Approximately 46,500 specimens of Laparocerus were collected in the field and identified by the first author, unless otherwise specified in Appendix 1. At least two specimens of each presumably distinct species or subspecies were directly introduced in absolute ethanol, and in many cases this was repeated for different localities. Ethanol was replaced the next day and samples preserved at -20°C. Voucher card-mounted dry specimens and specimens preserved in ethanol, from the same locality and date, share the same database collection number. This information for the specimens included in the present study is provided in Appendix 1. All this material is kept in the Machado Collection, La Laguna, Tenerife.

Eleven known species were not found alive in nature and for this reason they were excluded from the present analysis. From Madeira: Laparocerus (Lichenophagus) acuminatus (Wollaston, 1854), L. (Atlantodes) navicularis (Wollaston, 1854), L. (Atlantodes) lanatus (Wollaston, 1854), and L. (Anillobius) porctosantoi (Franz, 1970); from the Selvagens L. garretai albosquamosus Machado, 2011; and from the Canaries: L. (Purpuranius) fraterculus Machado, 2012 and several hypogean species or subspecies of the subgenus Machadotrox, which are normally scarce and difficult to obtain: L. zarazagai zarazagai García and Oromí, 1997; L. iruene García and Machado, 2011; L. machadoi García and González, 2006; L. idafe García and Alonso Zarazaga, 2011, and L. cavernarius Machado, 2011. This set of missing taxa in the analysis represents nearly about 5% of the total of known Laparocerus (237).

Plant genera mentioned in the text, and their respective families (Mabberley 1997) are listed in Appendix 3.

In our first contribution (Machado et al. 2008) fragments of the mitochondrial cytochrome c oxidase subunit II (COII, 598 bp), the 16S ribosome large subunit RNA (16S rRNA, 427 bp), and, only for representatives of species groups, the nuclear elongation factor 1 alpha (EF-1α, 611 bp) were used as genetic markers. In addition to the aforementioned molecular markers, in this study we also use the mitochondrial 12S ribosomal RNA (12S rRNA, 344 bp) and the nuclear 28S ribosomal RNA (28S rRNA, 762 bp) genes both sequenced for all Madeiran and Canarian OTUs (see Appendix 1).

DNA was extracted using DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA Inc) following the instructions of the manufacturer. All PCR reactions were carried out in a Veriti Thermal Cycler (Applied Biosystems, USA) in a final volume of 25 µl containing 1× buffer (GeneAll, Korea), 150 µM of each dNTP, 0.2 µM of each primer, 0.5 U AmpONE™ Taq DNA polymerase (GeneAll, Korea) and 10–20 ng of DNA template. Thermal profile for COII, 16S rRNA and EF-1α fragments were as described in Machado et al. (2008). PCR conditions for 12S rRNA and 28S rRNA were as follows: 2 min at 94°C followed by 35 cycles of denaturation at 94°C for 10 s, annealing at 54 or 56°C respectively for 20 s, and extension at 72°C for 30 s, with a final extra extension step at 72°C for 5 min. Subsequently, the primers and nucleotides were removed with Illustra^TM ExoProStar^TM 1-Step (GE Heathcare, Life Sciences) according to the manufacturer’s instructions. Primers used for the PCR are listed in Table 1. Sequencing was carried out by the sequencing facilities of the company MacroGen Europe (Amsterdam).

One specimen, preferential from the type locality, was analysed for each of the 225 taxa sampled, except when the taxon was present in different islands, in known putative vicariance regions within the same island (e.g. Teno/Anaga in Tenerife), or when morphological differences associated with marginal localities were noticed. Due to such situations, a total of 30 additional sequences was included in the analysis. Moreover, in order to minimise laboratory errors (contamination, mislabelling, etc.) sequencing was repeated for discordant results, particularly, with taxa strangely placed according to traditional morphology (occasionally a second specimen from the same locality was used). Sequencing with both the forward and reverse primers was performed only in cases of not clean or incomplete chromatograms. For a few species (L. aethiops, L. auarita, L. canariensis, L. morio, L. vespertinus), several individuals from the same locality were sequenced for COII to get a more accurate idea of the range of local intraspecific genetic divergence with this marker.

A total of 1425 sequences was obtained: 441 for the COII, 322 for the 16S rRNA, 294 for the 12S rRNA, 290 for the 28S rRNA, and 78 for the EF-1α. All duplicate and redundant sequences – from the same or different localities – were removed from the combined matrix of COII+16S rRNA+12S rRNA for the final analysis, which ended up with a total of only 256 OTUs, representing 223 different Laparocerus taxa and two outgroups. This final set of sequences has been deposited in GenBank (www.ncbi.nlm.nih.gov/Genbank) with following accession numbers: EF583315 – EF583371, FJ495251 – FJ495253, KX551687 – KX551907 for the COII; KX550955 – KX551210 for the 12S rRNA; FJ495254 – FJ495256, KX551211 – KX551431 for the 16S rRNA; KX551432 – KX551686 for the 28S rRNA; and EF583372 – EF583389, KX551908 – KX551958 for the EF-1α.

The relationships among the many genera and tribes of Entiminae are still unsolved and pose long endeavour ahead (videOberprieler et al. 2007, Hundsdoerfer et al. 2009). Very few phylogenetic studies attempting to organise the present chaos include any Laparocerus or are little conclusive (e.g. Davis 2014, Stüben et al. 2015).

Alonso-Zarazaga and Lyal (1999) listed eight genera for the tribe Laparocerini Lacordaire, 1863, of which six have later been removed and assigned to other tribes (Machado 2010, Alonso-Zarazaga 2013). The only genus left besides Laparocerus, was Straticus Pascoe, 1886 which presumably belongs also to the African Peritelini as do some of the other putative Laparocerini. Moreiba Alonso-Zarazaga, 2013 recently described from the Canary Islands, is a true Laparocerini with presence of spiculum relictum in the male sternite VIII, but the internal sac of aedeagus has the gonoporus in normal position and the tibiae are lacking mucro. It was tested as an outgroup and will be commented upon below.

In the Madeiran clade analysis (Machado et al. 2008) Rhyncogonus excavatus Van Dyke, 1937 from the island of Rurutu in the French Polynesia, was selected and used as a formal outgroup. In Machado (2010) several other genera of Entiminae were checked for closer relationship with Laparocerus based on 16S rRNA sequences obtained by us or taken from the NCBI (National Centre of Biotechnology Information). Unfortunately, the number of different genera sequenced is limited and Laparocerus does not relate with confidence (Bayesian support) to any other genus. Nonetheless, we selected new outgroups in order to avoid long-branch attraction (s. Anderson and Swofford 2004), those being Barypeithes indigens (Boheman, 1834) (Sciaphilini), and Brachyderes rugatus Wollaston, 1864 (Brachyderini), two of the less divergent (p-distance) genera tested.

Moreiba was also tested directly as outgroup, but it showed unstable behaviour jumping from the Madeiran clade to the Canarian clade or outside both of them, depending on the individual gene or combination of genes used. Moreiba is clearly related to Laparocerus from the morphological point of view (Alonso-Zarazaga 2013), but lacking DNA from other genera of Laparocerini that could help fixing its position, we disregarded it as an outgroup and excluded it from the final analysis.

DNA sequences were viewed, edited and assembled using MEGA 6 (Tamura et al. 2013). Alignments were achieved using the program Muscle (Edgar 2004) with default parameters as implemented in MEGA 6 and tuned by eye. Each marker was tested for hipervariational loci with GBlocks (Castresana 2000); no fragments were removed. The plausibility of the alignment of COII and EF-1α sequences was verified at the amino acid level. The entropy-based index as implemented in Dambe 5.2.78 (Xia et al. 2003) was used to assess substitution saturation within the mtDNA and 28S rRNA sequences, with negative results.

Alignment of 16S rRNA and 28S rRNA included 5 and 17 indels, respectively. These positions were considered as missing data for all analyses. In the case of the 12S rRNA, shared indels seemed to express relations judging from the known taxonomy (e.g. same subgenus) and were coded (1 or 0) with FastGap 1.2 (Borchsenius 2009), increasing the sequences from 344 bp to 364 bp.

Genetic divergence of all sequence pairs (genetic distance, gamma distributed with invariant sites G+I), the p-distance means between and within each subgenus, and the means between and within the Canarian and Madeiran subsets were calculated with MEGA7 (Kumar et al. 2016) after removal of all positions containing gaps and missing data, and eliminating duplicate taxa by keeping only the sequence of the type-locality specimen or that with the higher divergence value (223 taxa in the final dataset). In the calculation of within subgenera divergence, species grouped as incertae sedis and Lichenophagus (only one species available) were not considered.

Phylogenetic relationships were reconstructed using Bayesian inference (BI). Nucleotide substitution model parameters were obtained with jModelTest 2.1.4 (Darriba et al. 2012) using the Bayesian Information Criterion (Schwarz 1978), with the following results: TIM1+I+G for the mtCOII; TIM3+I+G for the 16S rRNA; HKY+I+G for the 12S rRNA, and for the nuclear markers TVM+I+G for the 28S rRNA, and TIM2+IU+G for the EF-1α.

Analyses were conducted using Bayesian Markov chain Monte Carlo inference (Yang and Rannala 1997) as implemented with MrBayes 3.2.3 (Ronquist et al. 2012) running on the facility Mobyle SNAP Workbench at the North Carolina State University (Monacell and Carbone 2014). We used the combined dataset of the mitochondrion markers partitioned (256 OTUs, 1389 bp), and the previously determined models of nucleotide evolution. Parameters were treated as unknown variables with equal a priori probability and subsequently estimated by the programme during the analysis. Starting trees were randomly chosen. Two independent 10,000,000 generation runs of eight Monte Carlo Markov chains ‒ two cold, six heated at 0.02 ‒, were conducted (nswap = 5), and trees being sampled every 100 generations for a total of 100,000 trees in each of the initial samples. Variations in likelihood scores were examined graphically with the Tracer v1.6 application (Rambaut et al. 2014) and the first 2,500,000 generations were discarded, having ensured that stationarity was reached. Accordingly, the first 25,000 trees were discarded as burn-in, and the following 75,000 trees were used to estimate topology and tree parameters.

