Teniky: enigmatic architecture at an archaeological site in southern Madagascar

ABSTRACT

Teniky is an isolated inland archaeological site in the Isalo massif of southern Madagascar, with enigmatic rock-cut architecture that is unique in all of Madagascar and the wider East African coast. In the first half of the twentieth century visitors described archaeogical structures within a fluvial cirque (i.e. an amphitheatre-shaped valley formed by fluvial erosion), which included man-made terraces, rock-cut niches in the steep cliffs and a rock shelter delimited by walls consisting of carved sandstone blocks. Our investigations at Teniky have led to the discovery of further archaeological structures beyond the cirque. These structures include further terraces, rock-cut niches, stone basins and carved sandstone walls on a hill 2.5 km to the west, as well as sandstone quarries and dry stone walls in the valley in between. AMS radiocarbon dating of charcoal found during excavations at the newly discovered rock-cut niches and carved sandstone walls indicates that they were constructed in the late first millennium/early second millennium c. tenth to twelfth century AD. Pottery sherds at Teniky include Chinese and Southeast Asian ceramics dating to broadly the eleventh to fourteenth centuries, concordant with the radiocarbon ages. The presence of imported ceramics indicates that those living at Teniky participated in Indian Ocean trade networks in medieval times, despite being over 200 km from the nearest coast. The closest stylistic parallels to the enigmatic rock-cut architecture at Teniky are found in present-day Iran, particularly in the Fars region, where rock-cut niches dating to the first millennium AD or earlier have been attributed to Zoroastrian communities. We tentatively interpret the rock-cut architecture at Teniky as part of a former necropolis made by settlers with Zoroastrian origins. More archaeological research is required to test this hypothesis and to address questions such as: where and when did these settlers first arrive on the coast of Madagascar; why and when did they move inland; how did they live and how did they interact with other populations on the island; and why and when was the site deserted?

RÉSUMÉ

Teniky est un site archéologique isolé à l’intérieur des terres dans le massif de l’Isalo au sud de Madagascar, avec des vestiges rupestres énigmatiques qui sont uniques dans tout Madagascar et sur la côte plus vaste de l’Afrique de l’Est. Dans la première moitié du vingtième siècle, les visiteurs ont décrit des vestiges archéologiques dans un cirque fluvial (c'est-à-dire une vallée en forme d’amphithéâtre formée par l’érosion fluviale), comprenant des terrasses artificielles, des niches creusées dans les falaises abruptes et un abri sous roche délimité par des murs constitués de blocs de grès taillés. Nos recherches à Teniky ont permis de découvrir d’autres vestiges archéologiques au-delà du cirque. Ces vestiges comprennent d'autres terrasses, des niches creusées dans la roche, des bassins en pierre et des murs en grès taillés sur une colline située à 2.5 km à l’ouest, ainsi que des carrières de grès et des murs en pierres sèches dans la vallée intermédiaire. La datation au radiocarbone AMS des charbons de bois trouvés lors des fouilles des niches creusées dans la roche et des murs de grès sculptés récemment découverts indique qu’ils ont été construits à la fin du premier millénaire/début du deuxième millénaire (vers les dixième au douzième siècles) après J.-C. Les tessons de poterie de Teniky comprennent des céramiques chinoises et d'Asie du Sud-Est datant du onzième au quatorzième siècles après J.-C., ce qui concorde avec les âges radiocarbones. La présence de céramiques importées indique que les habitants de Teniky ont participé au réseau commercial de l’océan Indien à l’époque médiévale, bien qu’ils aient vécu à plus de 200 km de la côte la plus proche. Les parallèles stylistiques les plus proches de l'architecture énigmatique de Teniky se trouvent en Iran, en particulier dans la région de Fars, où des niches creusées dans la roche datant du premier millénaire après J-.C. ou plus tôt ont été attribuées à des communautés zoroastriennes. Nous interprétons provisoirement l’architecture taillée dans les roches à Teniky comme faisant partie d’une ancienne nécropole construite par des personnes ayant des origines zoroastriennes. Des recherches archéologiques supplémentaires sont nécessaires pour tester cette hypothèse et répondre à des questions telles que: où et quand ces migrants sont-ils arrivés pour la première fois sur la côte de Madagascar; pourquoi et quand se sont-ils déplacés vers l'intérieur des terres; comment ont-ils vécu et comment ont-ils interagi avec d’autres populations sur l’île; et pourquoi et quand le site a-t-il été déserté?

