlessi Zabka & Kovac??0Zabka & Kovac (1996), J. R. W. Woon and D. Li, unpublished data
Phaeacius malayensis Wanless101Jackson (1990d), Li (2000)
Phaeacius yixin Zhang & Li101Zhang and Li (2005), D. Li, unpublished data
Portia africana (Simon)113Jackson & Hallas (1986a), Li et al. (1997)
Portia fimbriata (Doleschall)113Jackson & Blest (1982), Li & Jackson (1996b)
Portia heteroidea Xie & Yin113D. Li, unpublished data
Portia jianfengensis Song & Zhu113D. Li, unpublished data
Portia labiata (Thorell)113Jackson & Hallas (1986a), Li et al. (1997)
Portia sp.113D. Li, unpublished data
Portia quei Zebka113D. Li, personal observations
Spartaeus jianfengensis Song & Chai100D. Li, personal observations
Spartaeus platnicki Song, Chen & Gong100D. Li, personal observations
Spartaeus thailandicus Wanless100Wanless (1987), Jackson & Pollard (1990)
Spartaeus wildtrackii Wanless100Wanless (1987), D. Li, personal observation
Yaginumanis wanlessi Zhang & Li???Zhang and Li (2005)
Asemonea sichuanensis Song & Chai100Wanless (1980a), Jackson (1990e), D. Li, unpublished data
Lyssomanes viridis (Walckenaer)100Jackson, 1990e
Onomastus nigrimaculatus Zhang & Li100Zhang and Li (2005); D. Li, unpublished data

Prey-catching webs

The term ‘web’ most appropriately refers to silk structures that are considerably larger than the web-building spider and are used in prey capture (Jackson, 1985a; Shear, 1994). Except for some spartaeines, lyssomanines and a few salticoids, web building is absent in salticids. The spinning of cocoon-like nests is common in the Salticoida, but these cocoons function as nests used for resting, moulting, mating and egg laying (Jackson, 1979). Few web-building salticids are known within salticoids, and they are Plexippus paykulli (Jackson & Macnab, 1989; also see Hallas & Jackson, 1986), Pellenes arciger (Lopez, 1986), Euryattus sp. (Jackson, 1985b) and Simaetha sp. (Jackson, 1985c). The only spartaeines known to build space webs are Portia and Gelotia, whereas other spartaeines, and lyssomanines only spin small silk sheets (Jackson, 1990e, 1992), or in the case of Spartaeus (Jackson & Pollard, 1990) sheets of moderate size. In some instances, these silk sheets may play a role in prey capture.


Specialized predation on other spiders (araneophagy) is another unusual predatory strategy of spartaeines. Various salticid species occasionally feed on other spiders (Jackson, 1986; Edwards & Jackson, 1993), but araneophagic salticids have prey-specific capture behaviour and show a distinctive preference for spiders as prey (Jackson, 1992; Jackson & Pollard, 1996; Li et al., 1997; Jackson et al., 1998; Jackson & Li, 1998; Jackson, 2000; Li, 2000). Araneophagy is found in the spartaeine genera Brettus, Cocalus, Cyrba, Gelotia, Holcolaetis, Mintonia, Neobrettus, Phaeacius and Portia (see Table 2 and Jackson, 1990b; Li & Jackson, 1996a; Jackson et al., 1998; D. Li, unpublished data). It ranges from hunting or ambushing of spiders outside the webs (e.g. on tree trunks and boulders: Holcolaetis and Phaeacius) (Jackson & Hallas, 1986b; Jackson, 1990d; R.R. Jackson, unpublished data) to the capturing of host spiders after invading their webs (Jackson, 1992; Jackson & Pollard, 1996). Web-invading araneophagic spartaeines enter alien webs to prey on the web-building spider, and have glue immunity; i.e. they have the remarkable ability to walk, without adhering, across cribellate and ecribellate sticky webs. This is a surprising feature because no salticid builds sticky webs (Jackson & Pollard, 1996).

