Taxonomic Synopsis of the Ponto-Mediterranean Ants of Temnothorax nylanderi Species-Group

In the current revisionary work, the Temnothorax nylanderi species-group of myrmicine ants is characterized. Eighteen species belonging to this group in the Ponto-Mediterranean region are described or redefined based on an integrative approach that combines exploratory analyses of morphometric data and of a 658bp fragment of the mitochondrial gene for the cytochrome c oxidase subunit I (CO I). The species group is subdivided into five species complexes: T. angustifrons complex, T. lichtensteini complex, T. nylanderi complex, T. parvulus complex, T. sordidulus complex, and two species, T. angulinodis sp. n. and T. flavicornis (Emery, 1870) form their own lineages. We describe seven new species (T. angulinodis sp. n., T. angustifrons sp. n., T. ariadnae sp. n., T. helenae sp. n., T. lucidus sp. n., T. similis sp. n., T. subtilis sp. n.), raise T. tergestinus (FINZI, 1928) stat.n. to species level, and propose a new junior synonymy for T. saxonicus (SEIFERT, 1995) syn.n. (junior synonym of T. tergestinus). We describe the worker caste and provide high quality images and distributional maps for all eighteen species. Furthermore, we provide a decision tree as an alternative identification key that visually gives an overview of this species-group. We make the first application to Formicidae of the Semantic Phenotype approach that has been used in previous taxonomic revisions.

Thanks to intensive studies during recent years [22], [23], [24], [25] the European ant fauna is relatively well explored. Temnothorax has long been the focus of ant taxonomy, and earlier taxonomic studies are available for particular geographic areas [26], [27], [28], [29]. The Temnothorax nylanderi species-group is particularly diverse in the Ponto-Mediterranean region (i.e., the basin of the Mediterranean Sea and Black Sea) and its systematics is still very challenging. Those few species that range as far northwest as England and Central Europe are well understood and adequately described by modern taxonomic methodology [30], [31], [32], [33]. In North Africa and in part the Iberian Peninsula, the T. nylanderi species-group is replaced by the Temnothorax laurae species-group, its vicariant sister group [34], [35].
The high diversity of the T. nylanderi species-group in the Ponto-Mediterranean region is coupled with a large number of cryptic species [36], [33], whose broadly overlapping morphological characters hamper the definition of species boundaries based on conventional techniques. Therefore, in the present study we delineate species boundaries in an integrative fashion [37], using multiple lines of evidence, including Numeric Morphology Based Alpha Taxonomy (NUMOBAT) and molecular analyses.
In our study we initially recognized morphological patterns using the exploratory data analysis tool NC-clustering [38], which allows inference of the boundaries of morphological entities without prior hypotheses about the number of clusters or the classification of a particular sample. The performance of such exploratory data analysis tools allowed formulation of sound hypotheses about formerly unrecognized morphologically cryptic taxa in ants [39], [33]. The initially inferred patterns are subsequently tested by confirmatory data analyses. To strengthen the consistency of conclusions by the complex morphometric analyses, we in parallel analyzed a fragment of the mitochondrial gene for the cytochrome c oxidase subunit I (CO I) in specimens from all identified morphological clusters.
Our data reveal cryptic diversity in this species-group of the genus Temnothorax and help to better understand the biogeographic patterns of this genus in the Ponto-Mediterranean region.

Materials and Methods
In the present study, we recorded 22 continuous morphometric traits in 1693 worker individuals belonging to 526 nest samples. Specimens for the present study were borrowed from public and private collections, see the list of institutions below. Our study did not involve endangered or protected species. Label information of type material is given for each taxon in the section type material investigated. Non-type material that was morphometrically examined in this revision is listed in S1 Table. Samples displayed in the dendrogram are characterized by samplespecific abbreviations (e.g., GRE:Levidi-10S-20000427-123) generated from the original label information as follows: A three-letter country code (in capital letters separated from the following part by a colon), the nearest settlement given in the label (separated from the following parts by hyphens), the distance and direction from it, the sampling date in alpha-numeric format and a final unique field sample identifier.
All images presented are available online on AntWeb (http://www.antweb.org) and can be uniquely identified by their specimen-level codes affixed to each pin. Information about the taxonomic history of the taxa redefined below is based on B. Bolton in AntCat (http://www. antcat.org). Distribution maps were generated by QGIS 2.4.0 software [40]. For statistical comparisons of morphological data we used R [41]. Synopsis of the Ponto-Mediterranean Temnothorax nylanderi Group PL: total petiole length measured in dorsal view; distance between the dorsalmost point of caudal petiolar margin and the dorsalmost point of anterior petiolar peduncle at the transversal level of its strongest constriction. Positioning of petiole as in NOdL (Fig 1F).
PoOC: postocular distance. Use a cross-scaled ocular micrometer and adjust the head to the measuring position of CL. Caudal measuring point: median occipital margin; frontal measuring point: median head at the level of the posterior eye margin (Fig 1A).
PPH: maximum height of the postpetiole in lateral view measured perpendicularly to a line defined by the linear section of the segment border between dorsal and ventral petiolar sclerite ( Fig 1D).
PPL: maximum length of the postpetiole measured in lateral view perpendicular to the straight section of lateral postpetiolar margin (Fig 1E).
PPW: maximum width of postpetiole ( Fig 1C). SL: maximum straight line scape length excluding the articular condyle ( Fig 1A). SPL: minimum distance between the center of propodeal spiracle and the subspinal excavation measured in lateral view (i.e. the same view that is applied to measure ML). Note: in lateral view propodeal spiracle and the caudal margin of propodeal declivity might not be in the same focal level, hence slight adjust might be necessary while measuring SPL between the two endpoints ( Fig 1E).
SPBA: the smallest distance of the lateral margins of the spines at their base. This should be measured in dorsofrontal view, since the wider parts of the ventral propodeum do not interfere with the measurement in this position. If the lateral margins of spines diverge continuously from the tip to the base, a smallest distance at base is not defined. In this case, SPBA is measured at the level of the bottom of the interspinal meniscus ( Fig 1C).
SPST: distance between the center of propodeal stigma and spine tip. The stigma centre refers to the midpoint defined by the outer cuticular ring but not to the centre of real stigma opening that may be positioned eccentrically (Fig 1E).
SPTI: the distance of spine tips in dorsal view; if spine tips are rounded or truncated, the centers of spine tips are taken as reference points (Fig 1C).
SPWI: maximum distance between outer margins of spines; measured in same position as SPBA (Fig 1C).
The deviation of propodeal spines from longitudinal mesosomal axis (i.e. angle between the spines and the mesosomal axis) is often of high diagnostic value within the Temnothorax nylanderi species-group. Longitudinal mesosomal axis is defined by Seifert et al. [39] as follows: in lateral view as straight line from the center of propodeal lobe to the border point between anterior pronotal shield and propleuron.
Hereby we propose two new template EQ expression as an expansion of the semantic statement types described by Balhoff et al. [49] and Mikó et al. [50]: Angle between anatomical lines Angles between anatomical lines are often used in taxonomic treatments of Formicidae: NL: Dorsal profile of petiolar node contour line angle value to frontal profile of petiole contour line in lateral view: 72-82°" We expressed angle value between anatomical lines using PATO's angle and UO's degree with the combination of our previously proposed relative and absolute measurement phenotypes: SP: 'has part' some ('dorsal petiolar profile contour line' and (bearer_of some (angle and ('is quality measured as' some (('has measurement unit label' value degree) and ('has measurement value' some (float[> = 72.0f] and float[< = 82.0f])))) and (towards some 'anterior profile of petiolar node contour line')))) Coloration We present color hue and intensity as separated entity attributes: NL: Body color: brown SP: 'has part' some (body and ('bearer of' some brown)) and NL: Body color pattern: mesosoma, antenna and legs excluding femora, waist and anterior region of 1st gastral tergite lighter than head, femora and posterior region of gaster.