Similar BI analysis were repeated, 16,000,000 generations, adding to the mitochondrial matrix the 28S rRNA sequences (total 2.151 bp) and, for some selected OTUs (78), the EF-1α (total 2.762 bp). The trees obtained were used to check consistency with the mitochondrial only based results. We confirmed that there is no significant incongruence between the information provided by each gene using the partition homogeneity test of Farris et al. (1994) as implemented in PAUP*4.0b10 (Swofford 2002), with 500 replicates. The nuclear genes were the most incongruent (with p = 0.10, but lower than p = 0.05) and the values obtained for the three mitochondrial genes (p = 0.894) and for the whole set of five mitochondrial and nuclear genes (p = 0.868) reflect total congruency.

Maximum likelihood trees for all markers and combined sets were also reconstructed using RAxML (Stamatakis 2014) as implemented in Mobyle SNAP Workbench under the above-mentioned models. No differences with the BI trees were found.

The BI final phylogram was edited with TreeGraph 2 (Stöver and Müller 2010) and, due to its length, the tree was fully collapsed or divided in separate pieces for presentation. Clades and subclades are organised from older (bottom of tree) to younger (top of tree). Species names are coloured in the phylograms to indicate the island of origin of the specimen sequenced (a legend is provided with each subtree). In most cases, Laparocerus species are single-island endemics and that gives an overall idea of their distribution. Species that live in more than one island have been sequenced for each island, with the exception of L. ellipticus from El Hierro, which failed.

The colonisation pathways have been inferred from the tree topology under criterion of parsimony, assuming the uncertainty derived from having analysed mostly one specimen per species, and lacking total knowledge about extinctions.

A molecular clock test was performed with MEGA7 by comparing the ML value for the given topology with and without the molecular clock constraints under GTR model. The null hypothesis of equal evolutionary rate throughout the tree was rejected at a 5% significance level (P = 0) for all individual markers. Consequently, a timetree was built using the program RelTime (Tamura et al. 2012) under relaxed molecular clock hypothesis (local clocks) and using the combined mitochondrial tree obtained from the BI analysis. This module of MEGA7 calculates divergence times using the Maximum Likelihood method based on the General Time Reversible model (Nei and Kumar 2000). The relative times are converted to absolute divergence times based on constraints of the calibration point supplied.

Twenty potential calibration points were tested giving preference to the nodes of vicariant species present in El Hierro or La Palma, or radiations within these islands which are the youngest in the Canaries, with a geological age of 1.12 Ma (Carracedo 2011) and 1.72 Ma respectively (Guillou et al. 2001). Species with a plausible ancestor from a source-island that was not sequenced or is unknown were disregarded. For instance, Laparocerus (Amyntas) incomptus is endemic to El Hierro and shows a basal position as sister taxon to the rest of species in this subgenus that inhabit much older islands. However, there is no Amyntas known from La Gomera (9.4 Ma) from where it probably originated. In the case of Madeira, nodes implying radiations within the main island (4.8–5.2 Ma, f. Schmincke 1998) were chosen. Following the idea of Near and Sanderson (2004), but without fossils, we crosschecked the calibration points, disregarding those chronograms when any of the points surpassed the age of their island.

Finally, a single calibration point (see black triangle in Fig. 7) was selected, namely the radiation event of four species of the subgenus Machadotrox which are endemic to La Palma (two of them hypogean and blind), not allowing it to be older than the island age of 1.72 Ma, or younger than 0.21 Ma. This minimum age constrain was obtained by starting with 0.01 Ma and increasing it until the estimated age of Node P of Madeira (Atlantis and Pseudatlantis species) dropped below the age attributed to this island. That gives an ample margin for colonisation of La Palma to happen after the island emerged.

In the chronogram obtained (see Suppl. materials 1–3) bars around each node represent 95% confidence intervals which were computed using the method described in Tamura et al. (2013). The estimated log likelihood value of the topology shown is -32807.3327. A discrete Gamma distribution was used to model evolutionary rate differences among sites (6 categories (+G, parameter = 0.5622)). The rate variation model allowed for some sites to be evolutionarily invariable ([+I], 42.7318% sites).

In the summarized Table 2 mean values of age estimates are included for the main nodes and MRCA (most recent common ancestor) of subclades representing subgenera, as they should be more reliable than divergence times at leaf levels subject to individual variability. Having worked mostly with one specimen per species implies risk of depicting relations linked to the particular individual sequenced, and that may inflate or deflate inferred times and rates of evolutionary divergence if there is some kind of underlying polyphyly in the species (Funk and Omland 2003). Our date estimates must to be taken with some caution.

For the COII, 326 out of 598 positions (54.5) were variable and 272 informative (45.5%); for the 12S rRNA raw fragment 184 out of 344 positions (53.5%) were variable and 145 informative (42.28%), and for the 16S rRNA fragment, 163 out of 427 positions (38.2%) were variable and 122 informative (28.6%). Third positions of COII showed low G composition (1.7%) as is typical in insect mitochondrial DNA coding genes.

Xia’s index for substitution saturation in COII produced values of 0.074 (first and second codon positions) and 0.46 (third codon position) which were significantly lower than the critical value for symmetric topologies (0.69–0.78, P < 0.001; 0.69–0.77, P < 0.001 respectively), suggesting that sites have reached little saturation and sequences can be reliably used for phylogenetic reconstruction. In the case of the 28S rRNA and the other genes saturation was also amply disregarded (data not shown).

The overall mean genetic divergence (p-distance) among species is 11.7% in COII, 5.4% in 12S rRNA, 5.2% in 16S rRNA, 1,0% in 28S rRNA, and 8.2% in the combined set. To gain a rough idea of the local genetic divergence, we sequenced the COII of five species with four specimens each collected in the same locality, obtaining the highest value of 1.3% in the case of Laparocerus canariensis from El Portillo, Tenerife.

The overall mean p-distance in the mitochondrial 3-gene combined matrix is 8.5% in the Madeiran subset and 7.3% in the Canarian subset, with a maximum of 13.3% between a Canarian and a Madeiran species in two cases: Laparocerus colonnellii/L. calcatrix and L. rugosivertex/L. chaoensis (from Bugio).

Mean p-distance within subgenera ranks from 2.9% in Fernandezius to 7.0% in Atlantis, with an overall mean average of 4.6%. Mean p-distance between groups ranks from 5.4% between Bencomius or Belicarius and Canariotrox to 12.0% between Aridotrox and Atlantis, with a global average value of 9.2% (9.4% in Madeiran groups and 7.6% in Canarian groups). For more information on genetic divergence, see Appendix 2.

In order to facilitate readability and exposition of results, the global phylogenetic tree obtained for the combined set of three mitochondrial markers is displayed in Figure 3 in a collapsed summary form, and thereafter expanded trees of subclades or groups of them are presented individually. The complete phylogram, as well as the 4-gene tree (with 28S rRNA added to the set), are available as supplementary materials to this paper.

In Figure 3 compact subclades are collapsed and labelled after subgenera (established or newly described here). There is one case of paraphyly in the Madeiran clade (one species of Atlantis Wollaston, 1854 clusters with species of Pseudatlantis Machado, 2008), and the subclade ‘Pecoudius’ has not been divided in subgenera pending further study, so we adopt in this contribution a wide sense (s.l.) for Pecoudius Roudier, 1957. Basal nodes are identified with letters for referring purposes.

The Bayesian support is high (BPP > 0.95) in most cases and it rises up from 0.81 to 0.97 in Node N, from 0.94 to 0.99 in Purpuranius, from 0.88 to 0.94 in Fortunotrox, and from 0.92 to 0.99 in Machadotrox, when the nuclear marker 28S rRNA marker is added to the analysis.

A general picture of the estimated ages of the main lineages expressed as mean values is provided in Table 2, with indication of the island distribution of each of lineage (see the complete chronogram with confidence bars in the supplementary files).

For the combined set of three mitochondrial markers we have calculated an overall divergence rate of 3.1 Ma^-1 by dividing the between groups mean divergence (12.2%) by the average group age (3.98 Ma). In this case, the divergence values between sequences used to obtained the means have been corrected following the Maximum Composite Likelihood model (Tamura et al. 2004), with a Gamma distribution (shape parameter = 0.6) for the rate variation among sites (Appendix 2). Its equivalence in nucleotide substitution rate is 0.0153 Ma^-1.

The phylogram of the genus Laparocerus has two basal branches originating in Node A (age 11.2 Ma): one gives rise to the Madeiran clade (Node M), and the other to the Canarian clade (Node C), which contains also species from the Selvagens and from Morocco. Both clades show sequential polytomies that group together a few or several subgenera with the lineage that splits the next. These solid polytomies represent basal star-shaped radiation events.

The Madeiran clade was presented and discussed in Machado et al. (2008) and Machado (2008a). The addition of the 12S rRNA gene modified slightly the topology and increased the support values, plus the dating performed, and a few extra OTUs justify including here the new version of the tree (Figure 4) and some comments.

The Madeiran clade splits sequentially in time starting with Node M (8.5 Ma), followed by Node N (7.2 Ma), Node O (6.3 Ma), and Node P (5.1 Ma), each giving rise to one or two monophyletic subclades recognised as subgenera and morphologically identifiable; six in total plus Anillobius (not included in the tree). Mean p-distance within subgenera ranks from 3.3% in Wollastonius to 7.0% in Atlantis (average 4.6%).

