Holotype. —National Museum of Natural History, Washington DC (USNM 486962), male collected by F. B. Gill, at 1,100 m elevation, 8 km south-southwest of Saint-Denis (20°52′44″S, 55°26′53″E), northern Reunion, in the south-west Indian Ocean, on 17 November 1964 (Figs. 1–2).

Paratypes. —USNM 486964, female collected by F. B. Gill, at Nez de Bœuf (21°12′20″S, 55°37′14″E), Reunion, on 22 November 1964; USNM 486965, male collected by F. B. Gill, at Nez de Bœuf, Reunion, on 22 November 1964; and USNM 486966, male collected by F. B. Gill, 6 km north-northwest of Le Vingt-Septième (c.21°11′38″S, 55°29′47″E), on 23 November 1964 (Fig. 1).

Diagnosis. —Based on the rather small sample sizes available, saffordi is separable from nominate francicus using plumage characters and biometrics (see Table 1, Fig. 3). For biometrics, saffordi has an on average smaller / narrower pale rump patch than nominate francicus. Additionally, tail fork is potentially a useful separating character, with that of saffordi being on average considerably (c.20%) shallower, although judging specimens in comparable states of feather wear unquestionably can limit the reliability of this feature in the museum. Nevertheless, it can be obvious when comparing images of live birds on the wing (Fig. 4). In plumage, the rump patch of saffordi is far less contrasting by virtue of the often many fewer pure or almost pure white feathers therein, in addition to its being narrower. As noted by M. R. Browning (in Chantler & Driessens 2000), the wings, especially the flight-feathers, of nominate francicus appear substantially more glossed (greenish in some lights, but bluer in others) than those of saffordi, and this remains obvious even in much older specimen material from Mauritius (see Fig. 1). The head-sides and ear-coverts of nominate francicus typically appear much more solidly dark than those of saffordi, with the result that its cap seems much larger, whereas saffordi appears to display much more of a paler and contrasting neck-collar. Finally, specimens of saffordi appear considerably more patterned below than those of nominate francicus, displaying much more obvious shaft-streaks on the lower breast and belly, and a sharper division between the darker throat and breast vs. paler (and greyer) remainder of the underparts; in nominate francicus this pectoral band effect is much reduced, as well as being much higher on the breast. A. f. saffordi is separated from the similar but larger and darker Aerodramus elaphrus of the Seychelles by most of the same characters as can be used to distinguish the latter from nominate francicus, namely the shorter wings and bill of saffordi (A. elaphrus has wing 115–119 mm and bill to feathers 7.5–8.0 mm; Rocamora 2013), its narrower wing base and smaller eye. In contrast to nominate francicus, the rump patch of saffordi being rather less obvious within the otherwise dark upperparts, is more similar to A. elaphrus.

Description of holotype. —See also Figs. 1–2. Colours correspond to those in Smithe (1976). Very dusky above, and paler, slightly greyer below. Head basically shows four tones: the cap which is blacker (between Color 119 Sepia and 119A Hair Brown) and quite well demarcated reaching level with lowest part of eye; the lores which are even blacker (closest to Color 89 Jet Black); the paler greyish-brown ear-coverts (close to Color 119B Dark Drab) and chin (slightly paler than the ear-coverts); with the throat being the palest part of the head (closest to 119C Light Drab). Upperparts: largely dusky brown (Color 221 Vandyke Brown), contrasting with ill-defined paler grey rump (between Color 119C Light Drab and 119D Drab-Grey), with very few whitish bases and edges exposed; rectrices, remiges and some of the coverts and tertials are darkest part of wing, almost blackish brown (Color 119), with very slight greenish gloss/iridescence. Underparts: lower throat to breast 119C Light Drab, or slightly darker (119B Dark Drab) forming a very broad and somewhat diffuse breast-band, below which the remainder of the underparts (belly to undertail-coverts) become paler (between 119C and 119D); especially from the mid-breast across the belly, shaft-streaks are very well pronounced and there are dark feather centres exposed between the lower throat and upper breast forming a broken mesial stripe. Underwing-coverts as top of head (nearest to Color 119), but underside of remiges marginally paler (between Color 119 Sepia and 119A Hair Brown). Bare parts all dark post-mortem (original colours unknown).

