A completely resolved phylogenetic tree of British spiders

The recent accumulation of increasingly densely sampled phylogenetic analyses of spiders has greatly advanced our understanding of evolutionary relationships within this group. Here, this diverse literature is reviewed and combined with earlier morphological analyses in an attempt to reconstruct the first fully resolved phylogeny for the spider fauna of the British Isles. The resulting tree highlights parts of the group where data are still too limited for a confident assessment of relationships, proposes a number of deviations from previously suggested phylogenetic hypotheses, and can serve as a framework for evolutionary and ecological interpretations of the biology of British spiders, as well as a starting point for future studies on a larger geographical scale.


11
In recent years, the number of large-scale phylogenetic analyses of spiders has been rapidly 12 increasing. Molecular studies, as well as classical morphological work and integrative "whole-13 evidence" analyses, are covering large parts of spider diversity with increasing density. The 14 last years have seen the publication of several comprehensive phylogenetic studies of the entire 15 order, based on continuously increasing species coverage and ever-larger amounts of (mostly 16 molecular) data (e.g., Agnarsson et al. 2013;Bond et al. 2014;Dimitrov et al. 2017;Fernandez 17 et al. 2014Fernandez 17 et al. , 2018Garrison et al. 2016;Hedin et al. 2019;Kulkarni et al. 2020;Opatova et 18 al. 2020;Ramírez 2014;Ramírez et al. 2019Ramírez et al. , 2021Shao & Li 2018;Wheeler et al. 2017). 19 Subsets of the order, from superfamilies to individual groups of genera, have also been the target 20 of various analyses (e.g., Crews et al. 2020;Godwin et al. 2018;Hedin et al. 2018;Kallal et 21 al. 2020; and numerous publications cited below for individual families). In addition, "DNA 22 barcode" projects have provided a plethora of molecular genetic data for a wide range of spider 23 species (e.g., Astrin et al. 2016;Blagoev et al. 2016), which can serve to complement earlier 24 morphological analyses in an attempt to resolve phylogenetic relationships, especially within 25 spider genera (Breitling 2019b,d). 26 It would be interesting to see how all these studies in combination can inform our understand-27 ing of the evolutionary relationships within a local spider fauna. Most ecological and faunistic 28 work on spiders is done at a local level and would benefit from a clear and explicit evolutionary 29 framework, i.e., a phylogenetic tree of the local spider fauna. While such a tree would be pioneer-30 ing for arachnology, megatrees for local floras and faunas have been successfully constructed for 31 many other groups, from European tetrapods (Roquet et al. 2014) and butterflies (Wiemers et 32 al. 2020), to the vascular plants of the British Isles, Germany, the Netherlands, and Switzerland 33 1/29 (Durka & Michalski 2012), as well as for British birds (Thomas 2008). They are considered 34 an essential ingredient for evolutionarily informed studies in ecology and conservation science 35 (Roquet et al. 2013), e.g., providing the necessary phylogenetic framework for understanding 36 the evolution of egg shell pigmentation in British birds (Brulez et al. 2016), or for identifying 37 the effect of species' traits on their population changes (Sullivan et al. 2015). 38 The general benefits of a phylogenetic framework are obvious: just imagine what ecological 39 or faunistic studies would look like if all phylogenetic information were discarded: families 40 and genera could not longer be used to structure the information. The ubiquitous pie charts 41 that present survey results according to spider family would disappear. It would no longer be 42 possible to state that the relative abundance and diversity of linyphiids increases towards higher 43 latitudes, or that lycosids dominate in a pitfall sample. The resulting challenges would clearly be 44 wide-ranging. 45 A fully resolved phylogenetic tree provides the same kind of information as the family and 46 genus assignments, but with much finer granularity and without being restricted to arbitrary 47 taxonomic categories. Ecological and faunistic studies can ask the same kind of questions about 48 each of the clades in the tree that they would routinely ask about families or genera. This is a 49 rich opportunity -and, importantly, the usefulness of a tree for this kind of analysis would not 50 increase if it included a global set of species, but it would decrease if any local representatives 51 were missing or if it were less than fully resolved. 52 Moreover, applying the phylogenetic results at a local level should help identifying gaps in 53 our current understanding of the spider Tree of Life: is it even possible to plausibly reconstruct 54 all the evolutionary relationships within a local spider fauna given the currently available data? 55 Here, I attempt to answer this question for one particularly well-studied spider fauna, that of the 56 British Isles. Great Britain and Ireland have a long and distinguished history of arachnological 57 research. The fact that much of their fauna and flora was acquired after the glaciations of the 58 ice ages has resulted in a relatively impoverished, but still interestingly diverse, spider fauna, 59 which largely consists of species that are common and widespread across the Palaearctic. As a 60 result, the vast majority of British spiders has been studied, illustrated and described repeatedly 61 in great detail, providing an excellent starting point for determining their relationships. Many of 62 the British species or their close relatives have been included in published phylogenetic studies, 63 and a considerable fraction has been the target of comprehensive DNA barcoding projects. The 64 systematics of the British spiders is reasonably stable, and their nomenclature and classification 65 provide another, implicit source of information about the phylogenetic relationships within the 66 group.

67
But isn't it somewhat pointless to reconstruct a phylogenetic tree for such a small, stochastic 68 subset of global biodiversity? Certainly not. For a start, all published phylogenetic hypotheses 69 refer to limited subsets of global spider richness; a complete fully resolved tree of all 40,000+ 70 spider species is a distant dream -and if it were constructed today, it would still lack all the 71 undiscovered and most of the extinct species that are equally valid members of the great Tree of 72 Life. Compared to the typical trees shown in the arachnological literature, the tree presented 73 here is exceptionally large and densely sampled. But, more importantly, for the reasons listed 74 above, the British spider fauna is far from a random subset of spider diversity: as a result of 75 ecological history, it is strongly enriched in abundant, generalist and widespread species (and 76 this still holds true for its recent arrivals). And, as a result of arachnological history, it is also 77 strongly enriched in well-studied, carefully described and comprehensively analysed species, all 78 of them repeatedly revised and illustrated and many of them barcode-sequenced; it is furthermore 79 strongly enriched in type species of globally distributed genera, and type genera of many globally 80 important families. Consequently, the resulting phylogenetic tree contains just the kind of species 81 that will make it useful as a point of reference for expanding the tree, first to the Palaearctic fauna, 82 and in the future to phylogenies on a global scale (even if these will more likely be generated 83 from scratch -the synthesis of the literature provided here should give a first idea of the work 84 2/29 still required to achieve this ultimate ambition). 85 Yet, a concern could remain that the geographically defined scope of the tree is inappropriate: 86 after all, phylogeny does not respect geography, spiders do not know national borders, and 87 the British spider fauna has certainly not evolved in situ, but is merely a part of global spider 88 biodiversity, closely related to that of the nearby continent. However, this is also true for the 89 many other taxa where local phylogenetic trees have proven to be of tremendous value. I would, 90 moreover, argue that nobody doubts the value of checklists of the local spider fauna; for British 91 spiders, such checklists are published regularly, with geographic scales ranging from individual 92 nature reserves, to single counties and the entire country. Knowing which species are present 93 in a local area provides the necessary context for any observations on individual groups on the 94 list. It would be considered bizarre if such a list were sorted alphabetically by species epithet: at 95 the bare minimum, the species will be grouped according to genus and family. In the case of 96 British spiders, there has in fact been a longstanding tradition to arrange checklists even further 97 by phylogenetic affinities, reflecting presumed evolutionary relationships in the order of species 98 within genera, genera within families, and families within Araneae in general. A fully resolved 99 phylogenetic tree takes this idea only a small and natural step further, making the proposed 100 relationships explicit, unambiguous and testable.