Subgenus Wollastonius, four small sized species (< 4 mm), showed a solid basal position (BPP 1) in the COII–16S rRNA phylogram; clustered with Atlantodes when adding the 12S rRNA, and recovers its basal position if the 28S rRNA is added. It is the oldest individual lineage (7.2Ma) in our tree and originated probably in Porto Santo, which was the only emerged island at that times, but it radiated within Madeira much more recently (3.3 Ma). Laparocerus waterhousei has been also recorded from Deserta Grande (Wollaston 1854, Roudier 1958) and without analysing specimens from that dismantled islet (we searched for it in 2000 and 2008 without success), it is impossible to infer whether the route followed was from Madeira to the Desertas or in the opposite direction.

Subgenera Laparocerus and Atlantodes cluster together at Node N (Fig. 4). We missed sequencing L. navicularis from Porto Santo, which is, from the morphological point of view, the sister species of L. colasi, endemic to Madeira. Its absence in the tree could explain the abnormal basal position of L. colasi within Atlantodes, a group that presumably started also in Porto Santo. Laparocerus (s. str.) is the only subgenus that is distributed in the whole archipelago, and the role of Ponta de São Lourenço, in the extreme east of Madeira, was already discussed in Machado et al. (2008: 423). The fauna of this peninsula in the eastern extreme of Madeira has greater affinity with its extending arc of islets of the Desertas and with Porto Santo (L. cryptus, L. schaumi), than with the rest of Madeira, supporting the hypothesis that the Ponta de São Lourenço was a separate islet that recently fused with Madeira along the valley of Machico.

Laparocerus undulatus clusters with Pseudatlantis but not with Atlantis, the subgenus to which it was attributed by Machado (2008a) based on phenetical consistency of the shape of male metatibiae (Fig. 2E) and the structure of aedeagus. This position is maintained when the nuclear 28S rRNA is added to the analysis. However, in the previously published COII-16S rRNA phylogram, it joined within Atlantis with low support (PPB 0.79), and when it was excluded from the analysis the PPB of Atlantis raised up to 0.99. If L. undulatus is also removed from our 3-gene analysis, the separate group of L. clavatus clusters with that of L. lamellipes, shaping Pseudatlantis and Atlantis as monophyletic subgenera. This ‘disturbing’ effect of L. undulatus is likely linked to an old hybrid origin.

The nominal species of Pseudatlantis Machado, 2008 have a very characteristic aedeagus structure similar to that of Atlantodes and Wollastonius, with the gonoporal poach inserted apically (Fig. 2L), not laterally, in what has to be assumed as a case of parallelism, not of homology. Moreover, they do not show sexual differences in the tibiae and have a more rounded body shape, etc, that justified its separate subgenus status. Consequently, we maintain the concept of Pseudatlantis and Atlantis as established in 2008, including in the latter subgenus the four now ‘outplaced’ species, with a MRCA at 5.1 Ma. This is the only case of paraphyly in subgenera of Laparocerus.

After our first Madeiran phylogram was published (Machado et al. 2008), we were able to sequence one specimen of the rare and blind tiny Anillobius solifuga Fauvel, 1907 but only for the COII and 16S rRNA, that being the reason it was excluded from the 3-gene phylogram. Nonetheless, taking into account these two markers it clearly falls within the equivalent of Node O (BPP 0.99) in parallel to Lichenophagus and Atlantis+Pseudatlantis, supporting the attribution of Anillobius as a subgenus of Laparocerus as proposed by Machado (2013). The former species is just a strongly modified Laparocerus adapted to endogean life.

The same problem was faced with Laparocerus hobbit because only COII and 16S rRNA sequences were available, and both differ only in one nucleotide each from those of L. lamellipes, questioning the validity of the former species. The peculiar characters of the tarsi highlighted in the description (Machado 2008a) may represent a hoxgen-mutation, but mitochondrial introgression or incomplete lineage sorting could as well be a plausible explanation for this case.

Laparocerus noctivagans and L. lauripotens are widespread and endemic to Madeira, variable in their morphology, and very difficult to separate. Wollaston described both species in 1854, synonymised them a few years later (Wollaston 1857) and re-established them in 1871 after a careful morphological examination. In our phylogram both species are clearly separated. In addition, specimens of L. lauripotens from the type locality (Curral das Romeiras), in the lee side of the island do not join with specimens from Santana, in the North (incomplete lineage sorting/introgression?). Also, L. noctivagans from the extreme west region of Madeira are more strongly striated and of black colour, representing perhaps a separate taxon (p-distance 1.6%). This group of Atlantis seems to be in active speciation and merit a deep and detailed phylogeographic analysis before taking further taxonomic decisions.

The Canarian clade of Laparocerus shows its first radiation event (Node C, Fig. 3) at ca. 7.7 Ma ago, shortly (0.8 Ma) after the Madeiran radiaton. The clade overall mean genetic distance of species is 7.3%, slightly lower than in the Madeiran clade (8.5%), and the mean distance between both clades is 11.0% (net distance 3.1%).

Subsequent radiations in the Canaries occurred at 5.3 Ma (Node D) and 4.6 Ma (Node E), each generating several monophyletic subclades, interpreted here as subgenera (Figure 3). The final outcome of these star-shape branching processes are 13 subgenera in the Canaries versus 7 in Madeira, in coherence with the lesser number of islands and overall minor surface of the latter archipelago. The average within subgenera genetic p-distance in the Canaries is 4.5%, ranking from 2.9% in Fernandezius to 5.7% in Purpuranius.

Basal lineages from Node C like Purpuranius and Aridotrox inhabit Fuerteventura and Lanzarote, which are the oldest islands, while younger subgenera like Belicarius and Bencomius (Node E) are restricted to the most distant and younger Western Canaries. There is a general shift from East to West, with some back-colonisations, which seems to follow the pattern of decreasing island ages and increasing distance to continental Africa associated with the prevailing hypothesis of a hot-spot origin for this archipelago (Carracedo 2011).

Laparocerus garretai from the Selvagens has a basal position (2.8 Ma) in one of the groups of species of the subclade ‘Pecoudius’, in agreement with the hypothesis of a Canarian origin for the extant Selvagens Islands’ biota (cf. Machado, 1992). These residual Macaronesian islets are very old in origin (29 Ma), but went through a large submerged phase in the Miocene/Oligocene (Geldmacher et al. 2001), the actual emerged land being younger, approximately 14 Ma.

Laparocerus susicus, the only known species from the continent (NW Morocco) join with Canarian endemics in the subgenus Aridotrox, and not in a basal position. This would indicate a back-colonisation event at nearly 1.2 Ma, justifying the name of Canarian clade used in this study.

Subclade ‘Purpuranius’ (Fig. 5). This subclade (BPP 0.94 (0.99)) of Node C was the first that radiated within the Canarian clade ca. 5.9 Ma ago, and it includes five rather distinct species endemic to Fuerteventura (L. fraterculus not sequenced) and L curvipes with one subspecies in Fuerteventura, one in Lanzarote, and another one in Tenerife (L. curvipes curvipes). This is the single case of a Laparocerus present simultaneously in the Eastern and the Western Canaries. More striking is its presence in the leeward side of Tenerife and not inhabiting the intermediate island of Gran Canaria, where it may have gone extinct or is waiting to be found. The genetic divergence among the three subspecies ranks from 0.6% (Fuerteventura- Lanzarote) to 1.6% (Fuerteventura-Tenerife).

The mean p-distance of 5.7% within this new subgenus is the highest recorded in the Canarian clade. Laparocerus calvus and L. longipennis are adelphotaxa and the oldest Canarian species (4.9 Ma), but they are quite different morphologically (with/without scales and hairs, female ovipositor, body shape, etc.). This may represent the outcome of a long lasting parallel anagenesis or, more likely, that extinction has been most severe in this group, only a relictual set of few species remaining.

Some species of this group dwell in xeric and semi-arid lowland with Chenopodiaceae and Launaea, while others are restricted to the sheer summits of the oldest and highest mountains of Fuerteventura (807 m) and Lanzarote (671 m) where remnants of the past thermo-sclerophyllous vegetation persist.

The new subgenus is described in the next section.

Subclade ‘Aridotrox’ (Fig. 5). This subclade of Node C (BPP 1) radiated ca. 4.8 Ma ago and clusters four species from the eastern Canary Islands and one from western Morocco, all of similar outlook and living in xerophilus mountain or flatland habitats.The Moroccan Laparocerus susicus clusters with the Canarian species L. rasus – L. xericola with high support (BPP 1) and does not take a basal position within the clade like L. dispar or L. colonnelli. This supports the hypothesis of a back-colonisation from the eastern Canaries to Morocco some 1.2 Ma ago.

The presence of preapically notched male metatibiae in L. susicus inexpectatus is likely to be related to it being a plesiomorphy in L. colonnellii and the group of L. rasus (species bearing this character are marked with an asterisk in Fig. 5). Nonetheless, there are contradictory associations of conspecific taxa within the L. rasus group that suggest incomplete lineage sorting in peripheral isolates and deserve a deeper phylogeographic analysis to clarify their relationships and dispersal routes. The group seems to have started in Lanzarote.

Mean p-distance within Aridotrox is 5.3%. The description of this new subgenus is in the next section.

Subclade ‘Pecoudius’ (Fig. 6). This branch of Node C is always fully consistent (BPP 1) regardless of which marker or combination of markers is used. It first radiated ca. 4.2 Ma ago and clusters five morphologically disparate groups of species which could merit a subgenus each. Nonetheless, we opted for naming the subclade after the subgenus Pecoudius Roudier, 1957, type species Laparocerus compactus Wollaston, 1864, and expand its concept until an expanded morphological and genetic study is conducted.

The evolution of this clade is likely to reflect the convulsive geological history of Gran Canaria, an island with an age of 14.6 Ma that underwent catastrophic volcanic activity between 3–3.5 Ma ago, and a later much milder re-activation of it (Carracedo 2011). The species groups seem to have started at an age (3.3–4.2 Ma) before or close to the volcanic paroxysm, but they radiated only after it ceased: 0.6–2.7 Ma. The ancestral species of these groups are possibly the survivors of the mass extinction, and the somewhat loose position of some of the extant taxa in the tree may be the effect of missing species that disappeared.