Table 2.   Taxonomic and locality information of the specimens included in the molecular analyses and GenBank accession numbers.
Asemonea sichuanensisChina: SichuanEF419017EF419051EF418986EF419082NANANA
Cocalus murinusSingaporeEF419019EF419053EF418988EF419084EF419116EF418959EF419140
Cosmophasis umbraticaSingaporeEF419020NANAEF419085EF419117EF418960EF419141
Cyrba algerinaKenyaEF419021EF419054EF418989EF419086NAEF418961EF419142
Cyrba ocellataChinaEF419022EF419055EF418990EF419087NAEF418962EF419143
Cyrba sp.KenyaEF419023EF419056EF418991EF419088NANANA
Gelotia springopalpisChina: HainanEF419024EF419057NAEF419089EF419118NANA
Holcolaetis vellereaKenyaEF419025EF419058EF418992EF419090EF419119EF418963EF419144
Ligurra latidensSingaporeEF419026EF419059EF418993EF419091EF419120EF418964EF419145
Lyssomanes viridisUSA: FloridaEF419027EF419060EF418994EF419092EF419121EF418965EF419146
Mintonia ramipalpisSingaporeEF419028EF419061EF418995EF419093EF419122EF418966EF419147
Neobrettus tibialisMalaysia: Genting HighlandsEF419030EF419063NAEF419095EF419124NANA
Onomastus nigrimaculatusChina: YunnanEF419031EF419064EF418997EF419096EF419125EF418968EF419149
Paracyrba wanlessiMalaysia: GombakEF419033EF419066EF418999EF419098NANANA
Phaeacius malayensisChina: YunnanEF419034EF419067EF419000EF419099NAEF418970EF419151
Phaeacius yixinChina: HainanEF419035EF419068EF419001NANAEF418971EF419152
Plexippus paykulliSingaporeEF419036NAEF419002EF419100EF419127EF418972EF419153
Portia africanaKenyaEF419037EF419069EF419003EF419101EF419128NANA
Portia fimbriataSingaporeEF419038EF419070EF419004EF419102EF419129EF418973EF419154
Portia heteroideaChina: SichuanEF419039EF419071EF419005EF419103EF419130EF418974EF419155
Portia jianfengensisChina: HaiananEF419040EF419072EF419006EF419104NAEF418975EF419156
Portia labiataSingaporeEF419041EF419073EF419007EF419105EF419131EF418976EF419157
Portia queiChina: YunnanEF419042EF419074EF419008EF419106EF419132EF418977EF419158
Portia sp.China: SichuanEF419043EF419075EF419009EF419107EF419133EF418978EF419159
Rhene sp.Malaysia: Cameron HighlandsEF419044NAEF419010EF419108EF419134EF418979EF419160
Spartaeus jianfengensisChina: HainanEF419045EF419076EF419011EF419109NAEF418980EF419161
Spartaeus platnickiChina: HainanEF419046EF419077EF419012EF419110EF419135EF418981EF419162
Spartaeus thailandicusChina: YunnanEF419047EF419078EF419013EF419111EF419136EF418982EF419163
Spartaeus wildtrackiiMalaysiaEF419048EF419079EF419014EF419112EF419137EF418983EF419164
Thiania bhamoensisSingaporeEF419049EF419080EF419015EF419113EF419138EF418984EF419165
Yaginumanis wanlessiChina: SichuanEF419050EF419081EF419016EF419114EF419139EF418985EF419166
Cheiracanthium sp.SingaporeEF419018EF419052EF418987EF419083EF419115NANA
Hibana sp.GenBankAY297295NANAAY297422NAAY296713AY297358
Misumenops nepenthicolaSingaporeEF419029EF419062EF418996EF419094EF419123EF418967EF419148
Oxyopes birmanicusSingaporeEF419032EF419065EF418998EF419097EF419126EF418969EF419150

Two kinds of web invasion are used by spartaeines. In web invasion without aggressive mimicry (Cocalus), Cocalus murinus stalks very slowly across webs to prey on the resident spider (Jackson, 1990b). Cocalus generally do not venture far into the web and sometimes stay on the edge for hours (Jackson, 1990b). All the remaining web-invading spartaeines (Brettus, Cyrba, Gelotia, Neobrettus and Portia) deploy aggressive mimicry with Portia’s strategy being the most complex (see Table 1 and Wilcox et al., 1996; Tarsitano & Jackson, 1997; Clark & Jackson, 2000; Jackson et al., 2002; Jackson & Li, 2004). Here the web invader uses signalling (vibrations on the web) to control the behaviour of host spider.

