Gephyromantis tohatra sp. nov. (Figs. 2–3) Holotype. ZSM 422/2016 (ZCMV 15245), adult male, collected at a site along the main trail leading from Camp Simpona to the summit of the Marojejy Massif, Sava Region, northeastern Madagascar, at geographical coordinates 14.44400°S, 49.73791°E and an elevation of 1758 m above sea level (Fig. 2, 3a–b) by M. D. Scherz, M. Vences, J. H. Razafindraibe, and M. Bletz, in the evening of the 19th of November 2016. Paratype. UADBA uncatalogued (ZCMV 15236), adult male, with the same collection data as the holotype. Diagnosis. The new species is assigned to the genus Gephyromantis and subgenus Duboimantis on the basis of its mostly connected lateral metatarsalia, moderately enlarged finger tips, type 2 femoral glands (Glaw et al. 2000), inner and outer dorsolateral ridges sensu Vences & Glaw (2001), presence of webbing between the toes, absence of a distinct white spot in the centre of the tympanum, and presence of small supraocular spines. It is characterised by the following unique suite of characters: (1) SVL at least up to 33 mm, (2) orange to yellowish ventral colouration in males, (3) type 2 femoral glands with 10–21 small granules, (4) strongly distinct inner and outer dorsolateral ridges sensu Vences & Glaw (2001), (5) a pale loreal region and broad pale marking covering most of the anterior edge of the tympanum, (6) second finger much shorter than fourth, (7) third toe much shorter than fifth, and (8) 7– 10 distinct pulsed notes per call. Gephyromantis tohatra sp. nov. differs from the majority of Duboimantis species by smaller body size: adult male SVL 33 vs.> 35 mm in G. granulatus, G. cornutus, G. redimitus, G. tschenki, G. salegy, G. tandroka, G. luteus, G. sculpturatus, G. plicifer, and G. zavona. It can also be distinguished from all other species of Duboimantis by its orange to yellowish venter (typically whitish in all other species). Additionally, the presence of distinct inner and outer dorsolateral ridges distinguishes the new species in particular from G. granulatus, G. redimitus, and G. zavona in which these folds (in particular the inner dorsolateral ridges) are absent or very weakly expressed. Gephyromantis luteus, G. plicifer, and G. sculpturatus have dark concave markings in the suprascapular region (absent in the new species) and much more extensive webbing on the feet. Gephyromantis cornutus, G. tschenki, and G. tandroka have strongly expressed interocular tubercles (absent in the new species). Three species of Duboimantis overlap with the new species in body size. Gephyromantis leucomaculatus is typically larger (adult male SVL 32–41 mm) and differs by absence of inner dorsolateral ridges and straight outer dorsolateral ridges (vs. anteriorly curved). Gephyromantis moseri (adult male SVL 27–40 mm) has a granular dorsum without or with interrupted dorsolateral ridges (vs. smooth or reticulated dorsum and continuous dorsolateral ridges in the new species) and interocular tubercles (vs. absence). Gephyromantis schilfi (adult male SVL 27–30 mm) occurs in close syntopy with the new species but differs from it by the absence of distinct inner and outer dorsolateral ridges (vs. presence in the new species). Bioacoustically, the new species differs from all species in Duboimantis by advertisement call structure. Most species in this subgenus emit regularly repeated calls, each of which consists of a single, typically pulsed or pulsatile note (G. granulatus, G. cornutus, G. leucomaculatus, G. moseri, G. redimitus, G. schilfi, G. tschenki, G. tandroka, G. plicifer, G. zavona). In many of these species, notes can also be more irregularly arranged in short series, but calls do not consist of a regular number of 7–10 short and rapidly repeated notes as in G. tohatra sp. nov. This characteristic is particularly helpful in its distinction from the sympatrically occurring and morphologically similar G. schilfi, which emits a single, densely pulsed note. Gephyromantis sculpturatus and G. luteus emit a fast series of squeaking, unpulsed notes that are very distinct from the call of G. tohatra. The only other Duboimantis with a surprisingly similar call structure to the new species is G. salegy, which however differs by a substantially larger body size of 45–48 mm in adult males and by a larger number of granules in femoral glands (36–38 vs. 10– 21). G. tohatra sp. nov. differs from its probable sister species G. schilfi (Fig. 1) by a substantial genetic differentiation of 4.3% uncorrected pairwise distance (p-distance) in the 16S gene, and from all other species of Duboimantis by a 16S p-distance>6%. The bioacoustically similar G. salegy differs by 6.8% 16S p-distance and is not the sister taxon of the new species. Holotype description. A specimen in a good state of preservation, the right thigh muscle taken for