Laparocerus propinquus, L. fraudulentus, L. semipilosus, L. microphthalmus and L. crassirostris – each belonging to a different ‘Pecoudius’ group have apparently formed in the Tamadaba massif (videMachado 2012a), and have their respective sister species outside, representing a remarkable case of asynchronous parallel vicariance.

Species of the basal group, L. vicinus, L. propinquus, L. brunneus, live in high mountain scrubland and in the understory of the pine forest (e.g. Cistus, Artemisia). By its aedeagus structure, it may be somewhat related with the group of L. tessellatus, if any. The similarity of aedeagus is more clearly shown in the set of L. semipilosus, L. inconspectus, L. grayanus, L. fraudulentus, and L. hystricoides. They should constitute a second group on a morphological basis even though they do not cluster together in our tree. Species are distributed along the northern coast, in the central mountains, or almost all around the island (L. grayanus). The small monticolous L. hystricoides has preference for Cistus, but the other and much larger species are clearly more polyphagous (Kleinia, Rumex, Salvia, Periploca, Argyranthemum, etc.).

The group of L. compactus, or Pecoudius sensu stricto, is formed by five species that dwell in the leaf-litter and do not climb on the vegetation as is the dominant behaviour in Laparocerus. Their broadened and compact body, with mesothoracic inter-coxal process abnormally swollen and protruding, is apparently an adaptation to move between the vegetal debris. The extreme case is the dynamic boat-shaped form of L. eliasenae, similar to that of L. distortus in Madeira, clearly useful to push and navigate through the fallen leaves (convergence). The former species and L. sulcirostris live in the laurel forest or humid environments, while the other species dwell in the coastal spurge formations or subhumid sclerophyllous forest. The radiation of the group occurred in the northern side of Gran Canaria at ca. 1.2 Ma, and the last Pleistocenic volcanic activity phase may have played a role in isolating Laparocerus franzi in the peninsula of La Isleta, in the NE.

The other two species groups are not exclusive to Gran Canaria. The widely spread species complex of L. tessellatus has L. rugosivertex from Gran Canaria as sister species with BPP 0.93, and both join with Laparocerus garretai from the Selvagens Islands with a low BPP 0.81. All these species show the same type of aedeagus, and when adding the 28S rRNA to the analysis, the above referred support values raise up to BPP 0.98 and 0.91, respectively. The group is present in Gran Canaria (four species), Tenerife and the western Canaries (five species) except La Gomera, perhaps reflecting an original association to the pine forest plant community, which is missing in this island. However, they live in intermediate zones of the leeward and windward sides of the islands, in habitats where Adenocarpus, Chamaecytisus or Cistus grow, but also in the interior of the laurel forest. Faria et al. (2015) postulated its origin in Gran Canaria and presented evidence for multiple founding lineages and genetic admixture in their evolution, which is consistent with our results (see disjunct position of two L. auarita specimens). The complexity of this group shows adaptive and non-adaptive radiations that deserve more intense studies, particularly with the nuclear genes (videMachado 2016).

The last group, that of Laparocerus lepidopterus (BPP 1; 0.7 Ma), represents an outstanding case in the evolution of Laparocerus. It is formed by seven species and one of them, namely L. lepidopterus, is present in the western and central Canaries, with minor morphological variations per island. In the 3-gene phylogram, all ten OTUs sequenced split directly from the clustering node without any pairing. The overall mean p-distance within the group is extremely low: 1.2% (maximum 1.7%) and contrasts with the average of 4.5% within the rest of ‘Pecoudius’. With such a low divergence one would expect a very morphologically homogenous group, but this is only the case in the superspecies L. lepidopterus - pecoudi - separandus, which are large broad Laparocerus with silky hairs on the elytra living in the laurel forest and ecotones with the pine forest (mixed forest). In contrast, L. crassifrons has no hairs and a flattened body adapted to hide below the bark of the Canary pine whose needles it feeds on (it resembles more the pine weevil Brachyderes rugatus than a Laparocerus); secondly, L. lopezi and L. soniae – species apparently not directly related – are adapted to subterranean life, with reduced eyes, loss of vestiture, and narrower body (the p-distance between L. soniae and its epigean parent L. separandus is 0.7%); and lastly, L. mulagua from La Gomera looks like a half-sized and cylindrical L. lepidopterus with very prominent eyes. Such remarkable morphological differences within the group do not match with the reduced genetic divergence of the mitochondrial loci examined. When adding the 28S rRNA to the analysis, some OTUs cluster with high support values (PPB 0.95) but with little geographical logic. However, cases like L. mulagua joining L. separandus with 1 BPP reflects a direct link from Gran Canaria to La Gomera that has been found also in ground beetle vicariants of the genera Gomerina and Cymindis (Machado 1992), in the darkling beetle genus Pimelia (Contreras Díaz et al. 2003), or in bush-crickets of the genus Calliphona (Arnedo et al. 2007). If bad taxonomy is disregarded, there is no simple explanation (admixture, incomplete sorting, paralogy, convergence, etc.) for the high-morphological/low-divergence discrepancies observed in this group other than inadequate phylogenetic information, or that we are missing some unknown underlying genetic process that is worth investigation. The age estimates (0.3–0.7 Ma) obtained for this challenging group is probably unrealistic.

Subclade ‘Faycanius’ (Fig. 7).This subclade of Node C with BPP 1, has five species (two of them with subspecies) all endemic to Gran Canaria. The radiation of Faycanius is estimated in 1.9 Ma ago and is likely to have occurred also after the violent volcanic activity (3–3.5 Ma ago) of Roque Nublo Complex which sterilised great part of the island (Marrero 2004).

Species of Faycanius live on low plants and bushes (e.g. Artemisia, Argyranthemum) always in open land avoiding forest and shady areas. They are distributed over the whole island separated in different watersheds or by altitude, but not always.

Mean p-distance within Faycanius is 3.5%.

Subclade ‘Fortunotrox’ (Fig. 7). This subclade of Node D radiated ca. 5.1Ma in the western Canaries forming three well supported groups of species (each with one BPP): that of Laparocerus puncticollis (five spp), that of L. orone (three spp., two ssp.), and that of L. gracilis (five spp) exclusive to La Gomera, with the addition of L. chasnensis that is responsible for the low BPP 0.88 of the whole subclade (support increases to BBP 0.94 when the 28S rRNA is added). This latter species, endemic to Tenerife, is basal and the adelphotaxon of the other groups, thus representing the hypothetical source lineage. It is located at mid-elevation in the SW slope near the outcrop of the oldest island shield in Adeje (Roque del Conde), that remained uncovered during the Pleistocene eruptions of the Teide volcano complex (Carracedo and Pérez-Torrado 2013). The mean p-distance within Fortunotrox is 5.4%.

The group of Laparocerus orone lives in lowland xerophilous vegetation and in the intermediate zone dominated by Euphorbia and Kleinia, feeding mainly on Artemisia, Argyranthemum or Rumex when it comes closer to the forest zone. It seems to have generated three species in La Gomera following the radial watershed system in the N/NW/W section of the island, and from there the nominal species produced vicariants (subspecies) in El Hierro and La Palma; same habitat.

The group of L. gracilis is richer in species and exclusive to La Gomera, occupying the other half of the island (E and SE slopes), almost separated from the previous group. Species live in the same habitat characterised by dendroid spurges, but also in the remnants of the sclerophyllous forest. Segregation is likely to have taken place in parapatry. Remarkable is the case of L. gracilis and L. depressus, two sister species that live almost in contact in Barranco de La Villa, at different altitudes. Their morphological distinctiveness is not under discussion, larger and depressed body versus smaller and subcylindrical body, but their sequences are almost identical (p-distance 0.07%). We checked and disregarded mitochondrial introgression with nuclear markers. The mitochondrial genes analised simply seem to have not yet differentiated.

The group of L. indutus is linked to the sclerophyllous and the laurel forests in La Gomera and El Hierro, but the vicariant species from La Palma shifted to the high mountain scrub vegetation. Fortunotrox seems to have also undergone a mix of ecologically adaptive and vicariant based radiation. The abundance of L. confusus and related species in La Gomera could be an explanation for the absence of a representative of the L. tessellatus group in this island by competitive exclusion (similar size and ecological requirements), but this has not been the case at least with L. puncticollis and L. bimbache in El Hierro, where they overlap sharing habitats and often the same plant. In La Palma L. decipiens has apparently shifted to the high mountain domains (> 1800 m) avoiding in part the bulk area of L. auarita.

Subclade ‘Machadotrox’ (Fig. 7). This subclade of Node D (BPP 0.92/0.99) radiated ca. 4.7 Ma ago in the western Canaries giving rise to 13 species (incl. 2 subspecies). Half of them are adapted to subterranean life (lava tubes, mesocavernous shallow substratum (MSS), and soil): Laparocerus hypogeus and L. cavernarius (not sequenced) in El Hierro, L. oromii in La Gomera, and L. zarazagai, L. machadoi (not sequenced), L. iruene (not sequenced), and L. dacilae in La Palma. The epigean species are large forest Laparocerus that climb the vegetation to chew the leaves.

The subgenus Machadotrox was established by Alonso-Zarazaga and Lyal (1999) to replace the non-available name of Wollastonicerus Uyttenboogaart, 1937 (= Wollastonia Uyttenboogaart, 1936 non Heer, 1852) and was characterised by the male protibiae being enlarged at the apex both to the interior and to the exterior (fan-like) (Uyttenboogaart 1936, 1937). Our molecular analysis has revealed that this shape of the protibiae is present in several lineages and is not an autapomorphy of the subclade `Machadotrox’. Several species formerly considered Machadotrox show up, for instance, in Faycanius Machado, 2012, in Purpuranius subg. n., or in Bencomius subg. n. A good common and unique characteristic of the new concept of Machadotrox is the blade truncated form of the female hemisternites (Fig. 2C) with gonostyli placed distant from apex and being very short (not surpassing the hemisternite).