The highly versatile and complex predatory strategies in the spartaeines have attracted considerable interest (Jackson & Blest, 1982). However, in the absence of a phylogenetic tree, all evolutionary inferences were highly speculative. A hypothesis concerning the joint evolution of high-acuity vision and intricate predatory strategies was outlined in the first detailed study of Portia’s predatory strategy (Jackson & Blest, 1982; Jackson, 1986). It was suggested that salticid ancestors, prior to the evolution of refined high-acuity eyes, were web-building spiders that lived in habitats where webs of various spider species were abundant and often contiguous. Another important part of the hypothesis (Jackson & Blest, 1982) was that ultrastructural modifications of salticid retinas underlying high-acuity vision evolved in conjunction with a web-invading predator that was becoming a wide-spectrum aggressive mimic. This hypothesis can be rigorously tested using phylogenetic tools.

Materials and methods

Taxon sampling

Our data set comprised 35 species, including 22 species from 10 genera of the subfamily Spartaeinae, three species from the subfamily Lyssomaninae, one species from the Cocalodes group, five from Salticoida, and four outgroup species from four other families that are considered closely related to Salticidae, Misumenops nepenthicola (Thomisidae), Cheiracanthium sp. (Miturgidae) and Oxyopes birmanicus (Oxyopidae) and Hibana sp. (Anyphaenidae) (Simon, 1901; Petrunkevitch, 1933; Bristowe, 1938; Lehtinen, 1967, 1975; Ono, 1987; Coddington & Levi, 1991; Table 1).

Three species of lyssomanines (refer to Table 1) were included because Lyssomaninae has been proposed as the sistergroup of Spartaeinae (Wanless, 1980a, b, 1984; Blest & Carter, 1987; Maddison, 1988; Rodrigo & Jackson, 1992; Maddison & Hedin, 2003; Maddison & Needham, 2006). Five species of the Salticoida were chosen to represent the major groups recognized by Maddison & Hedin (2003): euophryines (Thiania bhamoensis), heliophanines (Cosmophasis umbratica), marpissoids (Rhene sp. indt.) and plexippoids (P. paykulli).

DNA extraction and sequencing

The DNA from whole spiders or spider legs was extracted using a modified CTAB extraction protocol (Shajahan, 1995). PCR was used to amplify DNA fragments from six genes (primers in Table 3). Cycling conditions for the primers started with an initial 95°C denaturation, followed by 30 cycles of 1 min at 95 °C, 1 min at 55 °C (28S), 50 °C (18S and H3), 46 °C (16S/ND1) or 48 °C (COI) and 1.5 min at 72 °C, and a final 2-min extension at 72 °C. PCR was carried out using the Hotstart Ex-Taq (Takara, Shiga, Japan). The amplified fragment was purified with the QIAquick PCR Purification kit (Qiagen, Valencia, USA). The purified PCR products were sequenced directly in both directions using an ABI 3100 (Applied Biosystems, Foster, California).

Table 3.   List of primers
COIC1-J-2309 (Masta, 2000)
C1-N-2776 (Maddison & Hedin, 2003)
18S18Sa2.0 (Giribet et al., 1999)
18S9R (Giribet et al., 1999)
28S128S ÒOÓ (Maddison & Hedin, 2003)
28S ÒCÓ (Maddison & Hedin, 2003)
28S228SRd4.5a (Whiting et al., 1997)
5Õ-GAC TTC CCT TAC CTA CAT-3Õ (Hausdorf, 1999)
Histone 3H3aR (Colgan et al., 1998)
H3aF (Colgan et al., 1998)
16S/ND1N1-J-12261 (Maddison & Hedin, 2003)
LR-N-12945 (Maddison & Hedin, 2003)