DNA tissue samples. SVL 32.7 mm. For other measurements see below. Body gracile; head longer than wide, wider than body; snout pointed in dorsal and lateral view; nostrils directed laterally, protruding slightly, much nearer to tip of snout than to eye; canthus rostralis distinct, straight; loreal region concave and weakly oblique; tympanum indistinct, oval, its horizontal diameter 42% of eye diameter; supratympanic fold distinct, straight, following the line of the eyelid to above the insertion of the arm; tongue fairly narrow, posteriorly bifid; vomerine teeth clearly distinct, arranged in two small aggregations on either side of the midline of the palate at the level of the anterior edge of the eye, posteromedial to choanae; choanae small and rounded and laterally displaced. Dark, translucent dermal fold below each jaw starting at the level of the mid-eye. Arms slender, subarticular tubercles single, highly distinct; outer metacarpal tubercle small and indistinct and inner metacarpal tubercle relatively well developed; fingers without webbing; relative length of fingers 1 et al. (2000), consisting of roughly 16 granules on the right thigh and 21 on the left thigh. Measurements, all in mm: SVL, 32.7; HW, 9.5; HL, 12.0; TD, 1.9; ED, 4.5; END, 3.3; NSD, 2.7; NND, 3.8; UAL, 6.7; LAL, 8.0; HAL, 11.0; THIL, 17.7; TIBL, 19.4; TARL, 9.5; FOL, 18.2. In life (Fig. 3a,b), the dorsum was a medium brown with the external dorsolateral ridges distinctly tan, the internal ridges less distinct but also light in colour. The dorsal surface of head was not distinct from the dorsum, with a slightly darker brown band between the eyes. The flank fades to a more grey-brown ventrally. The dorsal surfaces of hindlimbs were as the flank in colour, with three dark brown crossbands on the thigh, three on the shank, and no crossbands on the foot. The dorsal surface of forelimb was similar to the lower leg in colour. No light annulus was present before the tip of each toe or finger. The lateral head was distinctly differentiated from the dorsal head, being almost black along the canthus rostralis, from the nostril to the tip of the snout, over the dorsal portion of the tympanic region beneath the supratympanic fold, and beneath the eye. An immaculate white spot was present anterior to the eye, with a more faded spot posterior to the eye over part of the tympanum. The iris was whitish above, and dark brown below the pupil. The lower lip was yellowish. Ventrally, the hands, forearms, tarsus and feet were all burnt umber in colour, with pale yellowish subarticular tubercles on the fingers, the subarticular tubercles of the toes being more orange in colour. The ventral chin and throat were burnt umber, especially on the jaw, making a distinct line anterior to the pectoral girdle, posterior to which the abdomen was orange with a few dark brown flecks anteriorly. The ventral thighs were also orange. The femoral glands were a creamy orange. The ventral humerus was semi-translucent. After six months in preservative, the coloration has faded. Dorsally, areas that were brown have become grey to silver, with light grey dorsolateral ridges. The head colouration is mostly retained. The ventral colouration has lost its orange hue, and is now a yellow-cream. All elements of the colour pattern are however retained. Variation. The paratype was not available for measurements but was of similar size as assessed in the field. Its colour in life (Fig. 3c,d) was generally similar to the holotype, but it lacked the bright whitish marking anterior to the eye, had only one dark brown marking below the eye (two in the holotype), had slightly broader crossbands on the legs, and its outer dorsolateral ridges were less distinctly coloured than those of the holotype. Ventrally it was also similar to the holotype but its abdomen was more yellowish, with yellow-cream patches posterior to the pectoral girdle. Etymology. The specific name is a noun in apposition, derived from the Malagasy word ‘ tohatra,’ meaning ‘stairs,’ in reference to the difficulty of hiking up Marojejy on very steep trails with stair-like stretches between Camp Marojejia and Camp Simpona, and especially from Camp Simpona to the summit, necessary to discover this species. Natural history. Calling males were heard at night at a site between Camp 3 (= Camp Simpona) and the summit of the Marojejy Massif, within montane rainforest of rather low canopy height and in an area of very steep slopes. Most specimens were calling in a dry headwater area of a small stream, in between very dense bushy undergrowth, at perch heights of 0.5–1 m above the ground from leaves and branches. In some areas the call could be heard from far outside the fores. Advertisement calls. Call recordings were made in an area where several (>5) males were calling relatively close