Because of their fragility, the fossil record of spiders is controlled by the occurrence of Konservat-Lagerstätten (i.e., exceptionally well-preserved fossil deposits, Seilacher, 1970). One well-known Lagerstätte is amber, the highly polymerized form of fossil tree resin (Fig. 1A). The majority (over 90%) of fossil spiders discovered to date derive from amber deposits of Cenozoic age (Selden et al., 2009), with Baltic and Dominican ambers predominating (Fig. 2). The oldest spiders found in amber are from the Cretaceous (Dunlop et al., 2010). More rarely, spiders are preserved in sedimentary rock strata, and there are nearly as many different preservational styles as there are araniferous Lagerstätten (Fig. 1B; Selden and Penney, 2010 for an extensive review). Sedimentary deposits preserving the oldest record of spiders date back to the Carboniferous (Selden et al., 2009).

Although the term Lagerstätte conjures up images of exceptionally preserved fossils, fine-scale anatomical details are often not visible. Identifying spider species is therefore frequently difficult due to preservational constraints. Even in amber, which preserves specimens with the highest fidelity, the genitalia may be distorted, obscured or absent, and this has interesting implications for species definition. Genitalia are even less likely to be preserved in sedimentary deposits because the three-dimensional palps and epigynes are reduced to two-dimensions. Preservational infidelity is potentially more troublesome when dealing with members of the monophyletic Entelegynae, characterized by more complex reproductive systems, compared to those specimens belonging to the Haplogynae (Coddington, 2005, Coddington and Levi, 1991, Foelix, 1996, Griswold et al., 1999, Hausdorf, 1999 and Platnick et al., 1991). Female entelegynes have epigynes with external copulatory openings, and males usually possess palps with two or three divisions in the bulb (e.g., subtegulum and tegulum). The parts of the entelegyne bulb are typically connected by thin membranous tissue that expands during copulation (Coddington, 2005). This delicate membrane may be distorted during fossilization, thus complicating interpretation and determination of homologous structures in extant and fossil palps. Generally, the relatively simple, pyriform bulb of a haplogyne is more likely to be fully preserved than the complicated and intricate details of an entelegyne palp.

Other anatomical features, including trichobothria, leg spination, eye pattern, and coloration may also be tricky, if not impossible, to observe in fossil specimens, meaning that species identification must proceed in a different manner than when dealing with extant taxa. Although confined to available morphological characters, we do not imply that palaeo-species differ from extant species; as mentioned previously, palaeontologists simply must employ (potentially) different epistemological means of species identification. For instance, molecular methods cannot be drawn on to delimit fossil species, although this does not differ from the majority of modern systematic studies still firmly rooted in morphology. Using molecules to parse out inter- and intraspecific relationships has been utilized more frequently within spiders (e.g. Bond and Stockman, 2008, Duncan et al., 2010 and Hedin, 1997), but claims that DNA barcoding can replace traditional taxonomy and identify species (e.g. Barrett and Hebert, 2005) has rightly met with resistance (Prendini, 2005, Scotland et al., 2003 and Seberg et al., 2003). As Prendini (2005) (p. 502) noted: “DNA-based methods are not demonstrably more objective, accurate, or useful than morphology or other sources of phenotypic data for species identification or other taxonomic purposes.”

We have discussed how genitalia are used as species-specific characters and can potentially reveal SMRSs within the Araneae, but does this imply that somatic characters cannot and should not be used in species identification? In other words, are these characters epistemologically useless for species delimitation? A review of the spider systematic papers from the three most recent issues of the Journal of Arachnology illustrates that somatic characters are employed as diagnostic criteria for extant species in nine out of 14 papers (over 60%). This is in accord with Huber's (2004) third tenet, where such characters are still considered relevant diagnostic criteria. We assert that even though copulatory organs may be more epistemologically meaningful for distinguishing species, no a priori evidence exists to discount somatic characters as species-specific. Further tests, such as molecular and field analyses, can be performed on a case-by-case basis in the modern to examine if particular somatic differences emerge at the species-level or if they are polymorphisms, a task difficult to perform in the fossil record. Huber (2004) noted that individuals with identical copulatory organs but with discontinuities in other characters are usually interpreted as polymorphic or polytypic species, yet this practice is perhaps dubious. Consequently, somatic characters should be treated as potentially distinguishing features of a species, given no other a priori evidence to the contrary. There is always a risk that epistemology may not equate to the ontology of a species. In other words, we may over-split, creating many species when only one exists, or we may lump many species into a single taxon, but as long as diagnoses are explicit, new information and further research can continually test these hypotheses.