101
As a purely intellectual exercise a tree like this meets an elementary scientific desire for 102 taxonomic order based on evolutionary relationships and with the maximum possible resolution. 103 But it also fulfils an important role as a didactic and mnemonic aid: it is much easier to learn and 104 remember the characters and features of the members of a local spider fauna, if one is aware of 105 their precise evolutionary relationships. In contrast to a determination key or other pragmatic 106 arrangements of the species, a phylogenetically-informed "mind map" (i.e., a phylogenetic tree) 107 is ready to grow by adding new species whenever needed, for instance because the taxonomic or 108 geographic scope is expanded.

109
Of course, not every arachnologist will be interested in the detailed evolutionary relationships 110 of their study animals; some may be satisfied with the coarse-grained picture provided by the 111 Linnaean taxonomic hierarchy. But, the majority of curious naturalists will want to know how 112 the species in their local patch are related to one another; initially, this interest will be quite 113 independent of any relationships to species in other parts of the world. The tree identifies 114 specifically the local sister group of each species or group of species, without being restricted 115 by the arbitrary levels of the Linnaean hierarchy. It allows examining whether ecological, 116 morphological or behavioural traits studied in a British spider community correlate with their 117 patterns of evolutionary relatedness. But its usefulness doesn't end there; even if this local tree 118 only represents a subset of the global spider biodiversity, it would be applicable to studies across 119 much of Northwest Europe, with little or no modification, and could easily be expanded to the 120 spider fauna of large parts of Central Europe.

121
The central data source of this paper is a comprehensive synthesis of the published literature. 122 As a result, the amount of original data will be limited, and a conscious effort was made 123 to minimise the reliance on novel unpublished pieces of evidence. This turned out to be 124 far from trivial: the emerging phylogenetic literature is surprisingly challenging to apply on 125 a local level. Even the largest studies will include only a tiny fraction of the entire spider 126 diversity; individual studies only partly overlap in the species included; and while there is a 127 trend towards consolidation in some areas of the phylogenetic trees, there remain numerous parts 128 where different studies arrive at widely different and inconsistent results, which are difficult to 129 reconcile. Below, these challenges are discussed for selected individual examples, emphasising 130 the necessarily subjective nature of some of the decisions made, in view of the still incomplete 131 evidence available.

133
The phylogenetic reconstruction presented here is based on a synthesis of a wide range of 134 published phylogenetic analyses, supplemented by the data in the taxonomic literature on British 135 spiders, as well as additional barcode data from public databases. Given the heterogeneity of 136 the available information, no rigid, formal approach could be expected to resolve the remaining 137 conflicts and uncertainties convincingly. The construction of the preferred tree consistent with 138 the cited data was done manually, and, in general, the results of studies including a denser sample 139 of species in the relevant part of the tree were preferred over those with a sparser coverage; multi-140 gene molecular studies were preferred over purely morphological analyses; integrated studies, 141 combining molecular and morphological data, over those using only one data type; studies with a 142 larger number and diversity of molecular markers were preferred over those studying fewer genes; 143 and barcode data from the literature and the BOLD database (Ratnasingam & Hebert 2007) were 144 only used when they were not directly contradicted by the morphological evidence. In addition to 145 recently published phylogenetic analyses, the full range of phylogenetic hypotheses implicitly or 146 explicitly proposed in the traditional taxonomic literature was considered, as were all published 147 morphological data, although no formal cladistic analysis of such data was attempted.

148
Morphological data for linyphiid spiders were obtained from Anna Stäubli's identification 149 key at the Spiders of Europe website (Stäubli 2020; https://araneae.nmbe.ch/matrixlinkey). 150 The character states for Nothophantes horridus were added on the basis of Merrett & Stevens 151 (1995, 1999, and those for the male of Pseudomaro aenigmaticus on the basis of unpublished 152 observations by A. Grabolle. Data on the distribution of British spiders at the hectad level 153 were downloaded from the website of the Spider and Harvestman Recording Scheme website 154 (http://srs.britishspiders.org.uk/portal.php) and mapped onto the phylogenetic tree. Subsequently, 155 for each of the subtrees, a Wilcoxon rank sum test was applied to test for significant differences 156 in the value of ecological variables of interest for the species within the subtree (clade), compared 157 to the rest of the species. Uncorrected p-values are reported, but these are all significant at the 158 0.05 level after Bonferroni correction for multiple testing based on the number of subtrees. Trees 159 were visualised using the iTOL web tool (https://itol.embl.de/ ; Letunic & Bork 2019). Data 160 were processed using custom-made R scripts, which are available from the author upon request. 161 In this article, the focus is on proposing a single fully resolved tree, expressing a single testable 162 phylogenetic hypothesis for each triplet of species in the tree. Obviously, as will be mentioned 163 repeatedly in the following discussion, not all of these hypotheses will be proposed with equal 164 confidence. In many cases, alternative proposals would seem equally plausible. By unequivocally 165 identifying one preferred hypothesis in each case, the proposed tree might be considered overly 166 audacious. The advantage of this approach is that it facilitates future discussion and provides an 167 unambiguous point of reference for necessary improvements and corrections.

169
The list of British spiders was obtained from the Spiders of Europe website (araneae.nmbe.ch; 170 Nentwig et al. 2020) in May 2020. The list uses the World Spider Catalogue (WSC) nomenclature, 171 which should be referred to for the currently accepted names of the species involved, as well as 172 the authorities and additional taxonomic references for each of them. However, in the tree itself, 173 the nomenclature has been adjusted to ensure that all genera are monophyletic, according to the 174 proposed phylogenetic hypotheses. This renaming, which applies to a small minority of cases, 175 largely follows the results of Breitling (2019a,b,c,d,e), with a few additional new combinations 176 based on more recent published analyses.