Morphometric hypothesis formation and testing
The complexity of the Temnothorax nylanderi species-group required a complex work flow for morphometric pattern recognition. Combined stepwise analyses were done to achieve the best results. Iterated confirmatory analyses (Steps 2 and 3) helped to support the position of various samples predicted by prior steps (Step 1).
Step 1: Exploratory analyses of morphometric data by NC clustering algorithms. We first used an exploratory data analysis of continuous morphometric traits to assess the occurrence of cryptic diversity within the East-Mediterranean Temnothorax nylanderi species-group. A hierarchical, agglomerative nesting method, Nest Centroid Clustering (NC clustering) was employed to reveal complex patterns in the complete dataset [38]. NC clustering allows assessing the number of morphologically meaningful clusters (i.e., taxa) and disclosing the structure of the whole species-group without prior hypotheses. Initial conclusions about morphological patterns were drawn from the output of NC-UPGMA clustering. The script for this analysis, written in R, is freely available at: http://sourceforge.net/projects/agnesclustering/.
Step 2: Deriving hypotheses from NC-clustering and checking these by a confirmatory Linear Discriminant Analysis (LDA). To increase the reliability of the recognized patterns we applied parallel runs of NC-UPGMA and NC-Ward clustering. In a confirmative LDA, run on the level of individual workers, classification hypotheses were imposed for all samples congruently classified by both methods. Wild-card settings (i.e., no hypothesis imposed) were given to those samples whose classifications differed between the methods. The confirmative LDA was run as an iterative process and sample means of posterior probabilities formed the basis for the finally accepted classification. No samples with ambiguous position remained (for details see [38], and all taxa in this revision meet the criteria of the pragmatic species hypothesis [51]. The model is validated by Leave One Out Cross Validation. Step 3: Application of confirmatory Linear Discriminant Analyses (LDA) on separate species complexes. After prior species hypotheses formulated by NC-clustering (Step 1) had been confirmed by cumulative LDA (Step 2), we applied an a posteriori step of Linear Discriminant Analysis on every species-complex with more than one species. Posterior analysis allows confirmation of a priori groups suggested by NC-clustering and replacing erroneously allocated samples. Taxa within species-complexes can thus be visualized in two dimensional scatterplots. The second purpose of this ultimate statistical step is to provide readers with the simplest numeric function for species determination with an acceptably low (< 5%) error rate. Researchers who determine ants on a daily basis particularly benefit from the reduction of the number of characters in numeric determination tools.
Cross Validation Decision Tree. The key to worker caste was generated by a Cross Validation Decision Tree algorithm through the package "rpart" [52] in R [41] based on nest sample means of body size ratios. Decision trees yield a dichotomous structure in which each node represents a test of continuous morphometric traits (decision taken after computing all attributes). The applicability of attributes for each node generated by the model was validated by Leave One Out Cross Validation. The reliability of characters for each node was also tested manually by t-tests.

Multivariate Analyses of Numeric Morphology-Hypothesis formation and Confirmatory analyses
An exploratory NC-Clustering revealed the existence of eighteen clearly separated clusters within the available specimens of Ponto-Caspian populations of the Temnothorax nylanderi species-group (Fig 2). This pattern was confirmed by a cumulative Linear Discriminant Analysis. The confirmatory analyses correctly classified 94% of the 1693 individuals. Validation by Leave One Out Cross Validation yielded a 93% classification success rate, which can be considered acceptable in such a complex, multidimensional discriminant space. The vast majority of species showed a perfect or nearly perfect separation from other morphospecies treated in this revision: T. angulinodis  Synopsis of the Ponto-Mediterranean Temnothorax nylanderi Group A few others yielded a lower classification success: T. crasecundus (89.6%), T. crassispinus (90.4%), T. helenae sp.n. (91.3%), T. nylanderi (90%), and T. tergestinus (89.9%). This is probably not surprising, given the the high morphological similarity among some of the closely related species. Some taxa could be separated convincingly only by confirmatory LDA of nest sample means: 98.5% of the 528 nest samples were correctly classified by this analysis. Again a few species, T. crasecundus (98.2%), T. crassispinus (97.47), T. laconicus (93.8%), T. lichtensteini (98.8%), T. subtilis sp.n. (98.2%) and T. tergestinus (94.0%) attained somewhat lower rate, but nest samples of all other species treated in this revision were correctly classified (100%).
The species hypothesis generated by the NC-clustering, with 18 existing morphospecies, was thereby confirmed by LDA and can be considered as the final species hypothesis in this revision.