The proceeding discussion on preservational media hints at the difficulty of comparing the rock and amber records of spiders, both to each other and to living species, because each recovers a potentially different (sub)set of characters. The question is then raised of how to deal with similar looking fossils found under different preservational conditions. We posit that if there are no obvious differences between the fossils, the specimens should be placed at the appropriate taxonomic level for which the available diagnostic characters allow. No designation should be made for depauperate specimens lacking synapomorphies and/or distinguishing morphological data. The possibility exists that specimens that appear different due to preservation will be named as different species, when in reality they are synonymous (NB: due to the scarce nature of the fossil spider record, this is unlikely to be a major problem). Again, incorrectly diagnosing species is not the exclusive problem of palaeontology, and if there is no a priori evidence to suggest the specimens belong to the same species, differences should be treated as such, or no designation should be made if relevant and diagnosable character complexes are not discernible.

Occasionally, specimens are discovered that do not easily fit into extant classification systems (i.e., family, genera), but which are clearly related to modern lineages (i.e., so-called plesion or stem lineages). When this occurs, how should these specimens be treated so that classifications are in accord with hypotheses of evolutionary history and poly- and/or paraphyletic groups are avoided? Should modern genera be diagnosed more broadly to accommodate closely related fossil forms, or should the fossil (plesion) taxa be maintained separately? We advocate the methodology of Wiley (1979) and Patterson and Rosen (1977), whereby the fossil specimens are placed in the groups that they are most closely related to (i.e., plesiomorphic to), but the corresponding taxon name is placed in shutter quotes, denoting the group's potential paraphyly. This maintains modern classifications without uprooting traditional diagnoses, but also allows for hypotheses of relatedness within a phylogenetic framework. For example, Saupe and Selden (2009) described a species most closely related to the spider family Mecysmaucheniidae, but the specimen lacked the complete set of diagnostic characters of the modern group (e.g., the fossil possessed four rather than two spinnerets). In this case, the specimen could be diagnosed as “Mecysmaucheniidae”, rather than placed as a separate stem-lineage family or necessitating a re-diagnosis of the extant group.

Since spiders are relatively scarce in the fossil record, species descriptions are primarily based on singletons or only a few specimens (rather than a series of similar specimens), and only one specimen may represent the fossil record of entire families/lineages. Of course, this makes it impossible to assess levels of within-species variation. Reliance on Lagerstätten also means that range extensions and ghost lineages are quite common within the Araneae. These terms refer to the history of taxa with no direct record, as elucidated within a phylogenetic framework (Edgecombe, 1992, Norell, 1992 and Norell and Novacek, 1992). For example, the presence of Cretaceous species within the family Oonopidae suggests the Orsolobidae (putative sister to the oonopids sensu Forster and Platnick, 1985 and Platnick et al., 1991) were also present during this time, although they have yet to be found as fossils (Penney, 2002a).

Unfortunately, an increasing number of Mesozoic specimens have been assigned to extant genera (e.g., Araneus, Gnaphos and Theridion in Chang, 2004 and Pisaura in Kim and Nam, 2008), with poor descriptions and inadequate illustrations (Fig. 1C). These reports are problematic because they may skew our view of spider evolution by incorrectly dragging back records of genera or families and by providing erroneous calibration points for phylogenetic studies. Diagnostic criteria should be met for taxonomic placement in existing groups, regardless of the status of the specimen being studied (i.e., fossil or extant); if the appropriate diagnostic characters are not discernible, no designation should be made.