The La Palma endemic Machadotrox clustering in our tree (two epigean and two hypogean species) have obviously evolved in situ and therefore were selected as a trustful calibration point for our chronogram, not allowing their most common ancestor to be older than the estimated age of the island (1.72 Ma). The estimated age finally obtained for that node is 1.0 Ma.

The sequenced specimen of L. laevis from the north of La Palma (Pinar de Garafía) does not cluster with a conspecific specimen collected in the type locality and younger part (Barranco del Riachuelo) that joins with L. sculptus. This may represent one more case of incomplete lineage sorting or, perhaps, of poor taxonomy (specimens from the north are on average of a larger size).

The mean p-distance within subgenus Machadotrox is 4.8%.

Subclade ‘Mateuius’ (Fig. 8). This subclade (BPP 1) of Node D radiated ca. 3.3 Ma ago. Six species cluster together, but Laparocerus teselinde does not join the group, branching directly from the Node D. This outside position does not contradict its belonging to Mateuius, an attribution that cannot be disputed morphologically, as recently revised by Machado (2015). Except for L. auctus inhabiting El Hierro, all other species are endemic to La Gomera. They live camouflaged in the leaf-litter below scrubs from the lower semiarid zone (Euphorbia, Rubia, etc.) but L. merigensis and L. buccatrix (and possibly L. quadratus) dwell under trees and shrubs from the sclerophyllous forest or humid laurel forest, where they often find refuge in the dead leaves hanging from woody Sonchus species, or under the rossetes of rupicole plants. On the whole, Mateuius resemble the Fernandezius species which live in the same way and habitats, but on the other islands.

The mean p-distance within subgenus Mateuius is 5.0%.

Subclade ‘Amyntas’ (Fig. 8). This subclade (BPP 1) of Node D radiated at nearly the same time (2.5 Ma ago) as Fernandezius. It is equally distributed in Tenerife, La Palma, El Hierro, and absent in La Gomera, but present in Gran Canaria with only the species L. subnebulosus, which, due to its position in the tree, clearly represents a back-colonisation to this island, source of the parental lineage of Node D.

This group seems to have originated in Tenerife, inhabited by seven of the eleven species known. Two of them are shared with La Palma: L. fernandezi and L. tanausu. In the first case a potential introduction to La Palma cannot be disregarded (specimens scarce and located in anthropic sites), but in the second case we postulate a back-colonisation from La Palma to Tenerife with no apparent differentiation. Laparocerus tanausu is found in Anaga – much less abundant than in La Palma – where it replaces L. tibialis, and lives also in the offshore uninhabited Roque de Anaga. Laparocerus tibialis and L. tanausu are very difficult to distinguish phenetically and the latter should count as a cryptic species.

It is remarkable that L. incomptus, endemic to El Hierro, which is the youngest island (1.12 Ma), has an estimated age of 1.8 Ma and takes a basal position in the subsequent radiation that colonised several much older islands. A plausible explanation for this incongruence is a direct first colonisation of El Hierro from Tenerife, or that L. incomptus derived from the missing Amyntas of La Gomera that went extinct (or has not been yet discovered).

Amyntas are robust insects, and those species with black dull integument have a somewhat darkling beetle outlook. They are rather polyphagous and distributed in the xeric habitats of the islands leeward side (0–1000 m altitude) and in the more humid coastal zone of the windward side (0–400 m). Only Laparocerus bellus and L. arrochai feed on Jasminum odoratissimum, a bush species linked to the sub-humid sclerophyllous forest. They are sister species, one in Tenerife, the other one in La Palma.

The mean p-distance within Amyntas is 3.8%.

Subclade ‘Fernandezius’ (Fig. 8). This subclade (BPP 1) of Node D clusters all species of Fernandezius, a group with similar ecology and adaptations as Mateuius, but with a much younger and wider radiation event (2.2 Ma). It presumably started from Tenerife and reached the Western Canaries, except La Gomera, where its niche seems to be occupied by Mateuius. Their morphological convergence to ground life was misinterpreted by Roudier (1957) who described and attributed both subgenera to Lichenophagus from Madeira, at that time a separate genus from Laparocerus.

In a recent revision (Machado 2015) of the now three subgenera of Laparocerus, five new species of Fernandezius were described. However, their species status, thirteen species/subspecies in total, all single island endemics, was assigned by the author as an eclectic solution until further and more expanded molecular analyses clarify the real underlying relationships within this complex. We have included type-locality individuals for L. tesserula (Puerto de la Cruz) and L. subnodosus (Aguamansa) in our phylogram and additional specimens from Anaga at the NE of the same island in order to illustrate the problem. Different morphological taxa from a given region group among them and not with their corresponding taxon from other regions. This happens in Tenerife and in La Palma (with L. campestris). Such a repeated geographical influence requires a plausible explanation better than occasional peripheral isolate speciation, admixture in budding species, or a set of cryptic species. New advanced techniques for extracting information out of the nuclear genome, like RAD-seq (Restriction-site associated DNA sequencing), could be possibly the best solution to gain understanding of this species complex, which is also noticeable for the frequency of malformations observed, cases of plausible hybridisation, and the evidence of a successful mutation in L. impressicollis that may represent the initial stage of a new species formation in sympatry (Machado 2015).

The mean p-distance within Fernandezius is 2.9%; the lowest in all established subgenera.

Subclade E & incertae sedis (Fig. 10). The well supported Node E (BPP 1) branches from Node D (5.3 Ma) at ca. 4.6 Ma ago, with Tenerife or La Gomera as plausible origins. It groups four monophyletic subgenera that follow, and a few unplaced species, here considered taxonomically as incertae sedis. Judging from their morphology, some of these latter species could be related among them, but others not, and differences are remarkable. They represent lineages that failed in radiating as their sister groups did, or maybe they are the last survivors of once richer lineages that suffered partial extinction.

From the morphological point of view Laparocerus heres (with subspecies in La Gomera and Tenerife) is totally independent. It resembles somewhat a narrow Mateuius or Fernandezius (eyes placed at mid-face, dull integument, etc.) but it does not comply with their other diagnostic characters, and feeds on bushes like Chamaecytisus, Erica, Cistus or Adenocarpus. It denotes its own differentiated lineage.

Laparocerus depilis Roudier, 1957 is also a remarkable and isolated species, living in the central mixed forest of Tenerife, and unusually rare being a silvicole Laparocerus. By shape and tegument vestiture, it resembles somewhat a bald L. lepidopterus and was originally described as a subspecies of this species. However, it has a rather unique aedeagus with the penis strongly sclerotised and almost closed dorsally, ending in an abrupt double square step (in lateral view), with two little acute dorsal flaps on the wall on each side of the ostium. This combination of aedeagal features does not match with any other known Laparocerus.

The rest of ‘hanging’ species, basically from Tenerife and La Gomera, share some characters, like silky hairs on the elytra in the case of L. ellipticus and L. inflatus (both strict laurel forest insects), or pubescent coxae and thoracic sternites in L. pilosiventris, L. bacalladoi, and L. sanctaecrucis (all xerophylic species). Only the two latter species cluster as adelphotaxa, and show a p-distance divergence of 3.9%. They live allopatrically in the coastal lee side of Tenerife, one in the S and another in the NE, and their estimated separation point of 2.0 Ma greatly exceeds the age of the mega-landslide of Valle de Güímar (830.000 a f. Carracedo 2011) that could have divided the ancestral population.

Laparocerus ellipticus is the only not clearly differentiated species inhabiting five islands: common in La Palma and Tenerife, less common in La Gomera and El Hierro (not sequenced), and very rare in Gran Canaria, where only a few tiny areas of the original vast laurel forests remain. The estimated splitting time is 0.11- 0.17 Ma ago. However, considering that the maximum p-distance of 1.1% between specimens from Gran Canaria and from La Palma falls within the potential limits of local variation, a hypothetical recent introduction by anthropic activities cannot be disregarded. Sticks of laurel forest trees have been traditionally exported from La Palma to Tenerife and Gran Canaria for use in agriculture; La Gomera imported nursery forest plants from Tenerife, etc. This potential shuffle of populations is a question that merits being clarified with a more extensive analysis.

The mean p-distance within the set of incertae sedis species is 7.0%.

Subclade ‘Guanchotrox’ (Fig. 10). This subclade of Node E has a support of BPP 0.94 and 0.98 when the 28S rRNA gene is added. It radiated in four branches ca. 4.2 Ma ago, with a total of 27 species (several polytypic). It is the species richest subgenus of Laparocerus, because the subclade ‘Pecoudius’ is not considered a formal subgenus (pending revision). Until now, only the type species L. canariensis, from the upper regions of Tenerife (above 1,800 m altitude) and its vicariant on La Palma L. astralis were attributed to Guanchotrox.

The group of L. canariensis inhabits Tenerife (three species), Gran Canaria (three species), and La Palma (one species), all single-island endemics covering all main habitats: the lowland xerophilous scrub formation (e.g. Kleinia, Rubia), the laurel and sclerophyllous forest (e.g. Phyllis, Aeonium), the pinewood (e.g. Sideritis, Echium) and high mountain leguminous scrub (e.g. Spartocytisus, Adenocarpus). It is a mixed case of vicariant and adaptive radiation.

The group of L. tenellus has four species in Tenerife and one in La Palma that do not all cluster together, but it is morphologically consistent (small, rounded species with strongly mid-constricted rostrum). Laparocerus tenellus lives on the north-side of Tenerife, in forest areas, while the other three local endemics are parapatrically distributed in the western and southern lee side of the island. The vicariant of L. buenavistae on La Palma, L. palmensis, is spot-located in the same type of semi-arid habitat, feeding on Argyranthemum.