Molecular data
Partial sequences of the mitochondrial gene CO I were available for specimens from most of the clusters recognized by morphometry. A Bayesian tree (Fig 3) fully supported most of the morphospecies well classified by LDA of nest sample means with Bayesian posterior probabilities close to 1 (T. ariadnae sp.n., T. artvinensis, T. flavicornis, T. helenae sp.n., T. lucidus sp.n., T. nylanderi, T. parvulus, T. schoedli, T. similis sp.n,). In addition, the molecular data also substantiated T. tergestinus, less well supported by morphometry, as a separate lineage.
Molecular analysis reliably distinguished neither T. laconicus and T. lichtensteini nor T. crasecundus and T. crassispinus. This matches the slightly lower support of these taxa in the morphological analysis. The various specimens of T. subtilis, though with high classification rate by morphometry, did not form a distinct cluster in the molecular analysis. The placements of a worker from Bistrits, Bulgaria-morphologically identified as T. crassispinus-amidst specimens of T. lichtensteini and that of one worker of T. subtilis from Akseki, Turkey, near T. schoedli and T. flavicornis remains unexplained and might be due to an accidental exchange of specimens.
It was not aim of the study to provide a robust phylogeny of the here investigated taxa. The basal nodes of the tree show only very weak support. Nevertheless, our analysis supports the T. nylanderi complex, the T. parvulus complex, and the T. lichtensteini complex as monophyletic groups. As mentioned above, the two taxa of the T. lichtensteini complex clustered together but could not be distinguished reliably.
The Temnothorax angustifrons complex is not supported by the molecular data. Instead, T. lucidus sp.n. forms a weakly supported outgroup to the T. nylanderi complex, and T. similis sp. n. falls into the T. parvulus cluster. Similarly, the taxa of the T. sordidulus complex are scattered throughout the whole phylogenetic tree.  (Bernard, 1957) Temnothorax tergestinus  stat.n. Temnothorax saxonicus (Seifert, 1995) syn.n.

Key to worker caste
The following dichotomous identification key for the worker caste of the Temnothorax nylanderi species-group was generated by Decision Tree in R [41] based on continuous morphometric traits. It provides an easy-to-use identification guide, which is also displayed graphically in Fig 4. The reliability of characters has been tested and the percentages at each node indicate the classification success. The key is based on nest sample means (which require the investigation of 2 to 3 individuals from each nest). However, because the range of morphometric traits calculated from nest sample means is more or less identical with the 5-95% percentile range of the pool of single individuals, the key is also expected to work on single specimens with high probability (>95%). Additional information is available in differential diagnoses, which helps to identify the most problematic cases.  Decision Tree, an easy-to-use graphical display of a dichotomous identification key of workers of Temnothorax nylanderi species-group. The classification success in percentages is given in brackets under each particular node. Means and minimum, maximum values of morphometric ratios are given for each branch in red. At two positions (nodes) the originally suggested continuous morphometric traits were replaced by more reliable data about geographic distribution.

Descriptions and redefinitions of species and species-complexes
The description and redefinition of species belonging to the Ponto-Mediterranean Temnothorax nylanderi species-group can be found following the diagnoses of species complexes to which they belong. Species complexes are defined based on morphological similarities, but not fully supported by molecular (mtDNA) data. Patterns formerly recognized by NC-clustering were tested by confirmatory Linear Discriminant Analysis (Steps 2 and 3). The results of complex-wise analyses (Step 3) are displayed in scatterplots for each species-complex defined below.
Recent taxonomic papers on Temnothorax taxa [33], [48], [39] were considered in preparing descriptions and redefinitions of the species treated in this work. The varying degree of inclination of pilosity follows [53]. Definitions of surface sculpturing are linked to [54].
Definitions of species complexes as well as descriptions of species belonging to a certain complex are given in alphabetic order.
Body size is given in μm and morphometric ratios are given as means. In the differential diagnosis, we give the means of discriminant scores, minimum and maximum, and 5-95% percentiles to prevent that abnormal cases are misidentified. Nest sample means of ratios of morphometric variables are given in S4 Table for each species with standard deviation in italics followed by minimum and maximum in parentheses. In the diagnosis of species complexes, minimum and maximum values of body ratios are indicated.

Diagnosis of Temnothorax angulinodis species-complex
Workers of the Temnothorax angulinodis species-complex can be distinguished from those of other complexes treated in this revision by the combination of the following salient features: brown to black color; longer than broad head (CL/CWb [1.171, 1.222]), sculpture of head dorsum shiny: with inconspicuously areolate ground sculpture combined with parallel costulate main sculpture ( Fig 5A); petiolar node in lateral view with a straight or weakly concave frontal profile meeting dorso-caudal plate in an acute angle (72-82°) with a sharp ridge, in dorsal view appearing as conspicuous semicircular anterior-lateral rim ( Fig 5B); long propodeal spines (SPST/CS [0.332, 0.369]), deviating from longitudinal axis of mesosoma by 32-38°( Fig 5C).
This peculiar species-complex consists of a single species, Temnothorax angulinodis sp. n., which is only known from the Peloponnese peninsula. This species is clearly defined by NCclustering and corroborated by LDA (Fig 2, S5 Table).
Type material investigated. Holotype worker labelled: The list of 15 non-type individuals belonging to 5 nest samples of other material investigated is given in S1 Table. Worker (Fig 5A-5C, S1 Table, S4 Table). Body color: brown. Body color pattern: mesosoma, antenna and legs excluding femora, waist and anterior region of 1st gastral tergite lighter than head, femora and posterior region of gaster. Absolute cephalic size: 594-657 μm Differential diagnosis. Due to the unique combination of the long spine, high petiole, and sharp transversal crest on the dorsum of the petiolar node this species is easily distinguishable from related taxa even by simple visual inspection.
Geographic distribution. This species is known only from the Peloponnese peninsula.