177
Subgenera. -In a few cases, subgenera are explicitly proposed for traditional monophyletic 178 4/29 groups within larger genera. Subgenera are currently rarely used in arachnology. This was 179 not always the case and seems to be largely a historical consequence of the way the most 180 widely used spider catalogues are organised; e.g., the WSC almost always ignores subgenus 181 divisions, leading to justified concerns that re-classification proposals below the genus level 182 might easily be overlooked. However, the use of subgenera increases the information content 183 of a classification considerably (Wallach et al. 2009), and it avoids the instability caused by 184 subdividing homogeneous and obviously monophyletic groups into smaller and smaller genera. 185 Many of the most acrimonious taxonomic debates of recent years could probably have been 186 avoided if subgenera had been more widely adopted as a useful category in spider systematics. 187 Semispecies. --Barcoding data have regularly revealed that closely related spider species 188 share mitochondrial DNA haplotypes (Astrin et al. 2016;Blagoev et al. 2016;Domènech et 189 al. 2020;Ivanov et al. 2018;Lasut et al. 2015;Nadolny et al. 2016;Oxford & Bolzern 2018). 190 This has often been dismissed as the result of a "barcode failure", potentially due to incomplete 191 lineage sorting of recently diverged species. However, this explanation is clearly untenable for 192 the majority of cases in spiders: almost always, the sibling species involved are highly mobile 193 and very common ecological generalists that are sympatric and regularly syntopic over huge 194 areas, often entire continents. Assuming a traditional allopatric model of speciation (Kraus 195 2000), this argues strongly against a recent divergence between the species.

196
Alternatively, it has been suggested that the lack of an interspecific barcode gap is the result of 197 mitochondrial introgression, potentially facilitated by Wolbachia-mediated gene drives. This may 198 well be the case; however, for such a scenario to be as common as it appears in spiders requires 199 regular fertile hybridization between the species involved, with minimal negative fitness costs 200 for the hybrid individuals. It has been suggested that such introgression would have negligible 201 effects on the stability of the species boundaries, as the nuclear genetic contribution of one of the 202 parents would be rapidly lost, only the maternally inherited mitochondrial genome remaining as a 203 trace of the hybridisation event (Oxford 2019). This is, however, not correct: in fact, assuming a 204 stable population size, it is the mitochondrial contribution that will be rapidly lost stochastically, 205 but due to recombination events, at least some of the alleles from both parents will remain present 206 in the nuclear genome of backcrossing generations for a long time. Some studies that have 207 examined examples of a missing barcode gap in more detail have indeed found that at least some 208 nuclear markers equally fail to differentiate the species involved (Lasut et al. 2015;Spasojevic et 209 al. 2016). The exchange of nuclear alleles is not necessarily obvious in the form of a gradient of 210 phenotypically intermediate individuals, if backcrossed hybrids rapidly approach the phenotype 211 of one of the parents (Oxford 2019).

212
The relevance of this phenomenon for the present analysis is that species that exchange 213 mitochondrial DNA barcodes are likely to also exchange favourable (or neutral) nuclear alleles 214 regularly (examples are known from crop pests exchanging pesticide resistance genes, but also 215 from mimicry complexes in butterflies; e.g., Valencia-Montoya et al. 2020 andZhang et al. 216 2016). Such groups do not form independent evolutionary individuals yet -and they are not 217 mutually monophyletic -, and consequently it would be meaningless to propose a fully resolved 218 phylogenetic tree for them. This is only of minor import for species pairs, but in several of these 219 cases, especially among wolf spiders, groups of three or more species are involved globally.

220
In the tree presented here, I propose to apply the semispecies concept to such cases. Semis-221 pecies are groups of organisms in the "grey zone" of speciation, which have completed some of 222 the necessary steps towards full species separation, but not all of them. Semispecies have often 223 been used to classify island populations (as a synonym of allospecies), and the rare examples in 224 arachnology also applied the concept strictly to allopatric populations (Kraus 2000); however, 225 such a narrow interpretation is not necessary, and in groups with a more mature taxonomy 226 (especially birds and butterflies) locally sympatric (as well as parapatric) groups of semispecies 227 are identified with some regularity (e.g., Helbig et al. 2002;Smith et al. 2010), albeit not without 228 lively debate regarding individual cases (e.g., Pfander 2011; Vane-Wright 2020).

5/29
Importantly, semispecies are not by definition species in statu nascendi; their evolutionary 230 future cannot be predicted with certainty. In some cases, it might be that disruptive selection, 231 reinforcement and related mechanisms will allow them to progress to full speciation in sympatry. 232 But it is equally possible that they will continue to maintain gene flow indefinitely or even merge 233 into a single homogeneous freely panmictic population. Interestingly, among members of the 234 Eratigena atrica species group, individuals in the overlapping part of the range show a tendency 235 towards intermediate phenotypes (Oxford 2019), in contrast to the expectation of "character 236 displacement", which would predict that sympatric populations would become more clearly 237 distinct phenotypically, e.g., to minimise harmful hybridisation or ecological niche overlap.

238
In the phylogenetic tree, semispecies are explicitly identified by including a "superspecies" 239 name, i.e., the name of the first member of the group to be described, in brackets before the 240 species epithet. This highlights areas of the tree where speciation may not yet be complete, and 241 where the question "one species or two (or three)?" is ill posed. Studies such as the work of 242 Ivanov et al. (2018), Domènech et al. (2020), and Oxford (2019 show that each of these cases 243 will be a rich ground for future evolutionary insights.

245
The deep phylogenetic relationships of spider families have recently been addressed by a series 246 of molecular genetic studies (Agnarsson et al. 2013;Bond et al. 2014;Dimitrov et al. 2017;247 Fernandez et al. 2014247 Fernandez et al. , 2018Garrison et al. 2016;Opatova et al. 2020;Ramírez 2014;Ramírez 248 et al. 2019Ramírez 248 et al. , 2021Shao & Li 2018;Wheeler et al. 2017). These show considerable agreement 249 with previous morphological analyses (e.g., the trees shown in Coddington 2005 or Jocqué 250 & Dippenaar-Schoeman 2006), and where they disagree (e.g., regarding the non-monophyly 251 of orb-web weavers or the placement of Mimetidae among Araneoidea), the strong statistical 252 support for the alternative hypotheses and the agreement of independent molecular analyses 253 indicate that the latter recover the true evolutionary relationships. Overall, the last few years 254 have seen rapid convergence towards a stable consensus, although a few details, such as the sister 255 group of Salticidae, remain somewhat contentious. In addition to the British species, the tree in 256 Figure 1 includes all families known to be present in Europe, to provide a broader context for 257 the phylogenetic position of the British fauna.

258
While establishing the family-level framework is by now relatively easy, reconstructing the 259 phylogenetic relationships within each family required the reconciliation of a diverse range of 260 published proposals and personal judgement calls, on a case-by-case basis. Once a monophyletic 261 group of N species has been established, N − 2 decisions (justifying N − 2 branch points) are 262 required to reconstruct a fully resolved phylogenetic tree. For the more than 600 British species in 263 the present tree, obviously, not all of these decisions can be elaborated in detail here. Importantly, 264 the evidence supporting each decision (including published cladograms, character matrices, 265 illustrations of the genitalia, and a variety of DNA sequences) could not be presented in the text; 266 doing so would require reproducing a large fraction of the taxonomic literature on British and 267 European spiders. Instead, the reader is referred to the data in the cited literature. However, 268 selected examples from each family with more than 2 members are discussed in alphabetical 269 order below. The tree for this family is based on the detailed study by Bolzern et al. (2013), combining 272 morphological and molecular data for a dense sample of species. The deep relationships of 273 the genera show major differences compared to the analysis of Wheeler et al. (2017), which 274 6/29 included individual representatives of the same genera. Crews et al. (2020) include a larger 275 number of species, and recover a tree that is similar to that reported by Wheeler et al. (2017), but 276 seems overall poorly resolved. These disagreements illustrate how fragile some of the results of 277 even the most recent molecular analyses still can be. The arrangement proposed by Bolzern et 278 al. (2013) is preferred, as the molecular results in this case agree quite closely with those of a 279 morphological analysis, while in the trees presented by Wheeler et al. (2017) and Crews et al. 280 (2020) in particular the placement of Coelotes deeply within the remaining Agelenidae s. str. is 281 unexpected (the genus is typically placed in the subfamily Coelotinae, which is sometimes even 282 considered a separate family Coelotidae). Agelena longipes is a phantom species as defined by 283 Breitling et al. (2015Breitling et al. ( , 2016, i.e., it was not rediscovered since its original description in 1900. It 284 is thus considered a nomen dubium and is not included in the tree.