The group of L. obscurus is formed by the nominate species and L. dissimilis; both have radiated within Tenerife, from the coast to the high mountains, covering the whole island. This complex of species with their subspecies merits a detailed phylogeographic study to elucidate the main speciation regions within the largest of the Canary Islands.

The group of L. obtriangularis is the widest spread, with ten species covering all the central and western Canaries. They are mid-sized Laparocerus with a shiny or brassy integument. Some species dwell on the bushy vegetation in the forest, and others in the more open scrub formations at lower and intermediate altitudes.

The mean p-distance within the Guanchotrox Alonso-Zarazaga & Lyal, 1999 is 5.1%.

Subclade ‘Canariotrox’ (Fig. 9). This sufficiently supported (BPP 0.96) subclade of Node E (4.6 Ma) branched ca. 3.3 Ma ago into three species groups. It is distributed in the central and western Canaries with a total of 14 species, half of them in Tenerife where it is likely to have originated. We tried hard to find evidence of its existence in La Gomera without success. The absence of Canariotrox in this island surrounded by others where it lives is still a mystery, like the case of the L. tessellatus group or Amyntas.

The oldest group is that of Laparocerus occidentalis. They are all large weevils inhabiting the understory of the laurel and pine forests, and marginal vegetation. Outstanding is the very scarce genetic divergence (p-distance 0.2%) between L. aeneotinctus and L. femoralis, two allopatric species from La Palma, but the morphological differences are clear enough (normal/inflated profemora, etc.) to justify a species or subspecies status and invoke an incomplete lineage sorting of the markers analysed. The four Tenerife species are parapatric, L. rugosicollis, from central parts of the island, being associated with L. crassus from Anaga and not with L. aguiari, its vicariant from Teno. Laparocerus tauce lives at high altitude in the scrub formations, on the western flank of the island.

The groups of Laparocerus vestitus (open scrub land) and of L. inaequalis (humid laurel forests) are clearly morphologically related, but do not cluster together because of low support (BPP 0.87). Both are present in three islands: La Palma, Tenerife, and Gran Canaria, with morphologically very similar populations within each lineage, and deserve a thorough revision underpinned by a phylogeographical analysis.

Mean p-distance within Canariotrox Machado, subg. n. is 3.9%. Its description follows in the next section.

Subclade ‘Bencomius’ (Fig. 11). This subclade of Node E with BPP 0.94 cluster several species – mainly from Tenerife – originally attributed to Machadotrox due to the expanded apex of the male protibiae on both sides. They were removed from that subgenus by Machado (2013), and were pending a new placement.

Bencomius Machado, subg. n. branched ca. 3.1 Ma ago in two groups of species and the isolated Laparocerus edaphicus, which is the only known case of endogean adaptation (eyes and vestiture reduced, etc.) in this subgenus, likely from a common epigean ancestor.

The group of L. grossepunctatus (2.5 Ma) is formed by five species from Tenerife and a putative single vicariant, L. combrecitensis, in La Palma. It is remarkable that a specimen from the latter species collected in the type locality (Barranco del Riachuelo) in the middle of the island clusters (BPP 1) with species from the N and NE of Tenerife and not with another specimen, sharing the same morphology, from the north of La Palma (Roque Faro), which clusters (1 BPP) with L. escaleraorum, endemic to the Teno massif (NW Tenerife). The genetic divergence between the two specimens is 4.2%, which is indeed high, and even higher than 3.9% detected by Faria et al. (2015) within their Laparocerus sp 1 from La Palma (= L. auarita Machado, 2016). Multiple colonising events and admixture has been reported for the formation of both L. auarita and L. bimbache from El Hierro (Faria et al. 2015). This could be a plausible explanation for L. combrecitensis, but also the presence of an unnoticed cryptic species.

The group of L. scapularis is younger and radiated 1.5 Ma ago producing four species in Tenerife and single representatives in La Gomera and La Palma. The majority live in open leguminous scrub or on understory plants (e.g. Adenocarpus, Lotus) of pine woodlands. Only L. bolivari from Tenerife seems to be related to humid forests.

Mean p-distance within the Bencomius Machado, subg. n. is 4.1%. Its description follows in the next section.

Subclade ‘Belicarius’ (Fig. 11). This subclade of Node E with BPP 1 is the last to undergo radiation (estimated 2.8 Ma) among those subclades recognised as subgenera. It is formed by thirteen species distributed throughout the western and central Canary Islands. Some species inhabit two islands, and the concentration of eight species in La Gomera suggests an origin in this island, completing the general shift from East to West already commented upon. Nonetheless, a phylogeographic analysis with more individuals would be needed to confirm this hypothesis and clarify the jumping pattern among the islands (see discussion).

The set of Laparocerus exiguus, L. morrisi, and L. rotundatus arise independently from the basal ‘Belicarius’ node due to low support. At least, the first two species, which are very small and live among terophytes on the ground (>1000 m altitude), are clearly related from the morphological point of view, and replace each other in La Gomera and La Palma. Laparocerus rotundatus inhabit intermediate zone scrublands in the lee side of La Gomera, feeding on Rumex lunaria and Argyranthemum.

Laparocerus sanchezi and L. magnificus form a clear separate recent group (split 0.64 Ma ago) with subspecies in La Gomera and El Hierro in the first case. Laparocerus magnificus is widespread in the north-western parts of La Gomera, but a spot population has been found in the facing Teno massif on Tenerife, in similar habitat – remnants of the sclerophyllous forest – which are not prone to an introduction of anthropic origin. Specimens from both islands look morphologically identical, and the divergence of the individuals sequenced is 1.1%. We postulate a jump from La Gomera to Tenerife followed by no differentiation.

The group of L. mendicus radiated to five islands even more recently (0.8 Ma) and is well supported (BPP 1). Laparocerus robustus has three vicariant subspecies in the lofty elevations of Gran Canaria, L. longiclava is present in La Gomera and Tenerife with no apparent differentiation (p-divergence 0.8%), L. feloi and L. tarsalis are endemics to La Palma living in different sides of the island (W/E), and L. mendicus is exclusive to El Hierro. These five taxa have in common the presence of a small preapical tumefaction in the female elytra.

Laparocerus exophthalmus and L. oculatissimus are allopatric Gomeran endemics (wet forest/drier open scrubland), share very protruding eyes and are closely related in the tree. However, Laparocerus mateui mateui from La Gomera clusters with the previous pair (BPP 0.97) and Laparocerus mateui tuberosus clusters with L. mendicus, both from El Hierro (BBP 1). The species is well characterised by elytra beset of big protruding tubercles, a feature that is unique in climbing Laparocerus (a parallel case is known in Rhyncogonus tuberosus van Dyke, from Tahiti (videMachado 2007a), also a forest living weevil). Sequencing was repeated with the same and other individuals, with equal results. The nuclear 28S rRNA did not show any resolution power either when analysed separately or added to the mitochondrial matrix. However, the nuclear elongation factor (EFα) alone groups both subspecies with BPP 0.92, and then with L. mendicus (BPP 0.86), pointing to a mitochondrial introgression from either L. exophthalmus or L. oculatissimus into L. mateui mateui in La Gomera.

Mean p-distance within Belicarius Machado, subg. n. is 3.4%. Its description follows in the next section.

The limitations imposed to our phylogenetic study ‒ by having selected basically only one specimen per species or subspecies ‒ have greater relevance at the tips of the phylogenetic tree, preventing us from detecting cases of mitochondrial polyphyly and paraphyly, or from invoking the appropriate mechanism involved in those contradictory cases that showed up. Funk and Omland (2003) reported species-level paraphyly or polyphyly patterns in 26.5% of 2,319 assayed arthropod species. The phenomenon is taxonomically widespread, and our partial results – and those already registered by Faria et al. (2015) – suggest that it is indeed a common phenomenon in the explosive radiation of Laparocerus, and possibly more frequent than in continental taxa. However, our main concerns in this contribution are the basal nodes, where admixture, incomplete lineage sorting and introgressions should have less impact. Adding a third mitochondrial marker (12S rRNA) to our first Madeiran two-gene approach, has increased the support of all basal branches in the general phylogram, ratified with the addition of the nuclear 28S rRNA and its coherent outcome. Therefore, we can trust the first split and subsequent solid basal polytomies (PPB > 0.94) of nodes A-E and nodes M-P as being consistent. These polytomies come from collapsing very close unresolved branchings that ought to represent star-shaped radiations, an evolutionary process that should not be uncommon in oceanic archipelagos if several islands or vacant niches are available for colonisation.

When discussing paraphyly and endemic plant genera of oceanic islands, Stuessy et al. (2014) concluded that genera should be based on cohesiveness, distinctiveness, and monophyly in a wide sense (including paraphyly and holophyly). Laparocerus shows strict monophyly, and the overall mean p-distance (3-gene dataset) divergence in the Madeiran (8.6%) and Canarian clades (7.3%) is rather similar. More meaningful for cohesiveness, are the homogeneous divergence values within the named subclades of each clade; 4.8%, and 4.5%, respectively (see Appendix 2). There are no big differences between the two basal clades contradicting cohesiveness in Macronesian Laparocerus.

Being the Madeiran and the Canarian clades both monophyletic and sister groups, as they show up, there is an option for establishing two separate genera. However, there are no obvious features for distinguishing or characterising these genera morphologically, and the question of distinctiveness can be even more striking. Species of subgenera from Madeira and from the Canaries like Lichenophagus and Mateuius, or Pseudatlantis and Aridotrox, may look more similar among them, than species of sister subgenera from the same archipelago. This is because the explosive radiation of Laparocerus (237 species and subspecies) has been triggered both geographically (vicariance radiation) and ecologically (adaptive radiation), producing in the latter case derived forms adapted to different niches – underground environment, leaf-litter, cloud forest trees, high mountains, etc. – which are not free from adaptive morphological convergence. Similar cases have been reported for Canarian Nesotes in the Tenebrionidae (Rees et al. 2001a; b). In our results we have included many comments regarding ecology and distribution of species to illustrate these circumstances.