Diagnosis of Temnothorax angustifrons species-complex
Workers of the Temnothorax angustifrons species-complex can be distinguished from those of other complexes by the combination of the following salient features: light yellow to light brown color; moderately longer than broad head (CL/CWb [1.135, 1.254]), sculpture of head dorsum shiny, with inconspicuously areolate ground sculpture combined with feeble costulate main sculpture; short to moderately long propodeal spines (SPST/CS [0.159, 0.267]), deviating from longitudinal axis of mesosoma by 47-52°; petiolar node in lateral view with a weakly concave frontal profile meeting dorso-caudal plate in an obtuse angle (95-105°) with a moderately sharp ridge, in dorsal view appearing as a visible (occasionally inconspicuous) anterior-lateral rim. Exploratory NC-clustering (Fig 2) revealed the existence of four species (Temnothorax angustifrons sp. n., T. lucidus sp. n., T. similis sp. n., T. subtilis sp. n.) within this complex, which was confirmed by LDA (S5 Table, Figs 6 and 7). Members of this species-complex are known to occur in Turkey and Crete. The occurrence of two samples found in Greece may be ascribed to anthropochory, i.e., dispersal by human activities (Fig 8).
Temnothorax angustifrons sp. n. The list of 59 non-type individuals belonging to 17 nest samples of other material investigated is given in S1 Table. Worker (Fig 9A-9C, S1 Table, S4 Table). Body color: yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than head dorsum and posterior region of gaster. Differential diagnosis. This species has the longest scape (SL/CS) and the narrowest frons (FRS/CS) of all species treated in this revision (S5 Table). Temnothorax angustifrons sp. n. strikingly resembles other members of this complex in Turkey. Therefore, the above specified characters may slightly overlap, but a simple ratio of their combination (FRS/SL) provides an excellent tool to separate this species from its relatives with an error rate of less than 5% for single individuals: T Simple ratios of the above-mentioned morphometric traits (SL/CS and FRS/CS) also separate T. angustifrons sp. n. from other species belonging to other species complexes in Turkey (S5 Table).
Geographic distribution. This species is known from Western Anatolia, Turkey (Fig 8). Its occurrence in Stenos, Greece, might be ascribed to anthropochory.
Temnothorax lucidus sp. n.   The list of 65 non-type individuals belonging to 22 nest samples of other material investigated is given in S1 Table. Worker (Fig 10A-10C, S1 Table, S4 Table). Body color: yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than head dorsum and posterior region of gaster. Antenna color pattern: clava concolorous funicle. Absolute cephalic size: 560-670 μm (mean = 618, n = 24). Cephalic length vs. Maximum width of head capsule (CL/CWb): 1.139-1.221 (mean = 1.173). Postocular distance vs. cephalic length (PoOc/CL): 0.368-0.400 (mean = 0.380). Postocular sides of cranium contour frontal view orientation: converging posteriorly. Postocular sides of cranium contour frontal view shape: convex. Vertex contour line in frontal view shape: straight. Vertex sculpture: main sculpture dispersed forked costate, ground sculpture inconspicuous areolate. Genae contour from anterior view orientation: converging. Gena contour line in frontal view shape: feebly convex. Differential diagnosis. This species differs from members of other species complexes treated in this revision by its smooth and shiny head dorsum and the relatively short propodeal spines. This character combination is shared with other species belonging to T. angustifrons complex: T. angustifrons sp. n., T. subtilis sp. n. and T. similis sp. n. Nest samples of T. lucidus sp. n. can be separated from those of T. angustifrons sp. n. by their shorter scape (SL/CS) and wider frons (FRS/CS). Their simple ratio (FRS/SL) provides an excellent tool to separate workers with less than 5% of error rate (see details under differential diagnosis under T. angustifrons sp. n.). Ratios of NOH/CS and PEH/CS help to distinguish this species from T. similis sp. n. on the level of nest samples. A discriminant (D4) function, including discriminant scores for separating single individuals with acceptably low error rate, is given in the differential diagnosis of the latter.
The spine length ratio (SPST/CS, see S4 Table) provides a good quick character to separate nest samples of T. lucidus sp. n. from those of T. subtilis sp. n., but in single workers this character may broadly overlap between the two species. A discriminant function with reduced character number (D4) yields 98.6% classification success rate (see details in differential diagnosis under T. subtilis sp. n.).
Temnothorax lucidus sp. n. can be easily separated from two additional species of the T. parvulus complex that occur in Crete, T. ariadnae sp. n. and T. helenae sp. n. based on shiny surface of the head dorsum. In exceptional cases, or if dust obstructs a clear view of the surface sculpture, nest samples of T. lucidus sp. n. can be separeted from T. ariadnae sp. n. by nonoverlapping ranges of body ratios (NodL/CS and NOL/CS, see S4 Table). Simple ratios do not help to dinstguish T. lucidus sp. n. and T. helenae sp. n., hence a simplified discriminant function (D3 = -0.0807 Ã POC +0.0896 Ã SL -0.0578 Ã SPTI -11.284) is recommended to separate nest samples of these specie. The same function yields 97.9% classification success rate between single individuals of these species.   The list of 23 non-type individuals belonging to 8 nest samples of other material investigated is given in S1 Table. Worker (Fig 11A-11C, S1 Table, S4 Table). Differential diagnosis. This species can be distinguished from other species belonging to different species complexes by its inconspicuously sculptured or smooth head and/or its shorter propodeal spines (SPST/CS). Temnothorax similis sp. n. differs from the superficially similar T. schoedli by non-overlapping ranges of PEH/CS and NOH/CS ratios on the level of nest sample means.
Though workers of T. similis sp. n. differ from those of other species in this complex by having a wider frons (FRS/CS) than T. angustifrons sp. n., a lower petiole (NOH/CS) than T. lucidus sp. n., and longer propodeal spines (SPST/CS) than T. subtilis sp. n. (see S4 Table), nest sample means of these characters slightly overlap between the latter two species and T. similis sp. n. In order to separate T. similis sp. n. from T. angustifrons sp. n., T. lucidus sp. n., and T. subtilis sp. n. a discriminant function with reduced character number (D4 = 0.0492 Ã SPST -0.1037 Ã PEH +0.0547 Ã ML -0.0417 Ã CL +2.2655) is provided, that yields 98.6% classification success rate for single individuals.
D4 scores for single individuals: Geographic distribution. This species is known from South and Central Anatolia, Turkey (Fig 8). According to the material available in this study, this species seems to occur in disjunct areas with a 600 km wide gap between the two known finding sites (S1 Table). Due to the fact that morphometric analyses have not revealed convincing morphological separation between the two populations we hold that the described disjunct distribution is more likely a result of sampling error rather than representing areas of two reproductively separated species.  The list of 158 non-type individuals belonging to 50 nest samples of other material investigated is given in S1 Table. Worker (Fig 12A-12C, S1 Table, S4 Table). Differential diagnosis. This species has the shortest propodeal spines of all species treated in this revision, and therefore can be separated from other species complexes by the non-overlapping SPST/CS ratio and its smooth and shiny head. Spine length ratio slightly overlaps with that of other species belonging to T. angustifrons complex yielding 94.5% success in distinguishing nest samples (S5 Table) from T. angustifrons sp. n., T. lucidus sp. n., and T. similis sp. n. Simple FRS/SL ratios help separating this species from T. angustifrons sp. n. (see differential diagnosis under the latter).
A discriminant function with reduced character number (D4) arrives at 98.6% classification success between single individuals and complete success for nest sample means of T. subtilis sp. n. and T. similis sp. n. (see differential diagnosis under T. similis sp. n.).
Though the SPST/CS ratio (see S4 Table) provides a fairly good quick key to separate nest samples of T. subtilis sp. n. from those of T. lucidus sp. n., this character may broadly overlap in single individuals of these two species. In order to determine single workers with high success, a discriminant function with reduced character number (D4 = +0.0717 Ã EL +0.0778 Ã NOH +0.0404 Ã SPST -0.0824 Ã SPBA -10.321) yielding 98.6% classification success rate can be used. Geographic distribution. This species is known from South Anatolia, Turkey, and Crete (Fig 8). The list of 44 non-type individuals belonging to 12 nest samples of other material investigated is given in S1 Table. Worker (Fig 13A-13C, S1 Table, S4 Table). Differential diagnosis. Workers of T. flavicornis might be confused with other longspined species, i.e., T. laconicus, T. lichtensteini and T. parvulus, but the coarse rugulose or rugulo-reticulate main sculpture on the head dorsum combined with a shiny ground sculpture help to distinguish T. flavicornis from related species by simple visual inspection. In case specimens are covered by dust, a simple ratio (SPBA/CWb) provides perfect separation of T. flavicornis and similar species (T. laconicus, lichtensteini and parvulus) at the level of nest sample means (see Fig 4). Additional morphometric data of nest sample means for T. flavicornis and related species (see S4 Table) provide further opportunities for safe separation.