285
Amaurobiidae 286 The topology of the tree within this small group is determined by the morphological affinities, 287 and confirmed by available barcode data for all three species.

289
The relations in this case are inferred on the basis of the similarities in pedipalp and epigynum of 290 Anyphaena accentuata and A. sabina, as well as the rather close similarity the barcodes for these 291 two species, which indicate that they are probably are pair of sister species, while A. numida is 292 more distantly related. ; for other genera the relationships are informed by barcode data, which are available 300 for all species. The placement of Zilla is only very weakly supported; it is based on Tanikawa's 301 assessment that this genus is closely related to Plebs/Eriophora (Tanikawa 2000), which in the 302 tree of Scharff et al. are closely related to Singa.

303
The genus Araneus in the traditional sense is clearly polyphyletic. The placement of Araneus 304 (s.str.) angulatus is based on barcode similarity to A. bicentenarius, which is placed apart from A. 305 diadematus and its relatives in Scharff et al. 2020. The close relationship between A. angulatus 306 and A. bicentenarius is confirmed by their morphological similarity, which led earlier authors to 307 consider A. bicentenarius a variant of A. angulatus (Levi 1971). The available genus name for 308 the A. diadematus group would be Epeira, and this is used in the tree.

309
The placement of the Atea species is unclear. Their barcode sequences indicate that they are 310 not closely related to A. angulatus or the A. diadematus group of the highly polyphyletic Araneus 311 s. lat.; as the genus Atea has historically often been placed close to Agalenatea, the two British 312 species are placed there, but with some reluctance, as there seems to be no convincing evidence 313 for a close relationship between the two genera. Relationships within this small group remain obscure. Wolf (1991) discusses Cheiracanthium 316 pennyi and C. erraticum as being more similar to each other than C. erraticum and C. virescens, 317 contra Helsdingen (1979). Barcodes for C. pennyi are not available, so the molecular data cannot 318 help resolving the case. The pedipalp without ridge and the epigynum with short insemination 319 ducts set C. pennyi apart, but these could be autapomorphic. The striking red dorsal line of the 320 opisthosoma is shared by C. pennyi and C. erraticum, but here the loss in C. virescens could be 321 autapomorphic. The preferred topology represented in the tree gives precedence to the genital 322 similarities, but without a sound cladistic basis for this decision.

324
The backbone of the topology within this family largely follows Breitling (2019c). Microclu-325 biona sensu Wunderlich (2011), i.e., the trivialis group sensu Mikhailov (1995) is considered 326 monophyletic (as in some of the barcode analysis results). Euryclubiona sensu Wunderlich 327 (2011) is also considered monophyletic, and is recovered as such in most of the barcode results; 328 the internal topology within this group is based on morphological affinities. For the reasons 329 discussed in Breitling (2019d), none of these subgenera is formally recognised here, but a future 330 subdivision of Clubiona into well-defined subgenera would certainly be highly preferable over a 331 splitting of this homogeneous and obviously monophyletic group into separate genera.

332
The position of C. rosserae as sister of C. stagnatilis is based on the original description 333 and Wiehle (1965), while the position of C. subtilis and C. juvenis is based on morphological 334 similarities to their proposed sister species. C. corticalis is placed basally within the genus, in 335 agreement with its unique morphology and most barcode analyses.

336
C. facilis, which is also listed as a member of the British spider fauna is possibly a synonym of 337 C. phragmitis, based on a malformed specimen, analogous to the case of Philodromus depriesteri 338 discussed in Breitling et al. (2015). The close morphological similarity to C. phragmitis 339 discussed in the original description (sub C. holosericea), as well as the "atrophic" appearance 340 of the epigynum in the accompanying illustration, would seem to support this interpretation. 341 Examination of the type material in the Pickard-Cambridge collection in the Oxford University 342 Museum of Natural History (Bottle 2312.1) could provide further insights, but for now the name 343 is considered a nomen dubium and is not included in the tree.  2020), Dictyna was recovered as 351 polyphyletic, but here it is assumed that the British representatives form a monophyletic group, 352 as do the other genera in these two families.

353
The placement of Altella is questionable. Wunderlich (1995aWunderlich ( , 2004a considers the genus a 354 junior synonym of Argenna, indicating at least a close relationship between the two genera. But 355 in the same works he also considers Brigittea (and Emblyna) as synonyms of Dictyna; for this to 356 be correct, it would be necessary to considerably expand the scope of Dictyna, to also include 357 the quite distinct genus Nigma.

8/29
Given the close morphological similarities (and barcode similarities) between Argenna, Altella 359 and Dictyna s.lat., Argyroneta is placed basal to a clade uniting these two groups, despite results 360 by Crews et al. (2020) that argue for joining Argenna and Argyroneta instead.

361
The internal topology within Nigma is supported by morphological affinity that indicates 362 a sister group relationship between N. puella and N. flavescens. The same argument applies, 363 with less conviction, in the case of Lathys. The topology within Dictyna s. str. is based on a 364 combination of morphological similarities and barcode data for a subset of the species.

365
Dysderidae 366 Platania et al. (2020) show that Harpactea in its present form is a polyphyletic assembly 367 of unrelated groups, the two British species being placed in deeply separated clades. Non-368 monophyly is also found for several other genera of Harpacteinae (Folkia, Dasumia), as well as 369 for some of the previously proposed species groups within Harpactea. This indicates that the 370 morphological recognition of monophyletic units within this family is unusually challenging. 371 Consequently, instead of separating the two British species into different genera, it appears 372 more pragmatic to extend the Harpactea genus concept and treat them as members of a single 373 Harpactea s. lat. The results of Wheeler et al. (2017) show that morphological data have so far failed to converge 376 on a stable and reliable phylogenetic reconstruction for Gnaphosoidea. Recent morphological 377 analyses by Rodrigues & Rheims (2020) and Azevedo et al. (2018) show fundamental differences 378 compared to the molecular analysis presented by Wheeler et al. (2017). For example, they place 379 Prodidomidae deep within Gnaphosidae; a placement that molecular analyses contradict with 380 strong support. On the other hand, in the molecular analysis, traditional Gnaphosidae are highly 381 polyphyletic.