Phylogenetic analyses

Fragments were assembled into contigs in Sequencher (Gene Codes Corp., Ann Arbor, MI, USA). Multiple alignments were carried out in Clustal X (Higgins & Sharp, 1988) for 28S and 16S using the gap opening/gap extension ratios recommended by Maddison & Hedin (2003). Afterwards the alignments were manually adjusted in MacClade 4.06 OS X (Maddison & Maddison, 2003). 18S was first aligned in Clustal X (gap opening/gap extension ratio: 15/6.66) (Higgins & Sharp, 1988) and manually adjusted in MacClade 4.06 OS X (Maddison & Maddison, 2003). The DNA sequences for protein-encoding genes were aligned based on amino acid sequences in DAMBE (Xia & Xie, 2001), using a gap opening/gap extension ratio of 10 : 0.1 and the protein weight matrix BLOSUM. The leading and trailing bases that had been discarded by DAMBE were re-inserted.

The data set was subjected to a parsimony analysis in TNT version 1.0 (Tree Analyses Using New Technology; Goloboff et al., 2000). To explore if the indels in the alignments affects the resulting phylogenetic hypothesis, two separate analyses were carried out. The first with gaps treated as missing data and the second with gaps treated as fifth character state. A driven search was carried out at level 99 on TNT (with sectorial search), and the minimum length was found three times. Support for internal nodes was assessed using bootstrap and Bremer support. Bootstrap values were calculated in TNT (100 replicates) by using the ‘New Tech’ option utilizing the same search parameters as mentioned above. Bremer support values were computed in PAUP* version 4.0b10 (Swofford, 2002) using TreeRot (Sorenson, 1999; 100 random sequence-addition repetitions, TBR for each constrained search).

A Bayesian analysis was conducted in MrBayes 3.1 (Huelsenbeck & Ronquist, 2001). The GTR + I + G model was favoured by the Akaike information criterion and hierarchical likelihood ratio testing was implemented in MrModeltest version 2.2 (Nylander, 2004). The data set was analysed for 3 000 000 generations and a tree was sampled every 300 generations. Chain stationarity was achieved after 1 200 000 (burn-in) and 4000 trees were subsequently discarded. Three independently repeated analyses resulted in similar tree topologies, comparable clade probabilities and substitution model parameters, which suggested that reasonable estimates of the posterior probability (PP) distributions had been obtained.

Analyses of gene partitions

To assess whether mitochondrial or nuclear data provided greater amounts of support, partitioned Bremer supports (PBS) were determined for mitochondrial and nuclear genes with gaps coded as fifth character state. The amount of homoplasy and divergence in the nuclear and mitochondrial genes was assessed using the consistency index (CI), retention index (RI) and pairwise distances, as determined in PAUP* version 4.0b10 (Swofford, 2002).

Characters pertaining to predatory behaviour and eye ultrastructure

An extensive literature review on the predatory behaviours of Spartaeinae and eye ultrastructure of Salticidae was carried out and provided the following characters that were mapped onto the phylogenetic tree using MacClade 4.06 OS X (Maddison & Maddison, 2003) (Table 1). (1) Web building is defined as building of silk devices that are larger than the host spider and are used in prey capture: absent = 0; present = 1 (see Jackson, 1985a, 1992; Shear, 1994; Jackson & Pollard, 1996; Li & Jackson, 1996a; Jackson & Wilcox, 1998; Harland & Jackson, 2004). (2) Araneophagy is defined as targeting and preferring spiders as prey in standardized prey-choice experiments: absent = 0; ambushing = 1; web invasion = 2; aggressive mimicry = 3. The character was once mapped as additive (ordered) and once as nonadditive (unordered). (3) Glue immunity is the ability to enter and walk across sticky webs: absent = 0, present = 1. (4) With regard to high spatial acuity eyes, the literature data are comparatively incomplete, and it is here assumed that congeneric species share similar eye ultrastructure morphology. Given these uncertainties, only one unambiguous character was defined. High spatial acuity eyes with elongated and single rhabdomere receptors optimally placed throughout the layer I retina: absent = 0, present = 1 (Blest et al., 1990). Such high spatial acuity has been confirmed for the following salticoid genera: Amycus, Corythalia, Fluda, Holoplatys, Itata, Jollas, Helpis, Metaphidippus, Myrmarachne, Phiale, Plexippus, Scopocira, Synemosyna, Thiodina and Trite and the spartaeine genus, Portia (Blest & Carter, 1987; Blest et al., 1990). A lack of high spatial acuity eyes is known for: Allococalodes, Asemonea, Brettus, Chinoscopus, Cyrba, Lyssomanes, Spartaeus and Yaginumanis (see Blest & Carter, 1987; Blest et al., 1990). Based on these data, we can here code the following genera as lacking high spatial acuity: Asemonea, Cyrba, Lyssomanes, Spartaeus and Yaginumanis, and the following as having high spatial acuity: Plexippus and Portia.