Diagnosis of Temnothorax flavicornis species-complex
Geographic distribution. This species is widely distributed in the Balkans and Italy.  Table, Fig 14). Colonies of this species-complex occur in Southern Europe from Spain to Bulgaria and in Turkey (Fig 15). The list of 41 non-type individuals belonging to 10 nest samples of other material investigated is given in S1 Table. Worker (Fig 16A-16C, S1 Table, S4 Table).     Dorsal region of petiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent. Dorso-caudal petiolar profile contour line in lateral view shape: straight; concave. Dorsal region of postpetiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent. Differential diagnosis. Temnothorax laconicus can be distinguished easily from other species by its very long propodeal spines (having the longest propodeal spines of all taxa surveyed within this revision) and its low deviation (20-25°) from the mesosomal axis.

Diagnosis of
This species is most similar to T. lichtensteini. The simple ratio SPST/CS does not overlap between the two species at the level of nest sample means (see S4 Table)   The list of 298 non-type individuals belonging to 84 nest samples of other material investigated is given in S1 Table. Worker (Fig 17A-17C, S1 Table, S4 Table).  axis by >35°. How T. lichtensteini and T. laconicus can be separated is described under the latter (see above).
Geographic distribution. Temnothorax lichtensteini is distributed throughout the Northern coast of the Mediterranean basin from Spain to Turkey (Fig 15). Morphological ad molecular analyses suggested the existence of two distinct parapatric metapopulations (Fig 14), "East Mediterranean cluster" (Austria, Bulgaria, Croatia, Greece, N-Italy, Turkey) and "West Mediterranean cluster" (France, Italy, and Spain) [33], which, however are quite similar in general appearence. Though the"Eastern" lineage appears to be more robust than the "Western" lineage in some traits (MW/CS, PEW/CS, PPW/CS, SPBA/CS, SPWI/CS, SPTI/CS) [33], these characters broadly overlap and do not provide a safe separation. Due to this similarity and their parapatric occurrence the two lineages have not been raised to species or subspecies rank.

Diagnosis of Temnothorax nylanderi species-complex
Workers of the Temnothorax nylanderi species-complex can be distinguished from workers of other complexes treated in this revision by the combination of the following salient features: yellow to light brown color (Peloponnese populations occasionally dark brown to black); slightly longer than broad head (CL/CWb [1.100, 1.196]), sculpture of head dorsum dull: with areolate ground sculpture combined with parallel costulate main sculpture; moderately long to long propodeal spines (SPST/CS [0.253, 0.356]), deviating from longitudinal axis of mesosoma by 32-42°; petiolar node in lateral view with a concave frontal profile meeting truncate dorsum in a right angle to an obtuse angle (88-115°) with a narrowly rounded transition, without a conspicuous sharp fronto-dorsal ridge on the petiolar node.