382
At this point, a conservative tree will largely follow the morphological results and traditional 383 arrangements. The morphological analyses do not fully resolve the relationship between the 384 subfamilies Gnaphosinae, Zelotinae, and Drassodinae. The preferred arrangement at this level is 385 based on the molecular data.

386
The placement of Urozelotes in the tree is tentative; it could with similar justification be 387 placed as sister to Drassyllus+Trachyzelotes, rather than Zelotes. The internal structure of 388 Drassodes, Drassyllus, Haplodrassus and Zelotes is based on a combination of morphological 389 similarities and barcode data. In the case of Haplodrassus, the placement of H. minor and the 390 deeper branches are ambiguous. The relationships within Gnaphosa are based on morphological 391 similarities only.

392
Micariidae are here treated as a separate family (stat. rev.), sister to Cithaeronidae, based 393 on the results in Azevedo et al. (2018) and Rodrigues & Rheims (2020). This separation 394 seems justified given the long-standing debate about the placement of Micaria, which often was 395 included in Clubionidae instead of Gnaphosidae. Given the chaotic results for Gnaphosidae in 396 Wheeler et al., this preference is obviously only weakly supported. The internal structure of the 397 tree for the genus Micaria follows Breitling (2017). The placement of M. albovittata is based on 398 Wunderlich's inclusion of the species (sub M. romana) in the pulicaria group (Wunderlich 1980). 399 The placement of M. silesiaca is based on its inclusion in the silesiaca group (Wunderlich 1980). 400

9/29
Hahniidae 401 The inclusion of Cicurina in this family is discussed above under Dictynidae. The inclusion 402 of Mastigusa is particularly weakly supported. The British species has formerly been placed 403 in Tetrilus, Tuberta, and Cryphoeca. Its placement in Hahniidae is based on its presumed 404 relationship to Cicurina (e.g., Murphy & Roberts 2015); Wunderlich (2004a) places it in 405 Cryphoecinae instead. The two forms of British Mastigusa are here conservatively considered 406 semispecies, rather than morphs of a single species, given their apparent ecological separation. 407 However, if this interpretation is correct, the name of the macrophthalmus form would probably 408 need to be changed, as the British specimens don't seem to belong to the species originally 409 described under this name from Eastern Europe (Wunderlich 1995b). The relationships between 410 the genera are based on analysis of the barcode data available for the family, as is the internal 411 topology of Hahnia. The internal topology of Iberina is based on morphological affinities, but 412 as the male of I. microphthalma is unknown, this remains tentative. This family is particularly difficult to analyse, not just because it is the largest of the British 415 spider families, but also because a large part of recent taxonomic work has been dedicated 416 to splitting genera into poorly supported smaller units on the basis of typological arguments, 417 instead of identifying convincing relationships between genera; together with the traditionally 418 poor genus concepts in this group, this has created such a degree of confusion that even a 419 considerable amount of detailed phylogenetic analyses (both molecular and morphological) have 420 not been able to completely clarify the situation, and the phylogenetic relationships of many 421 genera remain unresolved at all levels. Additionally, while the molecular and morphological 422 analyses show some convergence in a few important areas of the tree, a large fraction of the 423 published trees is still highly unstable, and the addition of new characters or species can lead 424 to major rearrangements (see, e.g., the discussions in Miller &Hormiga 2004 andPaquin et al. 425 2008). The proposal advanced here can only be a very first attempt at providing a comprehensive 426 phylogenetic hypothesis for this group (within strict geographical limits).

427
The framework for the proposed linyphiid phylogeny is provided by the molecular analyses 428 of Wang et al. (2015) and Dimitrov et al. (2017). This is complemented by the increasingly 429 comprehensive morphological analyses of the entire family or large subgroups in Duperré & 430 Paquin (2007), Gavish et al. (2013), Hormiga (1993, 1994, 2000, Hormiga & Scharff (2005), 431 Miller & Hormiga (2004), Paquin et al. (2008), andSun et al. (2012). Most importantly, the 432 relative placement of genera required a much larger degree of personal interpretation of the 433 traditional taxonomic and morphological literature. The British Linyphiidae were comprehen-434 sively analysed in terms of pedipalp morphology by Merrett (1963) and Millidge (1977), and less 435 comprehensively in terms of their female genitalia by Millidge (1984Millidge ( , 1993. This information 436 was complemented by the phylogenetic assessments implicitly (and rarely explicitly) contained 437 in the works of Wiehle (1956Wiehle ( , 1960 and Roberts (1987), as well as a thorough assessment of the 438 morphological data encoded in the interactive key of linyphiid species by Anna Stäubli (Stäubli 439 2020, http://www.araneae.nmbe.ch). The barcode analyses presented in Breitling (2019b) pro-440 vided additional information, but were mostly used for determining the relationships within 441 genera.

442
The internal topology of Agyneta is based on a careful interpretation of barcode data, in 443 conjunction with a morphological analysis. Agyneta is a good example of a genus where species 444 identification is challenging and the resulting mis-identifications cause difficulties in interpreting 445 barcode database information. In the British fauna, Meioneta and Agyneta seem to be mutually 446 monophyletic and could be maintained as subgenera, but in the global context, they should 447 remain unified (together with a number of smaller genera) in Agyneta s. lat.

10/29
Following Breitling (2019b), Saaristoa is considered a junior synonym of Aphileta, and 449 Centromerita a junior synonym of Centromerus. In both cases, the proposed phylogenetic 450 hypotheses support this synonymy, as it is necessary to maintain the monophyly of all named 451 genera.

452
Collinsia is treated as a junior synonym of Halorates, following Buckle et al. (2001), Millidge 453 (1977, Roberts (1987), and Tanasevitch (2009). As the proposed tree shows, it would be 454 impossible to maintain C. inerrans in the same genus as C. holmgreni / C. distinctus, if H. 455 reprobus is excluded. Joining the two genera in Halorates s. lat. seems more conservative in the 456 short run, than a splitting off of C. inerrans (in Milleriana), in the absence of a comprehensive 457 revision of this and several related genera. The barcode data indicate a general confusion in this 458 group, where most genera are not recovered as monophyletic. This is not fully reflected in the 459 proposed tree, which gives priority to the morphological similarities; e.g., in its COI barcode, 460 Mecynargus paetulus seems to be closer to H. inerrans than to the type of its genus, and H. 461 inerrans closer to M. paetulus than to H. holmgreni; complementary information from a larger 462 range of molecular markers would be required to justify such a major rearrangement.

463
Dicymbium is treated as a subgenus in a considerably expanded genus Savignia, resulting in 464 a number of new combinations, as shown in the tree. This change in rank is consistent with 465 earlier proposals by Millidge (1977) concerning the expansion of Savignia to include most 466 of the members of his "Savignya genus group". It is also supported by both molecular and 467 morphological evidence as discussed in Frick et al. (2010) and Breitling (2019d). Savignia 468 (Dicymbium) brevisetosa is certainly not a subspecies of S. (D.) nigra in the current sense, as the 469 two occur sympatrically. The genitalia are indistinguishable and the two forms are not clearly 470 ecologically distinct, although syntopic occurrence apparently is rare; it is therefore quite likely 471 that they are synonymous, brevisetosa merely being a geographically restricted variant of the 472 male prosomal morphology, as suggested by Roberts (1987). However, the genetic barcode data 473 show two clusters (BINs) among the Dicymbium specimens, which could indicate the presence 474 of two closely related species, one of which might correspond to the brevisetosa form, occasional 475 intermediate specimens being the result of sporadic hybridisation. The two forms are therefore 476 here considered conservatively as semispecies.