We obtained about 4.8 kb of sequences data for 35 species each (Table 2). Using equal weighting parsimony analysis of the gap = missing data set, we found five parsimonious trees (length: 6166) whereas using the gap = fifth character state analysis yielded three parsimonious trees (length: 6907; Fig. 1). The trees had similar topologies, bootstrap support (BS) and Bremer supports (BrS) (see Fig. 1). The posterior probability tree from the Bayesian analysis yielded a nearly identical tree. The only discrepancy is the lack of support in a few branches and conflict within the genus Spartaeus (Fig. 1). More than two-thirds of all the clades on the phylogram have 100% PP (see bold lines in Fig. 1).

Figure 1.

 Strict consensus tree of the MPTs (most parsimonious trees) from the analysis using gaps. (a) and (b) indicate areas that differ from the gaps = missing data tree. Bootstrap values are added below the nodes (bootstrap for gaps = fifth state/gaps = missing values). Numbers above each node refer to Bremer support values. Bold lines indicate nodes supported with 100 posterior probability in the Bayesian analysis.

The data support monophyly of Salticidae in all analyses, with a BS of 77, BrS of 11 and a PP of 100. The Spartaeinae is monophyletic (BS 85, BrS 5 and PP 100) (Fig. 1), with Holcolaetis being sistergroup to the remaining spartaeines (BS 87, BrS 13, PP 100) (Fig. 1). The Salticoida is monophyletic (BS 100, BrS 27, PP 100), whereas the Lyssomaninae are paraphyletic in the parsimony and Bayesian analyses. On the parsimony tree, Lyssomanes or Asemonea + Onomastus are sistergroup to the Spartaeinae and these two subfamilies are then sistergroup to the Salticoida (Fig. 1). On the Bayesian posterior probability tree, the higher level relationships within Salticidae remain unresolved.

Performance of gene partitions

Nuclear genes provided greater overall support than mitochondrial genes (Table 4). The overall highest support (ca 50%) comes from 28S1 (PBS = 147.9), which exceeds the support from all mitochondrial genes (PBS = 147.3) (Table 4). The second and third best performers were COI (PBS = 63.8) and ND1 (PBS = 56.6) respectively (Table 4). The poorest performer was 28S2 (PBS = 14.3). The mitochondrial gene COI had the lowest average CI (0.315) and RI (0.299) values of all the partitions, whereas the nuclear gene 18S had the highest average CI (0.638) and RI (0.650) values (Table 5). The 16S gene partition showed the greatest pairwise distance of all the partitions, whereas 28S2 partition showed the lowest pairwise distance (Table 6).

Table 4.   Partitioned Bremer support values for each gene partition (gaps = fifth character state).
All mitochondrial genes147.3
All nuclear genes214.7
Table 5.   CI and RI values for the seven gene partitions (gaps = fifth character state).
GenesCI/RI values for most parsimonious trees
  1. CI and RI values are averages. CI, consistency index; RI, retention index.

Table 6.   Uncorrected pairwise distances (in percentage).
GenesWithin SpartaeinaeAcross all species

Evolution of behaviour and spatial acuity

Araneophagy evolved once at the base of Spartaeinae and was lost twice. Web invasion with aggressive mimicry evolved from simple web invasion. When araneophagy is coded as unordered, there is a second equally parsimonious optimization where simple web invasion is derived from web invasion with aggressive mimicry. Silk sheets were at the base of the Spartaeinae and Lyssomaninae, and large space webs have evolved twice. Eyes with high spatial acuity have evolved twice in Salticidae.