Temnothorax crasecundus Seifert & Csősz 2014
Differential diagnosis. This species is easily confused with its parapatric relative, T. crassispinus. The D4 function that separates these two species is given under the latter.
Temnothorax crasecundus shares most of its surface sculpturing and general shape characteristics with T. helenae sp. n., despite the fact that they belong to different species complexes based on molecular phylogeny and morphometrics. In addition, the distribution of these species broadly overlaps in Bulgaria, Greece, and Turkey. A simple ratio (PoOC/NOH) yields a rather reliable discrimination with minor overlap between them (Fig 4). Other characters may also help in correct determination: T. crasecundus is larger (CS), has a wider frons (FRS/CS), higher petiolar node (NOH/CS), and longer propodeal spines (SPST/SC) than T. helenae sp. n. (see S4 Table).
Geographic distribution. This species is known to occur in the Balkans, Eastern Europe, and Turkey (Fig 19). Specimens collected as host of the slave-making ant Myrmoxenus tamarae in Daba, Georgia also appear to belong to this species [55]. The list of 119 non-type individuals belonging to 40 nest samples of other material investigated is given in S1 Table. Worker (Fig 21A-21C, S1 Table, S4 Table). Body color: brown; yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than head and posterior region of gaster. Antenna color pattern: clava concolorous funicle. Absolute Differential diagnosis. Temnothorax crassispinus may be confused with other long-spined species treated in this revision: T. angulinodis sp.n., T. laconicus, T. lichtensteini and T. parvulus. Temnothorax angulinodis sp.n. clearly differs from T. crassispinus by its sharply angulate petiolar node in lateral view (72-82°). In T. crassispinus, the frontal profile and the truncate dorsum of the petiole meet in an obtuse angle (100-115°). The deviation of the propodeal spines from longitudinal mesosomal axis (in lateral view) helps to separate T. crassispinus (32-42°) from T. laconicus and T. lichtensteini (20-25°). Temnothorax parvulus differs from T. crassispinus in the surface sculpturing on the head dorsum. If the samples are dust-covered, other Synopsis of the Ponto-Mediterranean Temnothorax nylanderi Group measures can also help: T. crassispinus is considerably larger (CS), has a higher petiolar node (SPST/CS), and a wider head (CL/CWb) than T. parvulus (S4 Table).
Temnothorax crassispinus shares most of its main characteristics, shape and surface sculpturing, with its siblings, T. crasecundus and T. nylanderi. The simple propodeal spine length ratio (SPST/CS) helps to separate nest samples of T. crassispinus from those of T. nylanderi without error, but the same character overlaps between nest sample means of T. crassispinus and T. crasecundus. The shortest discriminant formula (D4) that separates T. crassispinus from T. crasecundus with a classification success rate 95% in single individuals and 97% in nest sample means is D4 = +0.0392 Ã SL -0.0746 Ã SPST +0.0933 Ã SPL -0.0295 Ã SPWI -7.2179. Geographic distribution. This species is distributed from the Balkans to Central Europe (Fig 19). The distributional area of Temnothorax crassispinus lays between the ranges of its two parapatric relatives, T. nylanderi in the West and T. crasecundus in the East. The list of 60 non-type individuals belonging to 20 nest samples of other material investigated is given in S1 Table. Worker (Fig 22A-22C, S1 Table, S4 Table). Body color: brown; yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than head dorsum and posterior region of gaster. Antenna color pattern: clava concolorous funicle. Differential diagnosis. Temnothorax nylanderi has moderately long spines (SPST/CS) and therefore cannot be confused with long-spined T. lichtensteini. Non-overlapping SPBA/CS ratios help to distinguish it from T. flavicornis (S4 Table). Based on salient features, T. nylanderi might be misidentified as T. parvulus, but a simple ratio (SPTI/CS) reliably separate these species on the level of nest sample means (S4 Table). Temnothorax nylanderi can also be safely separated from weakly sculptured, lightly colored T. tergestinus using the ratios PoOC/CL and SPWI/CS (S4 Table). Geographic distribution. This species occurs in West Europe (Fig 19).