477
Erigone maritima is considered a separate species, distinct from E. arctica s. str., based on the 478 considerable barcode gap between Nearctic and Palaearctic specimens identified as "Erigone 479 arctica" s. lat. Whether the palaearctic species can be meaningfully subdivided into subspecies is 480 currently an open question; given the high mobility and vast range of Erigone species, which are 481 among the most frequent aeronauts, a relevant subspecific differentiation seems rather unlikely. 482 Many of the morphologically well-defined Erigone species show a surprisingly narrow barcode 483 gap, indicating relatively recent differentiation and arguing further against the probability of the 484 existence of morphologically all but cryptic subspecies.

485
Mermessus (sub Eperigone) was considered as probably closely related and possibly the sister 486 group of Erigone s. lat. by Millidge (1987), and the barcode data support this placement.

487
Erigone longipalpis meridionalis is a phantom species as defined by Breitling et al. (2015, 488 2016) and probably only represents intraspecific variation of E. longipalpis. It is thus considered 489 a nomen dubium and not included in the tree.

490
Frontinellina is considered a junior synonym of Frontinella, because of the close genetic 491 affinities between representatives of the two genera.

492
Hilaira is considered a senior synonym of Oreoneta. When separating Oreoneta from Hilaira, 493 Saaristo & Marusik (2004)  Maso and Pocadicnemis are strongly united in the barcode data; their position relative to other 499 higher erigonines is less clear. They are placed in the same group by Merrett (1963; Group E) and 500 Locket & Millidge (1953; all tibiae with 1 dorsal spine; with Tm IV), but these are rather large 501 groups, and the morphology of the two genera does not indicate a particularly close relationship 502 to each other or other genera.

503
Oryphantes is considered a senior synonym of Anguliphantes, Improphantes, Mansuphantes 504 and Piniphantes, following Breitling (2019b), and Palliduphantes antroniensis is also transferred 505 to Oryphantes s. lat., where it belongs on the basis of its genital morphology (Bosmans in Heimer 506 & Nentwig 1991), as confirmed by barcode information. As explained in Breitling (2019d), 507 the synonymy is also supported by the observation of Wang et al. (2015) that a combination 508 of a large number of genetic markers, including mitochondrial (COI and 16S) as well nuclear 509 sequences (18S, 28S, H3), recovers Anguliphantes and Oryphantes as mutually polyphyletic 510 with strong bootstrap support.

511
Millidge (1977) and Merrett (1963) point out similarities between Ostearius and Dona-512 cochara/Tmeticus, and Wiehle (1960) places Ostearius in his Donacochareae. However, this 513 traditional placement of Ostearius in a clade with Tmeticus and Donacochara has long been 514 dubious, and it is not supported by any of the recent analyses. Even the sister group relationship 515 between the latter two is not strongly supported by any of the newer data. Hormiga (2000) and 516 subsequent morphological assessments place Tmeticus far from Ostearius. The barcode data also 517 do not indicate a close relationship: there, Ostearius is sister to Eulaira, matching Millidge's 518 earlier morphology-based proposal (Millidge 1984).

519
Pelecopsis susannae is transferred to Parapelecopsis, based on similarity of genitalia and 520 absence of dorsal spines on its tibiae. As this indicates that the boundary between the two genera 521 is not quite clear, they are here treated as subgenera of Pelecopsis s.lat., and in the global context 522 Parapelecopsis should possibly be discarded altogether.

523
Poeciloneta is treated as a senior synonym of Agnyphantes and Obscuriphantes. While the 524 necessity of this merger is not obvious in the context of the British fauna, where each of these 525 genera is represented by a single species, the global analysis shows that this move is required to 526 obtain a monophyletic genus Poeciloneta.

527
In the case of morphologically homogeneous genera, where even the species boundaries have 528 long been ambiguous and species groups have been fluid at best, in the absence of genetic data 529 the proposed phylogenetic relationships can be little more than a poorly educated guess. The 530 genus Porrhomma is a good example of this situation. The preferred tree presented here is based 531 on a rather subjective assessment of the morphological affinities of the included species.

532
The placement of Pseudomaro as sister of Mioxena is based on unpublished data on the mor-533 phology of the males (A. Grabolle https://wiki.arages.de/index.php?title=Pseudomaro_aenigmaticus). 534 These indicate that the two genera may even be synonymous, but a formal synonymisation should 535 await a formal publication of the description of male Pseudomaro specimens.

536
Savignia is here considered in the broadest sense, as discussed in Breitling (2019d). It includes 537 the former genera Dicymbium, Minyriolus, Glyphesis, Araeoncus, Diplocephalus and Erigonella. 538 Various earlier authors, including Bosmans (1996), Frick et al. (2010), Holm (in lit. in Millidge 539 1977), and Millidge (1977 had already found that this group is so homogeneous and the genera 540 so poorly defined that they should probably be merged in a single genus. The barcode results 541 confirm this assessment. The subgenus assignments try to identify monophyletic groups, at least 542 within the context of the British fauna, but they are tentative only, given that no comprehensive 543 global analysis of the genus group has been performed, and their practical value could be 544 debated. Savignia connata jacksoni is considered an infrasubspecific variant of Savignia connata, 545 following Roberts (1987), and is therefore not included separately in the tree.  (2013), which analyses the densest sample of species, including 553 all genera found in the British Isles, presents a monophyletic Liocranidae s.lat. The preferred 554 tree presented here shows a compromise between the different analyses: while it proposes that 555 the British representatives of Liocranidae are united in a monophyletic group, it modifies the 556 arrangement of genera suggested by Bosselaers & Jocqué (2013) to match the observation by 557 Ramírez (2014) that Liocranum and Apostenus are more closely related to each other than to 558 Agroeca (which Ramírez wants to remove to Clubionidae). Scotina was not included in the study 559 by Ramírez (2014), but is morphologically closer to Agroeca, although historically, the species 560 of this genus have been placed in Agroeca, Liocranum, and Apostenus (S. palliardii in all three). 561

581
In Pardosa, the deep branches of the tree are inferred on the basis of morphological affinities, 582 while barcode data resolve the internal relationships within species groups, including a number 583 of semispecies complexes, which are particularly common in this genus.

585
The tree is based on the morphological assessments presented by Thaler et al. (2004), which 586 are fully consistent with the barcode data. The placement of Ero tuberculata, which remains 587 ambiguous in Thaler et al. (2004), is based on somatic and genital morphology, which suggests a 588 13/29 closer relationship between E. cambridgei and E. furcata, than between either of them and E. 589 tuberculata.