to one another, and were obtained from the holotype and the paratype. The calls are considered advertisement calls because they were loud, stereotyped calls emitted by males without any sign of conflict with conspecifics. Calls consist of 7–10 pulsed notes (Fig. 4). Calls are separated by long and rather irregular intervals. In the holotype, call duration was 547–640 ms (594 ± 38 ms; n = 9), inter–call interval 1258–2306 ms (1622 ± 398 ms; n = 8), dominant frequency 3014–3273 Hz, and approximate prevalent band width 2000–5000 Hz. Each note had three to four recognizable pulses, with the last note of the call having more pulses (about 8). Characteristics of notes were measured in two calls. Pulses were not separated by distinct silent intervals. Intensity of each note was highest in the first pulse of each note. Note and interval duration are only tentatively presented, as there remains a weak background sound energy during these periods. Call 1: Note duration 34–78 ms (53 ± 14 ms, n = 7), internote interval duration 19–40 ms (30 ± 8 ms, n = 6), pulses per note 3–8 (4 ± 2, n = 7). Call 2: Note duration 45–77 ms (57 ± 11 ms, n = 7), inter-note interval duration 7–36 ms (26 ± 12 ms, n = 6), pulses per note 3–9 (4 ± 2, n = 8). Calls of the paratype were similar in most parameters but differed in that the pulses in notes were neither distinct nor clearly distinguishable; clear silent intervals between notes were not present, possibly due to background noise, and a weak frequency modulation was apparent in each note. Call duration was 502–798 ms (698 ± 117 ms, n = 5), inter-call interval duration 1458–1461 (n = 2), notes per call 7–10 (9 ± 1 ms, n = 5), dominant frequency of first three notes in a call 2454–2713 Hz (2637 ± 124 Hz, n = 4), dominant frequency of last 2–3 notes in a call 3014– 3229 Hz (3122 ± 124 Hz, n = 4), time between start of one note and subsequent note 73–86 ms (82 ± 4, n = 17 notes from 2 calls). Comparative data for syntopic G. schilfi. A few individuals of G. schilfi were heard emitting advertisement calls very close to the site in which G. tohatra sp. nov. individuals were heard and collected, on the same day and ca. 15 minutes after collecting G. tohatra sp. nov. One adult calling male was recorded and collected (voucher ZSM 415/2016, field number ZCMV 15246) from dense vegetation at a perch height of ca. 1 m above the ground, in a rainforest area on a steep mountain slope. Measurements: SVL, 29.8; HW, 9.5; HL, 11.6; TD, 1.9; ED, 4.6; END, 2.9; NSD, 2.5; NND, 2.6; UAL, 4.8; LAL, 7.5; HAL, 9.3; THIL, 15.6; TIBL, 18.3; TARL, 9.2; FOL, 16.5. As previously described (Glaw & Vences 2000), the call is a single pulsed note emitted after long and rather irregular silent intervals (Fig. 5). Call duration is 483–538 ms (518 ± 25 ms; n = 4), inter-call interval duration is 2917–5121 ms (3672 ± 1255 ms, n = 3), one note consists of ca. 73–89 pulses (80 ± 7; n = 5). Pulses were not clearly separated by silent intervals and in some cases not unambiguously distinguishable from each other. The intervals between pulse intensity maxima were somewhat irregular, becoming longer towards the end of the call, with a duration of about 5 ms at the beginning of the call and 7–12 ms at the end of the call. Dominant frequency was 3402–3445 Hz and approximate prevalent bandwidth was between 2000–4000 Hz. Range extension for G. schilfi and G. tandroka. The goal of this study is not a comprehensive reassessment of the phylogeny within the subgenus Duboimantis and molecular analyses were thus limited to a subset of taxa. In exploratory analyses, the new species clustered with G. salegy, G. schilfi, and G. tandroka, three species that formed a highly supported subclade of Duboimantis in the analyses of Wollenberg et al. (2011) and Kaffenberger et al. (2012). For our analysis we thus expanded the sampling for these three species with additional individuals. We found molecular evidence for new populations of G. schilfi and G. tandroka, both of which were so far known only from Marojejy. Both these species were now also found on the Sorata Massif north of Marojejy, and had different degrees of genetic differentiation. While the populations of G. schilfi (Sorata vs. Marojejy) differed by only 0.6– 0.8% uncorrected 16S p-distance, those of G. tandroka had a higher divergence of 5.4% which suggests their taxonomic status requires revision and we therefore here flag the Sorata population as new Unconfirmed Candidate Species, Gephyromantis sp. Ca32, rationalised following Vieites et al. (2009). Another candidate species of Duboimantis (G. sp. Ca30 from the Bealanana district of northern Madagascar) that was also recently identified (Scherz et al. in press) was not included in our analysis here, but we have confirmed that it is not conspecific with G. sp. Ca32 (data not shown).