Salticid evolution

Resolving the relationships within the speciose Salticidae is one of the main priorities in reconstructing the tree-of-life for spiders, and much progress has been made in recent years (Maddison & Hedin, 2003; Maddison & Needham, 2006). In Maddison & Hedin (2003), the focus was on the relationships within the most speciose salticid clade, Salticoida. Our study is complementary in addressing the higher level relationships between Salticoida, Spartaeinae and Lyssomaninae. Maddison & Hedin (2003) discussed morphological evidence for the monophyly of Salticidae, which was also supported in their analyses based on molecular data. Here, we again find support for the monophyly of Salticidae (BS 77, BrS 11 and PP 100). The same applies to the Salticoida (BS 100, BrS 27 and PP 100) that were already strongly supported in previous studies (Maddison, 1996; Maddison & Hedin, 2003; Maddison & Needham, 2006). But in our study, we also present strong molecular evidence for the monophyly of Spartaeinae based on a comparatively large taxon sample (BS 85, BrS 5 and PP 100; Fig. 1). We find that similar to Maddison & Needham (2006), Holcolaetis, a member of the Cocalodes group (Allococalodes, Cocalodes, Holcolaetis and Sonoita) is sistergroup to the Spartaeinae sensu stricto, and should thus be included in the Spartaeinae.

However, one important issue in salticid systematics remains unsatisfactorily resolved. Ever since Blackwall (1877) first proposed a separate family for the lyssomanines, there has been uncertainty regarding its monophyly. Maddison & Needham (2006) suggested splitting the lyssomanines into two groups. Our tree similarly suggests that the Lyssomaninae is paraphyletic and forms a grade in the same clade that also includes the Spartaeinae. However, the support for this conclusion in the parsimony analyses is weak and lacking on the Bayesian tree. Overall, our tree is quite different from some previous hypotheses that considered both the Lyssomaninae and Spartaeinae to be basal grades within the Salticidae (Wanless, 1984; Maddison, 1988, 1996; Rodrigo & Jackson, 1992). In our analyses, the nuclear genes out-performed the mitochondrial genes in providing overall support (PBS). However, this is mostly due to 28S1, which provide more support than all the mitochondrial genes combined. Hence, the results show that different sections of the same gene (28S) can provide vastly different amounts of support.

Divergent evolution of predatory behaviour in Salticidae

Our study documents strongly divergent evolution of predatory behaviour in the Salticidae, where the behaviours range from cursorial hunting and building prey-catching webs to araneophagy involving web invasion and aggressive mimicry. When the multistate character ‘araneophagy’ is coded as ordered, araneophagy evolved once at the base of the Spartaeinae and was lost twice (in Spartaeus and Paracyrba; Fig. 2). This suggests that the ancestral spartaeine was araneophagic, but with araneophagy being expressed as hunting or ambushing spiders on tree trunks and boulders as seen in the spartaeines Holcolaetis and Phaeacius (Jackson, 1990d; Li, 2000; R. R. Jackson, unpublished data). Web invasion and aggressive mimicry evolved later within the Spartaeinae. This scenario requires that different modes of araneophagy be homologized across salticids. For example, araneophagy by ambushing is homologized with araneophagy by invading webs and one could argue against our treatment of the character and regard ‘araneophagy’ alone as insufficient evidence for homology. However, we submit that even though the behaviours for prey catching are very different, the search image and preference for spiders could be the homologous trait. Alternatively, one could regard ambush predation and web invasion as discrete characters, and then araneophagy would have evolved twice in Spartaeinae, once as ambush predation and once as web invasion in the web-invasion clade (see below).

Figure 2.

 Evolution of predatory behaviours in Salticidae (bold black lines = araneophagy present, thin lines = araneophagy absent, shaded lines = feeding behaviour unknown). ⋆ indicates Loss of web invasion, aggressive mimicry, and araneophagy. ♣ indicates building of large space webs. Numbers in brackets refer to character and states. Character 2 is mapped additively.