591
The backbone of the arrangement in this family is based on the barcode data available for three 592 of the species, as the genus present in Britain is morphologically rather homogeneous. The 593 placement of Zora armillata is based on its morphological similarity to Z. spinimana, implied in 594 the determination keys presented by Urones (2005) and by Wunderlich in Heimer & Nentwig 595 (1991). Of course, this is a rather weak argument, as these keys are intended as pragmatic aids to 596 identification, not as statements of phylogenetic hypotheses.

597
Nesticidae 598 Pavlek & Ribera (2017) illustrate the close relationship of Nesticus and Kryptonesticus, based on 599 morphological and molecular data. Nesticella is morphologically quite distinct, and barcode data 600 support the proposed arrangement.  2014) include the British genera in their analysis, but the resolution of their tree 606 is low and the phylogeny preferred here has only low support. Orchestina dubia could be a 607 considered a phantom species following the definition by Breitling et al. (2015Breitling et al. ( , 2016. In 608 contrast to other phantom species mentioned here, there remains a distinct possibility it will turn 609 out to be a valid species, and it is therefore included in the tree for completeness.

611
The basic relationships between genera are based on Wheeler et al. (2017), while internal 612 relationships are based on the data presented in Breitling (2019b).

613
The placement of Philodromus buchari is tentative, based on the morphological affinities 614 indicated by Muster & Thaler (2004).  Morphologically (in terms of palp, endogyne, colour pattern, and leg spines), Segestria bavarica 633 and S. florentina seem to be closer to each other than to S. senoculata, but barcode data for 634 S. florentina are not yet publicly available, and barcode distances indicate no especially close 635 relationship between it and S. bavarica. The similarities between the two species could all be 636 symplesiomorphic, but uniting them in the tree still seems the most plausible scenario.

649
The genus Steatoda in the usual sense is clearly highly polyphyletic in the analyses by Liu et 650 al. (2016); creating a monophyletic Steatoda s. lat. would require merging the genus with both 651 Crustulina and Latrodectus (and possibly other Asagenini); certainly not a desirable solution. 652 Instead, Steatoda is here tentatively divided into a number of smaller genera, all of which had 653 been postulated previously, as classic authors have long recognised the heterogeneity of the 654 genus. It is, however, important to realise that the molecular subdivisions are not closely aligned 655 to previous morphology-based ideas (e.g., Wiehle 1937;Wunderlich 2008), and the two sets of 656 results are difficult to reconcile.

657
Relationships within Enoplognatha are based on barcode data, E. oelandica being placed 658 based on morphology (but with low confidence). The semispecies relationship between E. 659 ovata and E. latimana is based on Lasut et al. (2015) who show convincing evidence based on 660 mitochondrial and nuclear markers indicating that the two forms are not yet fully reproductively 661 isolated, despite their obvious genitalic differences. Levi (1973) cites a personal communication 662 from V. Seligy stating that both forms of the genitalia can be found among siblings from the 663 same egg sac.

664
The internal topology of Robertus is also based on barcodes, but with weak support, the 665 15/29 placement of R. insignis being based on its similarity to the barcode-sequenced R. lyrifer 666 (Almquist 1978).

667
Phycosoma inornatum is here considered a member of Lasaeola, following Wunderlich (2020). 668 This placement is problematic, given that both of the genera, as well as Dipoena seem to be poorly 669 delimited. The transfer to Lasaeola is justified by the observation that European "Phycosoma" is 670 unlikely to be congeneric with the type species from New Zealand: it lacks an epigynal scapus 671 [present in true Phycosoma] and has a relatively large embolus and conductor [small in true 672 Phycosoma]; the male prosoma is also very different and matches that of other Lasaeola species. 673 The species thus lacks the most important genus diagnostic characters of Phycosoma. Otherwise, 674 the arrangement of the species is maximally conservative and maintains Dipoena and Lasaeola 675 as separate genera, despite concerns about their possible para-or polyphyly. Arrangements 676 within each genus are based on morphological similarities (e.g., Miller 1967).

685
Cryptachaea riparia is considered a member of Parasteatoda, based on the arguments detailed 686 in Breitling (2019d). The internal relationships between the Parasteatoda species are based on 687 barcode data and morphology.

688
The relationships between the Rugathodes species are based on morphological similarities.

689
The genus Theridion is clearly polyphyletic on a global scale, and its phylogeny inferred on 690 the basis of morphology and barcode data is not always consistent with more comprehensive 691 molecular phylogenies. To retain monophyletic genera in the tree, T. pinastri is placed in 692 Allotheridion, following Archer (1950): barcode data place the species very close to the type 693 species, and the general genus concept proposed by Archer seems validated by the molecular 694 data, including the closeness to Phylloneta. The transfer of T. hannoniae to the same genus is 695 based on its membership in the petraeum-group (Bosmans et al. 1994). Platnickina tincta is 696 returned to Theridion s. str., to be conservative and avoid changing the name of common species 697 that have always been placed in Theridion.

16/29
Ozyptila maculosa is a phantom species as defined by Breitling et al. (2015Breitling et al. ( , 2016, possibly 712 referring to a malformed specimen of O. atomaria. It is thus considered a nomen dubium and not 713 included in the tree.

715
The proposed topology is based on the informal, non-cladistic species groups proposed by 716 Bosmans (1997), which seem not entirely inconsistent with the limited barcode data available 717 for the family.  Figure 2 illustrates the phylogenetic distribution of a number of classical morphological 722 traits traditionally used in the identification of linyphiid spiders. The phylogenetic tree allows 723 organising a large and unwieldy data matrix to facilitate the identification of patterns. While 724 there is a high degree of homoplasy for all characters, many show clear trends in agreement with 725 traditional (typological) classifications of Linyphiidae, such as the loss of tibial spines (Wiehle 726 formula) and the distal movement of the trichobothrium on metatarsus 4 in "higher" linyphiids. 727 In many cases, these trends extend beyond genus boundaries, and their identification requires the 728 more detailed framework provided by the phylogenetic tree, e.g., regarding the the large body 729 size in the "basal" Linyphiinae (Linyphia, Frontinella and their relatives). Another obvious use 730 of this character map is its application as a starting point for correcting the tree and proposing 731 better alternative hypotheses. 732 Figure 3 uses the tree to search for potential correlations between phylogeny and ecological 733 traits. Data on the distribution of spiders in Great Britain (at the hectad level) were obtained 734 from the Spider Recording Scheme database and a number of ecological indicator values were 735 calculated for each species (excluding accidental introductions for which ecological analyses 736 would be inappropriate): (1) an estimate of the abundance of the species, based on the number of 737 occupied hectads; the weight of each hectad was decreased by 1% for each year since the latest 738 record from this hectad; (2) an indication of the recency of observations, based on the median 739 year of the latest record per hectad; (3) a characterisation of the North-South distribution based 740 on the median distance of the occupied hectads from the southern edge of the British Ordnance 741 Survey national grid; (4) a "Vulnerability Index", 1 minus the quantile of the average of the 742 recency quantile and the abundance quantile for each species, where high values (close to 1) 743 indicate species that are reported from only a small number of hectads and have few or no recent 744 records, while low values (close to 0) indicate species that are widespread and have often been 745 recorded recently; (5) a "Contraction Index", i.e., the quantile of the difference between the 746 abundance and recency quantiles, where high values (close to 1) indicate widespread species, 747 however with relatively few recent records, while low values (close to 0) indicate highly localised 748 species that have nevertheless a large fraction of recent records (e.g., new arrivals and expanding 749 species).