Stüben and Astrin (2010) presented the summarised phylogeny of the Atlantic Clade of Cryptorhynchinae (Curculionidae) using fewer but similar markers to ours (COI and 16S rRNA) and slightly different methodology, although comparable. This Macaronesian group of weevils, which are also flightless, is composed of 95 endemic and 2 introduced species (54 analysed), distributed in the Azores, Madeira, and the Canaries. They concluded that the Canarian and Madeiran archipelagos were colonised by continental Cryptorhynchinae at least seven times. Besides the two introduced species (Dichromacalles dromedarius and Echinoacalles franzi), four other lineages represent established genera (Calacalles, Onyxacalles, Echinodera, and Torneuma), and only one lineage underwent radiation forming a separate Macaronesian clade. This group of formerly Acalles is segregated in one new genus for Madeira (Madeiracalles) and nine genera and two subgenera within their Canarian clade. This profusion of taxa will reflect shifts to new habitats, like climbing forest trees (e.g. Dendroacalles, Silvacalles) or switching to different host-plant groups, even if there is only one representative of it (e.g. Echiumacalles, Ficusacalles, Pseudodichromacalles, or subg. Tolpiacalles). The argument is that in such endophytic larval dwelling weevils, the shift to a new host plant implies a parallel cladogenesis (Stüben 2000). That may be the case, even for the monotypic genera assuming they will success in the future, and a justification of splitting genera in Acalles, but it does not apply to Laparocerus, which have free-living larvae and are not host-plant specific.

More relevant for testing the single/multiple colonisation alternative hypothesis of the Macaronesian Laparocerus is the accuracy of the chronogram obtained in absence of fossils, particularly having acknowledged very disparate evolutionary rates among the different markers and between the different subclades/subgenera of Laparocerus. There are many inter-island and intra-island vicariant species that could have been chosen for calibration by taking the age of the island or of the disrupting geological event (e.g. lavaflow age) as a maximum time constraint; the problem lies in establishing a minimum time constraint. The shorter the timeframe, the smaller the uncertainty; but if it is too short, the risk of catching initial increased divergence arises (Penny 2005). Moreover, with vicariant sister species it is assumed that the split happened with or after the colonisation or geological event, and not before.

To circumvent these potential pitfalls, and after cross-validating different potential calibration points, we selected the radiation of Machadotrox in La Palma (maximum age 1.72 Ma), two epigean and five modified subterranean species (two sequenced), and tuned its minimum age constraint in function of not allowing the split of Node P ‘Atlantis-Pseudatlantis’ in Madeira to be older than the island age of 5.2 Ma (see methodology).

Stüben and Astrin (2010) used La Palma (2.0 Ma) and Madeira (4.8–5.2 Ma) as calibration constraints for their phylogeny of the Atlantic Cryptorhynchinae, and ended up with 11.6 Ma estimated for the split of Madeiran-Canarian Acalles-like genera. This age is very similar to the 11.2 Ma obtained for Laparocerus, whereas in the case of Cryptorhynchinae the Madeiran clade (Madeiracalles) with 7.3 Ma showed to be younger than the Canarian clade (several genera), with 10.9 Ma.

On the other hand, we have calculated an overall pairwise divergence rate per Ma, obtaining 3.1% for the combined set. These values are higher than the generally accepted standard of 2.34% from Brower (1994), based on uncorrected distances (our rate drops down to 2.3% if we use uncorrected distances), or the 2.1% used by Amorim et al. (2012) as the median value of the range of substitution rates they compiled for Coleoptera (0.7–3.5%). Pons et al. (2010) reported 2.6% just for protein-coding mtDNA.

Bromham and Woolfit (2004) explicitly tested if island radiations can speed up the molecular clock in a range of data sets, including Tarphius ironclad beetles, Pimelia darkling beetles and Dysdera spiders from the Canary Islands, but did not find support for that hypothesis. However, Papadopoulou et al. (2010) designed a careful study to estimate substitution rates in darkling beetles using the mid-Aegean trench separating of the western and eastern Aegean archipelagos (9–12 Ma), and obtained a divergence rate of 3.53% Ma^-1 for the COI gene and of 2.69% Ma^-1 when combined with the 16S rRNA gene. They used preferred partitioning scheme and a substitution model selected using Bayes mode, and also removed intraspecific divergence from the analysis as it may introduce higher divergence rates (Ho et al. 2005).

It seems that the rates obtained for different coleoptera or insect groups vary depending on the methodology used and the accuracy of calibration points. Some authors have applied directly the standard mutation rate of Brower, while most of the timetrees recently published for Macaronesia have used BEAST (Drummond et al. 2006), a software program that was originally designed for analysing population level genetic variations and possibly inflates age estimates when dealing only with species representatives and long timeframes. Calibration seems to have been problematic in many cases where very old colonisation events were forced to happen within a given island age, thus falling in tautology.

In this context, our divergence rate estimates can be considered as sound, thus giving additional confidence to the chronogram obtained.

In the previous paragraphs, a combined history of multi-vicariant and ecology-driven speciation and radiation events is envisaged for Laparocerus. Habitat shift is, among other factors, a process invoked for radiation success in oceanic islands, especially if niches are vacant (Juan et al. 2000, Emerson 2002). Otiorhynchus weevils are likely to represent the ecological counterpart of Laparocerus in neighbouring continental habitats, but no native species are known from the Canaries. Other potentially competing Entiminae are a few Sitona species that feed on leguminosae (Adenocarpus, Chamaecytisus, Cytisus) or Bituminaria (some are possibly introduced) and the large and endemic Herpisticus weevils (five species) present in semiarid habitats in all islands. More prone to sharing part of the feeding niche of Laparocerus are the darkling-beetle Nesotes (19 species and two subspecies), that climb the vegetation at night to feed, and have radiated similarly to Laparocerus, but with less success (Rees et al. 2001c). However, in these putative competitors the number of individuals per plant is far from being as high as it can be in Laparocerus (Machado and Aguiar 2005).

It is noticeable that if several sympatric Laparocerus species are beaten from the same plant – up to five species – they normally belong to different subgenera. A clear exception are three species of Purpuranius in Fuerteventura (to be discussed later), and it may happen occasionally in Bencomius (L. undatus and L. grossepunctatus) and in Belicarius. This empirical evidence suggests that several lineages underwent the same habitat shifts, and that competition among Laparocerus seems not to be a serious problem.

Original habitat. Excepting water and non-vegetated sand, lava and ash fields, Laparocerus are present in all natural habitat types of the islands: dune ecosystems (only in Madeira L. mendax and L. prainha), xeric Launaea steppes, semi-arid succulent shrub land, sclerophyllous forest, azonal cliff vegetation, laurisilva (laurel and heath forests), pine forest, high mountain scrublands and grasslands (including meadows in Madeira), and the subterranean environment. However, without knowing the ecology of the ancestral continental lineage and the paleo-environment of the islands in the late Miocene, it remains speculative to assess which was the original habitat.

The gradual closure of the Panama Straight (3.8–3.4 Ma ago), with its final closure nearly 2.5 Ma ago, modified the Gulf-Stream North Atlantic circulation which is responsible for the humid trade winds that arrive at Madeira and the Canaries, as well as for the colder sea-waters that contribute to ameliorate the climate extremes in these archipelagos (Haug and Tiedemann 1998, Molnar 2008). The trade-winds opened the ‘ecological window’ in Macaronesia for habitats like laurel forests. However, this climatic change also transformed the character of the vegetation in Africa and the Mediterranean (Blondel and Aronson 1999). Fossil gastropods found in Fuerteventura and dated in 4.8 Ma point towards a paleoclimate of oceanic-equatorial character (like in the Gulf of Guinea) much warmer than at present (Meco 2008). If Laparocerus arrival to the Canaries has been postulated at about 7.7 Ma ago, we simply do not know the habitat types available in the Canary Islands. It is easy to presume that they were not as dry as one tends to imagine judging from the present state. A recent reconstruction of the climatic conditions of Europe during the Tortonian (7.2–11.6 Ma) reveal mainly humid and subhumid summers, and no trace of a summer-dry Mediterranean climate even along the southern coasts (Quan et al. 2014).

The oldest Canarian Laparocerus group is Purpuranius (5.9 Ma) and its ecology could give some clues about the original habitat. Laparocerus longipennis dwells in arid scrubland on Chenopodiaceae (e.g. Salsola vermiculata), but L. calvus, L. maxorata and L. curvipes live sympatrically in the summits of the old Jandía massif (807 m), feeding mainly on the same woody plant, Asteriscus sericeus. There is a narrow habitat with relictual soil containing alophanes and maintained by the humidity of the tradewinds. The presence of isolated specimens of tree species known from the sclerophyllous forest and laurisilva (Rodríguez Delgado 2005) suggest that forests may have had a much larger distribution in the past, when the climate was more humid and the island much higher. An ancestral association of Laparocerus with some kind of forest cannot be disregarded.

Sclerophyllous forest. The sclerophyllous forest of the Canary Islands – also termed thermophyllous forest – is considered a species rich community of Mediterranean origin that occupied the transition zone between the succulent belt and the more humid laurel-forest, at 0–200 m and 500 m altitude in the windward side of the islands, and 300–500 and 700–900 m in the lee side (Fernández-Palacios et al. 2008). Unfortunately, we do not know what it looked like in natural conditions, as it has been almost completely destroyed by anthropic pressure.