Diagnosis of Temnothorax parvulus species-complex
Workers of the Temnothorax parvulus species-complex can be distinguished from those of other complexes treated in this revision by the combination of the following salient features: dirty yellowish to brownish color; slightly longer than broad head (CL/ CWb [1.147, 1.242]), sculpture of head dorsum dull: with uniformly areolate ground sculpture combined with inconspicuous (or the lack of) main sculpture; short to long propodeal spines (SPST/CS [0.205, 0.331]), deviating from longitudinal axis of mesosoma by 38-42°; petiolar node in lateral view with a concave frontal profile meeting truncate dorsum in a right angle to an obtuse angle (100-110°) with a narrowly rounded transition, without a conspicuous sharp fronto-dorsal ridge on the petiolar node.
Exploratory NC-clustering revealed the existence of three species, Temnothorax ariadnae sp. n., T. helenae sp. n., and T. parvulus (Schenck, 1852), which was corroborated by confirmatory analyses (S5 Table, Fig 23). Members of this species-complex are known to occur in Europe, the Caucasus, and Turkey (Fig 24).
Temnothorax ariadnae sp. n. Locality data of the single non-type sample with 3 individuals investigated is given in S1 Table. Worker (Fig 25A-25C, S1 Table, S4 Table). Body color: brown; yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than head dorsum and posterior region of gaster. Antenna color pattern: clava concolorous funicle.  . The geographical range of T. ariadnae sp. n. and T. parvulus (Fig 24) does not overlap and more complicated means of separating single individuals are therefore not needed.
The separation of T. ariadnae sp. n. and T. helenae sp. n. can be more difficult, because both taxa co-occur in Crete and various body ratios overlap. A discriminant (D4 = -0.0932 Ã PPL -0.0767 Ã POC +0.1203 Ã PL -0.0384 Ã SPWI +9.827) function is the shortest formula that yields reliable separation for single individuals (99.4%) and nest sample means (100%) of T. ariadnae sp. n. and T. helenae sp. n. Temnothorax ariadnae sp. n. can be easily separated from two additional species whose distributional range expands to Crete, T. lucidus sp. n. and T. subtilis sp. n., based on the shiny surface of the head dorsum of the two latter. In exceptional cases, or if dust cover obstructs a clear view of the surface sculpture, body ratios help to distinguish T. ariadnae sp. n. from T. subtilis sp. n. by the longer head (CL/CWb), the larger eyes (EL) and the longer propodeal spines (SPST/CS) and from T. lucidus sp. n. by the non-overlapping NodL/CS, lower NOL/CS ratio and the longer head (S4 Table).
Geographic distribution. This species is endemic to Crete (Fig 24). The list of 154 individuals belonging to 43 nest samples of other material investigated is given in S1 Table. Worker (Fig 26A-26C, S1 Table, S4 Table). Body color: brown; yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than head dorsum and posterior region of gaster. Antenna color pattern: clava concolorous funicle. Differential diagnosis. Temnothorax helenae sp. n. shares characters with T. ariadnae sp. n. and T. parvulus, differential diagnoses between these species are given under these taxa.
Separation of T. helenae sp. n. from T. crasecundus can be difficult if only sculpture and shape characteristics are considered, but a single ratio (PoOC/NOH) yields >95% classification success for nest sample means (see Fig 4). Several slightly overlapping morphometric traits (CL/CWb, ML/CS, NOH/CS, SL/CS see S4 Table) may provide further options for determination.
The geographical distribution of T. helenae sp. n. overlaps with those of T. lucidus sp. n. and T. subtilis sp. n. in a narrow zone in Greek mainland and in Crete. Temnothorax helenae sp. n. can be separated easily from the latter two species by the coarser sculpture on its head dorsum and the lack of a sharp antero-lateral ridge on its petiolar node of T. helenae sp. n. Both T. lucidus sp. n. and T. subtilis sp. n. have a shiny head dorsum and a sharp to moderately sharp antero-lateral ridge on the petiolar node. Due to the fact that head dorsum of T. helenae sp. n. is occasionally moderately shiny and the antero-lateral rim on the petiolar node may occasionally be inconspicuous in both T. lucidus sp. n. and T. subtilis sp., in a few cases determination might be doubtful. The list of 103 individuals belonging to 27 nest samples of other material investigated is given in S1 Table. Worker (Fig 27A-27C, S1 Table, S4 Table). Body color: brown; yellow. Body color pattern: mesosoma, antenna and legs, waist and anterior region of 1st gastral tergite lighter than  superimposed by dispersed rugulae. Frontal profile of petiolar node contour line in lateral view shape: concave. Dorsal profile of petiolar node contour line angle value to frontal profile of petiole contour line in lateral view: 105-110°. Anterodorsal rim of petiole count: absent medially. Dorsal profile of petiolar node contour line in lateral view shape: slightly convex. Dorsal region of petiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent. Dorso-caudal petiolar profile contour line in lateral view shape: straight; concave. Dorsal region of postpetiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent.
Differential diagnosis. Temnothorax parvulus can be separated easily from members of other species-complexes by its homogenously areolate microsculpture on the head dorsum. The head sculpture of other species complexes may vary from smooth to coarsely rugulo-reticulate but is never homogenously areolate. In exceptional cases, long-spined species of the T. lichtensteini complex might exhibit a homogenous areolate sculpture on an extended area of the head dorsum, but T. parvulus is distinguished from T. lichtensteini and T. laconicus by its more erect propodeal spines (38-42°vs. ca. 20-25°). Temnothorax parvulus can also be safely separated from weakly sculptured, lightly colored T. tergestinus samples using slightly overlapping NOH/CS, SPTI/CS and SPWI/CS ratios (see S4 Table).
Temnothorax parvulus shares most its surface sculpture characters with other species belonging to this complex, T. ariadnae sp. n. and T. helenae sp. n. Spine length ratio (SPST/CL) helps distinguish Temnothorax parvulus from T. ariadnae sp. n. and T. helenae sp. n.
Non Geographic distribution. The known distribution of this species ranges from Western Europe to the Black see coast and Turkey and from Italy and the Balkans to Central Europe (Fig 24).