Typically, the amount of accessible morphological data, particularly with respect to copulatory organs, differs for male and female specimens in the fossil record. For instance, many of the female genital structures used in species identification, such as coiled connecting ducts, seminal receptacles and fertilization ducts of the epigyne, are internal and usually not visible in specimens preserved in amber or rock (for an exception see Selden, 2010). The greater accessibility of male copulatory organs (i.e., the potential SMRS) may explain why the majority of described fossil species are based on males (∼95%, D. Penney, pers. comm.). Another, perhaps complementary explanation, however, is that males were preserved more often than females due to their wandering nature (Penney, 2002b and Penney and Langan, 2006) and thus are simply more abundant for study. The lack of information gleaned from females is, nevertheless, a hindrance and leaves many specimens unstudied. Similarly, it is often difficult to assign males and females to the same species. This is also true for the modern fauna (although neontologists can make field observations and study coloration patterns and internal genitalia). For example, a review of three spider families (chosen at random) revealed that original diagnoses were performed using both the male and female for only 39, 28, and 59% of the extant species contained within the Desidae, Selenopidae, and Cybaeidae, respectively (data from Platnick, 2010).

Fossil juvenile spiders are also hard (usually impossible) to identify to the species or even family level, and distinguishing between males and females in the absence of mature sex organs is often difficult. Again, these problems are not restricted to the fossil record but, in the modern, context clues and large sample sizes may aid in identification. Fossil juveniles are generally not studied, which leaves much diversity unexamined. Immature specimens are not entirely uninformative, however, as indicated by the Cretaceous fossil spider family Lagonomegopidae, which is described entirely from juvenile (and/or female) specimens (Penney, 2005, Penney, 2006 and Penney and Selden, 2005). This family is diagnosed as possessing cheliceral peg teeth and large eyes situated anterolaterally (Penney and Selden, 2005). More information on the group would be gleaned if adults were found, including making it possible to better evaluate its relationship with other spider lineages. Other strictly fossil spider families based on descriptions of juveniles have not withstood contemporary taxonomic scrutiny (Penney and Selden, 2005), and, in general, it is usually good practice to only name species based on adult specimens.

We do not attempt to address and provide solutions for every issue involved in the study of fossil spider species, nor is this likely possible. We endeavoured to provide an overview of the most relevant concepts and epistemological problems encountered. Although our preceding discussion does recognize that there are some difficulties involved with identifying and delineating fossil spider species, it is undoubtedly true that the fossil record of spiders has contributed in an important way to the understanding of the group's evolution (Penney, 2004 and Penney et al., 2003). Further, there are a plethora of instances where genitalia and other anatomical features are preserved in exquisite detail. For example, a member of the Plectreuridae was described from the Jurassic of Daohugou, China based on fine details of the palpal bulb and embolus and the presence of a macroseta on the tibia of the first leg (Fig. 1B; Selden and Huang, 2010). In fact, the morphology of this 165 million-year-old specimen is incredibly similar to modern species within the tristis group of the genus Plectreurys. Many amber specimens have palpal details preserved (Harms and Dunlop, 2009, Marusik and Wunderlich, 2008 and Penney and Ortuño, 2006, to name only a few) and can even be placed in extant genera based on these structures (e.g. Huber and Wunderlich, 2006, Penney, 2000, Penney, 2009 and Saupe et al., 2010).

Studies comparing modern and fossil forms are becoming increasingly common (for example, Harms and Dunlop, 2009 and Saupe et al., 2010), and we emphasize the benefits of doing so, particularly when carried out in a phylogenetic framework. Harms and Dunlop (2009) revised fossil mimetids and synonymized the extinct genera Succinero and Palaeoero with the extant genus Ero. Side-by-side examination of fossil and extant species allows for determination of changes in morphology through time (e.g., Eurasian species of Ero possess palps with conical lateral lobes between the straight cymbium in lateral view and the paracymbium, whereas the fossil forms do not) and a better understanding of biogeographic and evolutionary history (Fig. 1D).