Measurements of holotype .—Wing 107.0 mm, tail 51.0 mm, bill (to skull) 5.8 mm, bill (to feathers) 3.3 mm, bill depth (at feathers) 1.5 mm, tail fork 4.9 mm, and depth of white rump patch 9.9 mm; mass 8.6 g.

Variation in the type series. —The three paratypes (Fig. 1) are very similar both to one another and to the holotype, but USNM 486965 has the most visible white on the rump, whereas the holotype and USNM 486966 show least.

Measurements of paratypes. — USNM 486964 (female): wing 109.5 mm, tail 52.0 mm, bill (to skull) 5.9 mm, bill (to feathers) 3.3 mm, bill depth (at feathers) 1.6 mm, tail fork 6.5 mm, and depth of white rump patch 10.3 mm; mass 8.9 g. USNM 486965 (male): wing 111.5 mm, tail 49.0 mm, bill (to skull) 6.1 mm, bill (to feathers) 3.7 mm, bill depth (at feathers) 1.5 mm, tail fork 5.2 mm, and depth of white rump patch 12.1 mm; mass 11.8 g. USNM 486966 (female): wing 110.0 mm, tail 50.0 mm, bill (to skull) 5.7 mm, bill (to feathers) 3.0 mm, bill depth (at feathers) 1.8 mm, tail fork 5.0 mm, and depth of white rump patch 12.6 mm; mass 8.5 g.

Geographic distribution. —Endemic to the island of Reunion (2,152 km²), a French overseas territory, where it is widespread from sea level to the island's highest point (3,069 m, Piton des Neiges), although known breeding sites are comparatively few, and most support rather small numbers of birds. The largest are in the Ravine de la Grande Chaloupe, an Important Bird Area in the north-west of the island (Le Corre & Safford 2001) and at La Chapelle, Cirque de Cilaos, in the centre of Reunion (Cheke & Hume 2008). However, at least some caves (potential breeding sites) are almost certainly inaccessible and detailed surveys do not appear to have been attempted, in contrast to the situation on Mauritius (Safford 2013).

Etymology. —In naming this new swiftlet subspecies, it is gratifying to be able to pay tribute to the many contributions of Roger Safford to our avifaunal knowledge of the south-west Indian Ocean islands, especially the Mascarenes. His work there began in the late 1980s and culminated recently in the co-authorship of a superb new field guide to the region (Hawkins et al. 2015) and, even more importantly, the eighth adjunctive volume in The birds of Africa series, of which he was both an editor and a primary author (Safford & Hawkins 2013). Safford is currently a Senior Programme Manager at BirdLife International with a special responsibility for threatened species conservation. The name saffordi is a noun declined in its genitive singular form.

Taxonomic rank. —Employing a modern interpretation of the Biological Species Concept, we elect to describe saffordi at subspecies level based on the multiple but relatively modest morphological characters that distinguish the Reunion population of this swiftlet from that on Mauritius. This fulfils the notion expressed by Remsen (2010), in a defence of their usefulness in avian taxonomy, that subspecies should represent “geographic populations diagnosable by one or more phenotypic traits.” Given that the discriminant analyses correctly assigned all but one of 27 specimens from the two islands to population, including all of the Reunion material (see Results, Biometrics), it appears that Patten & Unitt's (2002) advocacy of a 95% rule for assessing diagnosability of subspecies can be met in the case of saffordi. Despite that the sum of these differences might be considered comparatively significant within a morphologically extremely conservative group such as swiftlets, given our current lack of vocal, behavioural or genetic evidence of additional differentiation, we contend that subspecies rank is most appropriate for the Reunion population. For the present we have no basis to believe that, other than as a function of geography, the two populations would function as reproductively isolated units, and we have no evidence of potential pre-mating isolation mechanisms. Morphological characters demonstrate that there are qualitative and, to a lesser degree, quantitative differences between the two populations. Traditional genetic data would merely give an indication of a lack of gene flow over a certain time period.