Web invasion, glue immunity and aggressive mimicry

Web invasion when mapped onto the salticid tree has evolved once in what we will refer to as the web-invasion clade (Portia + Cyrba + Gelotia + Cocalus + Neobrettus + Paracyrba; Fig. 2). There are two modes of web invasion, one without (Cocalus) and one with aggressive mimicry (Portia, Cyrba, Gelotia and Neobrettus), where the web invader uses signalling routines for controlling the behaviour of the resident spider. We find that web invasion has been lost once in Paracyrba (Fig. 2), which lives in the hollow interior of decaying bamboo internodes where it preys on aquatic insects, especially mosquito larvae (Zabka & Kovac, 1996). It also has a distinctive preference for mosquitoes over other insects; i.e. it has replaced one highly specialized predation strategy with another (J. R. W. Woon and D. Li, unpublished data).

Web invasion with and without aggressive mimicry only evolved in Spartaeinae. Based on the complexity of the behaviour, one may surmise that the simple web-invasion technique used by Cocalus may be intermediate between araneophagy through ambushing and araneophagy using web invasion and aggressive mimicry. We indeed find this hypothesis compatible with our phylogenetic tree, and it is unambiguously supported when araneophagy is coded as an ordered character. The Spartaeinae thus constitute a good case for the stepwise, gradual evolution of a complex behaviour. Aggressive mimicry evolved via, ambush araneophagy, simple web invasion before arriving at araneophagy via web invasion with aggressive mimicry. However, when the various modes of araneophagy are coded as unordered, the simultaneous origin of web invasion and aggressive mimicry in the web-invasion clade with a subsequent loss of aggressive mimicry in Cocalus is equally parsimonious.

One adaptation that is needed for web invasion and that is seen in all spartaeine web invaders is glue immunity. Experiments have revealed that the majority of salticids have difficulty walking across sticky webs, but some spartaeines have evolved glue immunity (Jackson, 1992; Jackson & Pollard, 1996). For example, ambush predators like Phaeacius never enter webs voluntarily and get stuck when experimentally placed onto sticky webs, whereas all web-invading spartaeines never adhere to sticky or nonsticky webs (Jackson, 1990d, 1992). Hence, glue immunity to sticky webs is an autapomorphy supporting the web-invasion clade, whereas the ability to enter nonsticky webs evolved earlier in spartaeines incapable of web invasion. All previous hypotheses had suggested that the spartaeine ancestor was likely to be a wide-spectrum aggressive mimic resembling extant Portia (Jackson & Blest, 1982). One important implication of our results is that web invasion and aggressive mimicry are not an ancestral trait of Salticidae, and instead evolved within the Spartaeinae.

Web building

We will here only consider as webs silk devices that are considerably larger than the resident spider and are used in prey capture (Jackson, 1985a; Shear, 1994). Most salticids do not build webs (Jackson, 1979). Within Spartaeinae, large space webs have evolved twice independently in Portia and Gelotia. The rest of the lyssomanines and spartaeines, build small silk sheets (Jackson, 1990e, 1992), or in the case of Spartaeus (Jackson & Pollard, 1990), sheets of moderate size, which often have a role in prey capture. Our tree suggests that the ancestor of lyssomanines and spartaeines built such silk sheets (Fig. 2). However, the ancestral character state for Salticidae remains uncertain. Several of the potential sistergroups of Salticidae (e.g. Thomisidae and Oxyopidae) are polymorphic with regard to building prey-catching webs. Resolving the ancestral state for Salticidae will thus not only require identifying the sistergroup of Salticidae, but is also dependent on having additional information on the distribution of prey-catching webs within the closely related families.

Convergent evolution of eye ultrastructure

We observe a remarkable case of convergent evolution of eye ultrastructure in the Salticoida and the spartaeine, Portia, which have evolved divergent predatory strategies (cursorial hunting vs. aggressive mimicry). Both have independently evolved a similar layer I eye ultrastructure that provides high spatial acuity (Fig. 2). The layer I fovea of Portia and the Salticoida are entirely made up of regularly packed light-guiding rhabdomeres, with receptors containing only a single rhabdomere that is positioned maximally distant from the rhabdomeres of adjacent receptors (Williams & McIntyre, 1980; Blest et al., 1990). This reduces ambiguity as it lowers the chance that light guided down one rhabdomere in a receptor cell gets passed into a rhabdomere in a neighbouring cell. Both Salticoida and Portia also have narrow receptors with small transverse profile and correspondingly less photopigment, making them less sensitive per unit length. This potential loss in sensitivity is at least partly abrogated in the salticoids and Portia by an increased receptor length and a reduction of the optical density of the cytoplasm surrounding the rhabdomere so that on entering the light is trapped within the rhabdomere by total internal reflection, i.e. they act as light guides. In salticoids, the cytoplasm surrounding the rhabdomeres contains no organelles other than microtubules, whereas in Portia even the microtubules are completely lost (Blest & Price, 1984).