Several Laparocerus species feed on plants and shrubs associated with this type of forest that survive scattered in cliffs and ravines (e.g. Convolvulus floridus, Carlina salicifolia, Bupleurum salicifolium, Maytenus canariensis, etc.). Good examples are Belicarius species in La Gomera (e.g. L. crotchi, L. subnodosus, L. gerodes, L. humeralis, L. magnificus) or L. (Amyntas) bellus from Tenerife and its vicariant in La Palma, L. arrochai, which feed on Jasminum odoratissimum, the only apparent case of monophagy recorded in Laparocerus (Machado 2003, 2009b) . However, the sclerophyllous forests are supposed to have emerged in the Mediterranean under hot dry summers and cool wet winters approximately 3.4 Ma ago (Jiménez-Moreno et al. 2010), under the influence of the climate change triggered by the closure of the Straight of Panama and well after the Messinnian Salinity Crisis in the late Miocene. In addition, Amyntas and Belicarius are recent lineages (< 2.8 Ma), and the number of characteristic plant components of this community that are not accepted as food by Laparocerus is much larger and significative: Bosea, Dracaena, Juniperus, Lavatera, Marcetella, Olea, Phoenix, Pistacia, Rhamnus, Ruta, etc. If there was an original association of Laparocerus with forest habitat, it probably was of a different type, and their presence in the sclerophyllous forest reflects a habitat shift or a survival strategy in the case of Fuerteventura.

Laurisilva. The Macaronesian laurisilva, traditionally termed ‘monteverde’ in the Canaries, covers laurel forests, heath forests and all their variations, but is generally referred to as laurel forest, in a wide sense. Stüben and Germann (2005:48) postulated for Cryptorhynchinae a start in the conspicuously more arid habitats of the coastal succulent belt, and continued much later (Dendroacalles and Silvacalles) to the shady and moist laurel forests (= laurisilva). Several lineages of Laparocerus have colonised the laurisilva (Bencomius, Guanchotrox, Machadotrox, Fernandezius, some groups of Pecoudius, etc.) also within the timeframe of the ecological window generated by the trade-winds. It has been confirmed that laurisilva has a Plio-Pleistocene origin, with a few older species from the Upper Miocene (Kondraskov et al. 2015). It is not a relict sample of the Tertiary flora as has been largely thought (Ciferri 1962, Axelrod 1975). Laparocerus can be found high in the canopy of trees (Laurus, Persea, Ocotea, Prunnus, Myrica, etc.), in bushes of all sizes, on the lower plants that grow in the shade or more exposed on the cliffs and openings (Aeonium, Cedronella, Geranium, Hypericum reflexum, Phyllis, Ranunculus, Rubus, Senecio, etc.). Very few of the twenty laurisilva tree species are left aside by Laparocerus (e.g. Arbutus, Visnea). The conquest of this habitat has been almost complete.

Pine forest. Laparocerus can be found only occasionally on Pinus canariensis (e.g. L. combrecitensis, L. tenuepunctatus), with a remarkable exception: Laparocerus crassirostris from Gran Canaria feeds almost exclusively on Pinus canariensis (also on Cistus, but less) and its flattened body resembles that of the Canary pine weevil Brachyderes rugatus, adapted to hide in the cracks of the bark. Both species coexist in the Tamadaba massif and may share the same tree.

The habitat shift to the pine forest has focused on the species growing in the understory (Chamaecytisus proliferus, Lotus spp., Cistus monspeliensis, Cistus symphytifolius, Adenocarpus foliolosus, etc.) where Laparocerus can be really frequent and extraordinarily abundant. There are no records of native pine trees, living or fossil, in the eastern Canary Islands, so it is likely that the habitat shift started early in Gran Canaria.

Subsurface habitat. The subterranean environment has been colonised by one specific lineage in Madeira (Anillobius) with two endogean single island endemics (not included in the phylogram), and by three lineages in the Canary Islands, which include epigean counterparts. The basal position of L. oromii from La Gomera in Machadotrox and of L. edaphicus from Tenerife in Bencomius suggest that the dispersal to the underground environment happened early after colonisation of the island. The isolated position of these species would represent the result of a long-lasting anagenetic evolution, or that we are missing other derived subterranean species, either because they have not been yet discovered, or because they went extinct. The concentration of hypogean forms in young islands, La Palma (four spp.) and El Hierro (two spp.), which are mainly large forms adapted to caves and the MSS or mesocavernous shallow substratum (Oromí et al. 1986) suggest that radiation happens in the early stages of island building, when this micro-cave environment is open for life before it gets filled by earth as the island becomes older and erosion prevails. An exception to this rule is L. idafe (not sequenced) from La Palma, which is small and has the typical subcylindrical body of edaphic forms (Villani et al. 1999), that are normally present in mature soil of old islands (Tenerife, La Gomera, Gran Canaria, Madeira and Porto Santo). Subterranean species have been marked with an asterisk in the phylograms, but we have been able to sequence only half of the 14 species known.

Subterranean insect life in volcanic terrains is much richer than originally attributed to oceanic islands. Far from being sterilised, the number of species found in lava tubes and the MSS in the Canary Islands is increasing constantly (Oromí 2010), despite the difficulties to prospect in these confined environments. If the hypothesis of an initial boom and bust speciation phase after island emergence is true, the volcanic mesocavernous shallow substratum (MSS) would be an ideal empty niche for occupation, and Laparocerus have shown adaptation capacity for doing so. Indeed, the presence of big, MSS/cave dwelling species only in El Hierro/La Palma and the presence of small, soil-dwelling species in older islands (plus one in La Palma) suggest a substitution of endogean species by hypogean ones as the islands get older. We are possibly facing a richer initial hypogean fauna, and the corollary hypothesis of a greater progressive extinction of Laparocerus under the ground rather than above the ground seems plausible, albeit hard to test.

Ecological diversity. Island size, distance from continent, island age and other factors have been traditionally analysed in shaping oceanic island faunas since the McArthur and Wilson (1967) seminal work on island biogeography. We have found a clear correlation (multiple R = 0.94, R² = 0.88) only between the ecological richness of the island and the number of Laparocerus species living in it. The slight deviations observed in Figure 15 are congruent with circumstances previously described, such as the size reduction of the island of La Gomera, or the plausible extinctions of species in Lanzarote and Fuerteventura.

This data suggest that, despite the historical background of colonisation and speciation on each island, it is ecological diversity that controls the extant local fauna.

Species are conceptually easy to understand as unitary lineages in evolution, but difficulties arise when it comes to establishing boundaries recognising species in practice. Homologies, cryptic species and budding species are usual nightmares of taxonomists, and Laparocerus covers the whole panorama. Molecular techniques have been welcomed to assist in systematic work, especially for unveiling real relationships and by offering quantitative data to fix taxonomic criteria, despite that a common ancestor is not the same for all molecular markers, especially if there was some kind of hybridisation during the segregation process. Species trees and gene trees are rarely identical at leaf levels (Nichols 2001).

It would have been practical to obtain some guiding ranks of genetic p-divergence for delimiting subspecies or species, but the reality in Laparocerus is more complex. Divergence may be very high in a species formed by admixture (e.g. 3.9 % in L. auarita); cryptic species may differ strongly in their divergence (e.g. 7.8% in L. cryptus/L. morio or 5.0% in L. tibialis/L. tanausu); conspecific specimens coming from two islands may show divergence reflecting only anagenetic evolution (0.2–1.9%); morphologically distinct species can show almost no divergence (e.g. 0.7% in L. depressus/L. gracilis); or the maximum divergence within species of a given group (e.g. 1.7% in the group of L. lepidopterus) may be less than intraspecific divergence or divergence in sister-species of other groups. There is no common criterion. Genetic divergence can assist taxonomic decisions about species boundaries in Laparocerus at most within a given group, and only in the integral context of morphological, biogeographical, and ecological information.

The species swarm of 236 extant Laparocerus in Macaronesia is the result of a blend of adaptive and non-adaptive evolution in old oceanic archipelagos with plenty of environmentally dissected islands with a dynamic and complex volcanic history of construction and deconstruction. Contingencies like sterilising eruptions and mega-landslides shall have played a decisive role in segregation, promoting allopatric and peripheral isolate intra-island speciation, as well as in punctuated dispersal. Oceanic islands are indeed species producing machines (cf. Rosenzweig 1995).

An obligatory question is: why has Laparocerus by large the record of island endemics in Macaronesia and not other groups that are almost as old, having evolved in the same scenario and are also flightless? The next in the list are Napaeus gastropods with 74 species (updated), Dysdera spiders with 72 species (including 16 pending description, Oromí pers. com., Crespo 2013), Tarphius beetles with 57 species, Dolichoiulus millipedes with 56 species, etc. (Fernández-Palacios 2011).

It has been argued, that some taxa show a greater inherent degree of genetic or morphological plasticity than others, or that they possess traits related to their breeding systems that favour rapid evolutionary change on islands (Whittaker and Fernández-Palacios 2007). Indeed, insects are known to be a successful group in evolution of life, Curculionidae within insects, and Entiminae within Curculionidae (Oberprieler et al. 2007). Entimine weevils are not prone to chemically mediated co-adaptations with new host plants as it occurs in other weevils with endophytic larvae, but they may have some pre-adaptive capacity to colonise isolated oceanic archipelagos, Laparocerus being the most striking case (237 species). Parallel but less spectacular radiative speciation is reported for some 130 species of Rhyncogonus in the Pacific (Machado 2007a, Gillespie et al. 2008), and nearly 100 species of Cratopus (mostly winged!) in the Indian Ocean (Kitson et al. 2013).

If oceanic islands have been traditionally considered as laboratories of evolution, Laparocerus will become the ideal guinea-pig for broadening in speciation processes of all kinds, and for studying the role that ‘geological turbulence’ has played in vicariant speciation or massive dispersal. They are flightless, monophyletic, have many endemic species in many islands, and are easy to find. Working with such a group is like getting a picture of nature with more pixels. We hope that several highlighted cases in this discussion (e.g. Atlantis, Aridotrox, Fenandezius, etc.) become stimuli for more intensive sampling and further phylogeographic research in this group. The answer to why there are so many Laparocerus is more or less clear; the how is now the challenge.

We are confident that in the near future Laparocerus will merit sharing the podium with Darwin´s finches or Drosophila in the studies of island evolution.