Diagnosis of Temnothorax sordidulus species-complex
Workers of the Temnothorax sordidulus species-complex can be distinguished from those of other complexes treated in this revision by the combination of the following salient features: brown to black color; slightly longer than broad head (CL/CWb [1.143, 1.278)], sculpture of head dorsum dull: with areolate ground sculpture combined with conspicuous parallel costulate or irregular reticulate main sculpture; moderately long to long propodeal spines (SPST/CS [0.220, 0.335]), deviating from longitudinal axis of mesosoma by 40-50°; petiolar node in lateral view with a concave frontal profile meeting occasionally indistinct truncate dorsum in an obtuse angle (110-120°) with a narrowly rounded transition, without a conspicuous sharp fronto-dorsal ridge on the petiolar node. Four species consist of this complex: Temnothorax artvinensis Seifert, 2006, T. schoedli Seifert, 2006, T. sordidulus (Müller, 1923) and T. tergestinus  stat.n.
Exploratory NC-clustering revealed the existence of four species (Temnothorax angustifrons sp. n., T. lucidus sp. n., T. similis sp. n., T. subtilis sp. n.) within this complex, which was confirmed by LDA (Figs 28 and 29). Members of this species-complex are known to occur in Turkey and Crete. The sporadic occurrence of two samples in Greece may be ascribed to anthropochory (Fig 30).
Type material investigated. Paratypes: 9 paratype workers of 3 nest samples were investigated from the type locality: Turkey, Artvin The list of 39 individuals belonging to 12 nest samples of other material investigated is given in S1 Table. Worker (Fig 31A-31C, S1 Table, S4 Table). Dorsal profile of petiolar node contour line in lateral view shape: slightly convex. Dorsal region of petiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent. Dorso-caudal petiolar profile contour line in lateral view shape: concave; straight. Dorsal region of postpetiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent.
Differential diagnosis. The very dark color of T. artvinensis prevents confusion of this species with any other Turkish Temnothorax treated in this revision.
Due to the fact that T. artvinensis is geographically well isolated from other dark brown or black species of the T. sordidulus complex, it is unlikely that this this species is confused with its siblings. However, for doubtful cases, or if anthropochory is suspected, a discriminant function after character reduction (D4 = +0.  (Fig 30). The list of 12 non-type individuals belonging to 4 nest samples of other material investigated is given in S1 Table. Worker (Fig 32A-32C, S1 Table, S4 Table). Body color: brown. Body color pattern: mesosoma, antenna and legs excluding femora, waist and anterior region of 1st gastral tergite lighter than head, femora and posterior region of gaster. Antenna color pattern: clava concolorous funicle. Absolute cephalic size: 572-696 μm (mean = 652, n = 11). Cephalic length vs. Maximum width of head capsule (CL/CWb): 1.143-1.196 (mean = 1.172). Postocular distance vs.   Dorsal region of mesosoma sculpture: areolate ground sculpture, superimposed by dispersed rugae. Lateral region of pronotum sculpture: areolate ground sculpture, main sculpture dispersed costate. Mesopleuron sculpture: areolate ground sculpture superimposed by dispersed rugulae. Metapleuron sculpture: areolate ground sculpture superimposed by dispersed rugulae. Frontal profile of petiolar node contour line in lateral view shape: concave. Anterodorsal rim of petiole count: absent medially. Dorsal profile of petiolar node contour line in lateral view shape: widely rounded or slightly angulate area. Dorsal region of petiole sculpture: ground sculpture areolate, main sculpture dispersed rugose. Dorsal region of postpetiole sculpture: ground sculpture areolate, main sculpture dispersed rugose. Differential diagnosis. Due to its dark brown to black color this species can only be confused with its European sibling, T. tergestinus. The other black species belonging to this complex, T. artvinensis is endemic to Turkey. Though T. sordidulus seems endemic to the Dinaran Alps, its distribution slightly overlaps with that of T. tergestinus in Slovenia and Austria. The darker, often black color of T. sordidulus in contrast to the brown color of Central European populations of T. tergestinus, its shorter spines (SPST/CS), shorter scapes (SL/CS), longer petioles (PL/CS) and petiolar nodes (NOL/CS) may help to separate T. sordidulus from Central European populations of T. tergestinus. If these ratios overlap and do not clearly separate between the two taxa, or if T. sordidulus shall be separated from Bulgarian or Greek Dorsal region of mesosoma sculpture: areolate ground sculpture, superimposed by dispersed rugae. Lateral region of pronotum sculpture: areolate ground sculpture, main sculpture dispersed costate. Mesopleuron sculpture: areolate ground sculpture superimposed by dispersed rugulae. Metapleuron sculpture: areolate ground sculpture superimposed by dispersed rugulae. Frontal profile of petiolar node contour line in lateral view shape: concave. Dorsal profile of petiolar node contour line angle value to frontal profile of petiole contour line in lateral view: 110-120°. Anterodorsal rim of petiole count: absent medially. Dorsal profile of petiolar node contour line in lateral view shape: slightly convex. Dorsal region of petiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent. Dorso-caudal petiolar profile contour line in lateral view shape: straight; concave. Dorsal region of postpetiole sculpture: ground sculpture areolate, main sculpture dispersed rugose; ground sculpture areolate, main sculpture absent.
Differential diagnosis. This species can be separated from members of other species complexes by the ruguloreticulate main sculpture on head dorsum that turns irregular on the vertex and the sides of the head. In some Western and Central European populations, the surface sculpturing might be less conspicuous, which may lead to possible confusion with T. parvulus and T. nylanderi, particularly if the body surface is covered by diffuse dust.
Weakly sculptured, lightly colored specimens of T. tergestinus can be safely separated from T. nylanderi using PoOC/CL and non-overlapping SPWI/CS ratios (S4 Table). Slightly overlapping NOH/CS, SPTI/CS and SPWI/CS ratios (S4 Table) help to distinguish T. tergestinus and T. parvulus samples. This species is not supposed to be confused with other members of other species complexes by combination of various traits. Temnothorax tergestinus shares most of its characters with T. sordidulus and T artvinensis. The latter is separated from T. tergestinus both by the broad gap in their distribution range ( Fig  30) and by discriminant (D4) function (for details see differential diagnosis of T artvinensis). A discriminant function (D7) that helps separating T. tergestinus from T. sordidulus is given in differential diagnosis under the latter.
Nomenclatural issues. A multivariate analyses of 22 continuous characters in 186 individuals of 64 nest samples (Fig 2, S5 Table, Figs 28 and 29) confirmed the earlier hypothesis [31] that two separate species exist: T. sordidulus (Müller, 1923) and T. saxonicus (Seifert, 1995). Our results, however, differ from the earlier [31] view in two major points: a) the Bulgarian populations do not belong to T. sordidulus, but grouped in a cluster that was formerly referred to as T. saxonicus (sensu Seifert [31]); b) in this novel classification, the type series of T. tergestinus  is nested in the T. saxonicus (sensu Seifert [31]) cluster with a posterior probability of p = 0.98. The type series of T. saxonicus (Seifert, 1995) fell in the same cluster with a posterior probability of p = 0.99. Therefore, we propose a new junior synonymy for Temnothorax saxonicus (Seifert, 1995) with Temnothorax tergestinus  stat. n.
The newly outlined species boundaries allow a considerably easier separation of T. tergestinus and T. sordidulus (Figs 28 and 29), which underpins the increased robustness of the new classification. A combination of 7 traits yields a perfect discrimination (see also differential diagnosis in T. sordidulus), in contrast to the 18 characters that was used to require earlier [31].
Geographic distribution. According to our newly recognized pattern T. tergestinus spreads from France to Bulgaria and Greece (Fig 30).
Supporting Information S1 Table. List of samples of ants Temnothorax nylanderi species-complex morphometrically investigated. All samples can be uniquely distinguished by sample-specific abbreviations