Within the Salticidae, the loss of organelles and/or microtubules in the cytoplasm is unique to Portia and Salticoida as is the simultaneous occurrence of narrow and long receptors in a layer I fovea that entirely consists of single-rhabdomere receptors. Other spartaeines may have only one or the other feature, such as somewhat lengthened rhabdomeres (e.g. Cyrba) or a mixture of single and double rhabdomere receptors in the layer I fovea (e.g. Spartaeus). Reconstructing the precise pathway that led to high spatial acuity vision within the Spartaeinae would be very rewarding. However, it will require the systematic study of additional species and more standardized morphological descriptions.

Evolution of behaviour and eye ultrastructure

We can use our tree to test previous hypotheses with regard to the evolution of prey catching behaviour and the origin of high spatial acuity. Previously, it has been hypothesized that high spatial acuity evolved in conjunction with web invasion and aggressive mimicry, as invading spiders with the ability to exploit a sensory system independent of web dynamics would gain an adaptive advantage over the host (Jackson & Blest, 1982). However, based on our phylogenetic tree, the ‘lyssomanines’ are more closely related to the spartaeines than to the Salticoida, and therefore, araneophagy is derived within this clade, and not the Salticidae. Furthermore, web invasion and aggressive mimicry are derived within the Spartaeinae. Hence, araneophagy and web invasion are not ancestral salticid traits. Furthermore, the exceptionally high spatial acuity seen in Portia has evolved within the web-invasion clade, probably from eyes of intermediate spatial acuities like those that are observed in Spartaeus. It is important to remember that even salticids that do not practice aggressive mimicry may be among the most vision-dependent spiders and generally have better spatial acuity than most web-building spiders.


Our study was able to resolve many outstanding issues in salticid evolution. We were able to establish the monophyly of Spartaeinae including Holcolaetis. However, more work including a larger taxon sample will be needed to resolve ‘lyssomanine’ relationships. We demonstrate that similar eye ultrastructure that is capable of supporting high spatial acuity has evolved at least twice in the jumping spiders, once in the Salticoida and once within the aggressive mimicry clade in the Spartaeinae; i.e. predatory behaviour specialized for web invasion preceded the origin of sophisticated eyes and even eyes with lower spatial acuity can support such behaviour. We were also able to demonstrate that this case of convergence involves spiders with vastly divergent predatory behaviours.


This study was in part supported by grants from National University of Singapore ARC to D. Li (R-154-000-199-112 and R-000-234-112) and grants from the Royal Society of New Zealand (Marsden Fund and Kames Cook Fellowship) to R. R. Jackson. Valuable and generous assistance was provided by ICIPE, Kenya and Xishuangbanna Tropical Botanic Garden (XTBG), Chinese Academy of Sciences in China. Special thanks are extended to H. M. Liu, Y. P. She, L. M. Li, J. Zhang, J. H. Mao and Z. H. Li from XTBG for generous assistance. Special thanks also go M. S. Zhu and M. B. Gu for assistance with the expedition to Hainan in December 2003 and for assistance with fieldwork in Hainan, China. We also wish to thank S. Q. Li for assistance with export permits from China. Import permits were provided by the CITES, Agri-food & Veterinary Authority of Singapore and J. X. Zhang helped collecting and identifying spiders. We further like to extend our thanks to the people of the Spider Lab and Evolutionary Biology Lab (DBS, NUS) for all their help and support through this study, in particular Sujatha N. K., G. Vaidya, M. L. M. Lim and J. R. W. Woon. We are particularly grateful to Gaurav Girish Vaidya and the National Grid Singapore for help and support with the computational work.