750
For each of the more than 680 subtrees a Wilcoxon rank sum test was applied to identify 751 clades that are significantly enriched in members with particularly high or low values for each 752 of these ecological indicators. In many cases, the results confirm previously reported informal 753 observations. For instance, the clade most significantly enriched in species with a northerly 754 distribution includes all the Linyphiidae compared to the rest of the species (Wilcoxon p-value 755 < 5 × 10 −26 ). However, the phylogenetic tree also allows an analysis at a much finer resolution. 756

17/29
In Figure 3 a few selected examples are highlighted. Some of the results follow traditional 757 family or genus boundaries and could be obtained without the help of a fully resolved tree. 758 For example, the Thomisidae are significantly enriched in species with a southern distribution 759 (Wilcoxon p-value < 5 × 10 −5 ).

760
More interesting, however, are those trends that are only possible to identify within the frame-761 work of a fully resolved tree: for instance, the clade combining Mimetidae and Tetragnathidae is 762 significantly enriched in species with a particularly low Vulnerability Index (Wilcoxon p-value < 763 5 × 10 −5 ), i.e., these species are on average more widely distributed and more recently recorded 764 than the members of other clades. The fully resolved tree also allows more detailed examination 765 of trends within families, beyond genus boundaries. For example, species in the two clades 766 from Floronia bucculenta to Poeciloneta variegata and from Wabasso replicatus to Erigone 767 atra in Figure 3 are highly enriched in species with a northern distribution (Wilcoxon p-values 768 < 3 × 10 −5 and < 3 × 10 −7 , respectively). This is one more example showing that ecological 769 traits are correlated with phylogeny, beyond the coarse-grained categories of traditional Linnaean 770 classification, as has been shown for other taxa before (e.g., Thomas 2008 for birds).

771
This is particularly interesting for those cases, where the ecological variables indicate potential 772 conservation concerns. For instance, in Figure 3, the clade from Evansia merens to Walckenaeria 773 mitrata is strongly enriched in species with a high Contraction Index (Wilcoxon p-value < 774 5 × 10 −5 ), i.e., these species show a surprising lack of recent records for such widespread species. 775 There are numerous potential explanations for this phenomenon, not all of which indicate an 776 immediate conservation concern, but this example nevertheless illustrates the potential of future 777 applications of the fully resolved tree for spider conservation research.

778
In conclusion, it is perhaps useful to reflect on the degree of confidence in the presented 779 tree. As stated repeatedly, there remain multiple areas where the available evidence does not 780 yet allow a highly confident decision between alternative phylogenetic hypotheses. A fully 781 resolved phylogenetic tree for the 680 British spider species could contain at most on the order 782 of 650 "mistakes" relative to the true phylogenetic history of the group (the maximum number 783 of branch moves required to transfer the tree into the correct one; Atkins & McDiarmid 2019). 784 The expected number of mistakes in a random tree would be on the order of 600.

785
Of course, one would hope that the tree proposed here is far from random and provides 786 a good initial approximation of the true phylogeny. Nevertheless, the number of mistakes is 787 probably still considerable, given the remaining instability and incompleteness of the underlying 788 datasets. This is a problem shared with almost all published phylogenies: for instance, Miller 789 & Hormiga (2004) report that their analysis of erigonine phylogeny has only 5-6 nodes (about 790 20%) in common with the topology proposed for the same genera just 4 years earlier by Hormiga 791 (2000), despite using very similar methodology. They suggest that 50-53 branch moves would 792 be required to convert the trees into one another. Even if this appears to be an overestimate, 793 considering the results of Atkins & McDiarmid (2019), it indicates that at least one of the trees 794 was still very far from reconstructing the true evolutionary history of this small sample of the 795 subfamily.

796
If the tree proposed here, for a much larger number of taxa, contains a similar number of 797 mistakes, this could be considered a major success. The number of errors in the presented tree 798 is difficult to estimate objectively, but the above calculations can provide a valuable point of 799 reference: each reader is likely to find some parts of the tree where their personal interpretation 800 of the data would lead them to prefer a different arrangement of the species. Each of them will 801 be able to count how many corrections ("branch moves") would be required to transform the 802 presented tree into their preferred one. They can then compare the number of corrections to the 803 number of errors expected in a random tree; the ratio between the two numbers provides a metric 804 of the (subjective) correctness of the proposed phylogeny.

805
Of course, phylogenetic relationships for British spiders have been proposed before, not only 806 18/29 for individual small groups, but implicitly also for the entire order, for instance in the arrangement 807 of species in the works of, e.g., Locket & Millidge (1951, 1953 or Roberts (1985Roberts ( , 1987. But 808 even when the proposals were made explicit in the form of tree diagrams, they sometimes failed 809 to achieve the level of precision and specificity that would make the suggestions testable in 810 an objective way. A good example is the tree proposed by Millidge at the conclusion of his 811 analysis of the (mostly British) Linyphiidae (Millidge 1977:fig. 200): here, the tree not only 812 consistently shows extant genus groups as "evolutionary precursors" of other groups and includes 813 a number of tentative alternative branches, but most importantly it leaves a large number of 814 unresolved polytomies. All this makes the proposals not only difficult to confirm or falsify, but 815 challenges even the simple, direct comparison of individual proposals. This changes only when 816 dichotomous trees are fully specified; this enabled, for instance, the detailed and quantitative 817 comparison of alternative trees in Miller & Hormiga (2004). But even recent studies, which 818 present fully resolved trees, often present multiple alternative topologies instead of identifying a 819 single preferred tree, thus again reducing the resolution of the results by implying polytomies 820 and diminishing the predictive content and testability of the hypotheses.

821
The explicit statement of all phylogenetic hypotheses in the form of a single completely 822 resolved tree, i.e., a tree that only includes bifurcations and avoids polytomies, down to the 823 level of individual species, makes the proposals of the present synthesis eminently testable. It is 824 hoped that future analyses will identify errors and omissions in the presented tree and suggest 825 alternative, better-supported hypotheses. At the same time, the tree might serve as a basis for 826 extended analyses, perhaps initially extending the geographic scope to neighbouring countries, 827 but ultimately resulting in a completely resolved tree of the global spider fauna. Testing the monophyly of the ground-dweller spider genus Harpactea Bristowe, 1939 (Araneae, Dysderidae) with the description of three new species.
Systematics and Biodiversity 18 (7)  Index is the quantile of the difference between the abundance and recency quantiles. Higher 852 values indicate widespread species with relatively few recent records; low values indicate highly 853 localised species that have nevertheless a large fraction of recent records (e.g., new arrivals and 854 expanding species). The clades highlighted in colour are discussed in more detail in the text. 855 The order of columns (from left to right) follows the order of the legends (from top to bottom