Biogeographical considerations. —The landbird faunas of Mauritius and Reunion are strongly characterised by: (a) being highly depauperate (at least following several centuries of human impact); (b) extremely high rates of endemism among their native birds; and (c) the majority of species on both islands being introduced. None of these characters can be considered surprising given their comparative isolation, lying c.2,000 km from the African mainland and 175 km from each other; their volcanic and independent origins (Mauritius is c.8–9 million years old, Reunion c.2–5 million years old: see Safford & Hawkins 2013: 15); as well as their comparatively long periods of continuous human colonisation and importance to traders.

Discounting the introduced taxa, under most previous taxonomic arrangements Reunion boasted six extant endemic landbird species (a harrier Circus maillardi, a cuckooshrike Coracina newtoni, a bulbul Hypsipetes borbonicus, a stonechat Saxicola tectes, and two white-eyes, Zosterops borbonicus and Z. olivaceus), one endemic subspecies (Terpsiphone b. bourbonnensis, a monarch flycatcher) and two other native landbird species—Aerodramus francicus and Phedina borbonica. The latter is distinguished by being subdivided into separate races on Madagascar (P. b. madagascariensis) and in the Mascarenes (Mauritius and Reunion; nominate). The Saxicola is sometimes treated as a subspecies of a much wider Old World species-complex, or as a race of a purely Afrotropical species, but under either arrangement, the taxon concerned, tectes, is endemic to Reunion.

Extant Mauritian landbird endemicity was previously measured by eight species (a kestrel Falco punctatus, a pigeon Nesoenas mayeri, a parakeet Psittacula eques, a cuckooshrike Coracina typica, a bulbul Hypsipetes olivaceus, two white-eyes Zosterops chloronothos and Z. mauritianus, and a fody Foudia rubra) and one subspecies (Terpsiphone bourbonnensis desolata). The only other native landbirds were the two species shared with Reunion (i.e. Phedina borbonica and Aerodramus francicus).

It is worth remarking, however, that when considering overall known diversity, taking into account extinctions, the basic biogeographical differences between the two islands shrink, and it becomes obvious that their avifaunas are much more similar than might be apparent from the above (Cheke & Hume 2008, Safford & Hawkins 2013). Nevertheless, most populations are congenerics are clearly differentiated to some extent. Consequently, it is unsurprising that Mascarene Swiftlet should in fact conform to the same pattern of unique taxa on the different islands. This leaves the Phedina as the only landbird native to both Mauritius and Reunion that displays no named morphological differentiation between the two islands.

Conservation. —Tire following is largely based on Safford (2013). In the mid 1990s the Reunion population of Aerodramus francicus was estimated to be in excess of 10,000 individuals (whereas Barré et al. 1996 thought numbers to be approximately 5,000 birds); indeed, the colony at La Chappelle (see Geographic distribution) was estimated to comprise more than 10,000 nests alone. This was believed to reflect a genuine increase in numbers since the 1970s (when the species was thought to be much less numerous on Reunion compared to Mauritius), rather than merely improved coverage. Several colonies on Reunion were considered to be threatened by the caving activities of speleologists, while the species' nests have recently acquired a reputation for enhancing effects of cannabis (M. Le Corre in Chantler & Driessens 2000, Cheke & Hume 2008). On Mauritius, A. francicus declined between the early 20th century and the 1970s, but by the time of the first detailed survey, in 1998, it was believed to be increasing. At this time, numbers were estimated at just 2,244–2,610 individuals, with no cave holding more than 600–700 and 19 of 34 caves known to harbour the species held fewer than 30 swiftlets (Middleton 1999). Varied forms of human persecution and exploitation, which certainly was ongoing in the 1990s, as well as the deliberate blocking of cave entrances, suggest that the Mauritian population is not only considerably rarer but also at greater risk than that on Reunion. Consequently, should any future taxonomic work determine that saffordi and francicus merit species rather than subspecies rank, then it appears certain that francicus sensu stricto would automatically be listed as globally threatened.