Introduction

Madagascar is situated in the southwestern Indian Ocean and is considered one of the last big islands to have been settled by humans. There is still considerable debate on the origin of the Malagasy population and when the first settlers arrived. Contributions of African, Asian, Indian, Austronesian, Arab and Persian populations have been suggested based on physical and cultural anthropology, oral traditions, linguistics and archaeology (e.g. Heiske et al. Citation2021). Genetic studies have identified African and Asian populations as the main contributors, yet with pronounced regional variability in the relative contribution of these two ancestries (Heiske et al. Citation2021). Bones with tool marks believed to indicate human butchery have been dated between approximately 10,000 and 1000 years BP and have been used to indicate an early human presence on the island (e.g. Gommery et al. Citation2011; Hansford et al. Citation2018, Citation2020). However, several authors have challenged reported evidence of human butchery prior to c. 1200 BP (Anderson et al. Citation2018; Mitchell Citation2019, Citation2020, Citation2022). Archaeology has so far only documented a relatively late settlement of Madagascar during the mid- to late first millennium AD, with possible exceptions provided by excavations at a rock shelter in the Velondriake area along the coast of southwestern Madagascar, where dated avian eggshell suggests a periodic campsite for mobile foragers at least 2000 cal. BP (Douglass et al. 2019), and others at two rock shelters in northern Madagascar, where Dewar et al. (Citation2013) infer the presence of stone tool-using foragers at least as early as 2000 BC. shows the location of archaeological sites dating to the period AD 700–1100 (Battistini and Vérin Citation1966, Citation1971; Wright and Fanony Citation1992; Dewar and Wright Citation1993; Radimilahy Citation1998, Citation2016; Rakotoarisoa Citation1998; Parker Pearson et al. Citation2010; Serneels et al. Citation2018, Citation2019, Citation2020, Citation2021; Anderson Citation2021).

Teniky (also known to as ‘Tenika’) is an inland site with enigmatic rock-cut architecture in central-southern Madagascar, some 200 km from the Mozambique Channel to the west and 275 km from the Indian Ocean to the east (). It is located in the heart of the Isalo National Park and can only be reached on foot, requiring a 20 km hike across rugged terrain dissected by steep canyons. Although the presence of archaeological structures at Teniky has been known for well over 100 years, until now they were only known from a fluvial cirque (i.e. an amphitheatre-shaped valley formed by fluvial erosion) that is partially enclosed by approximately 30–40-m-high cliffs and open to the southwest (). Paulian and Dommergues (Citation1950), Ginther and Hébert (Citation1963) and Ramilisonina and Rakotoarisoa (Citation1971) have succinctly described archaeological structures within this cirque, which include man-made terraces, dry stone walls and rock-cut niches of various sizes and shapes.

The best-known and most prominent structure at the cirque of Teniky is the ‘Grande Grotte’ ( and ), a large rock shelter delimited by neatly constructed sandstone walls (Ginther and Hébert Citation1963). In the early twentieth century, French naturalists Alfred and Guillaume Grandidier suggested that shipwrecked Portuguese sailors constructed the walls in the early sixteenth century AD (Paulian and Dommergues Citation1950) and the rock shelter is also known as the ‘Grotte des Portugais’. In their attempts to reach a port where Portuguese ships might be present they traversed the island, stayed in the Isalo Massif and supposedly built the walls at Teniky (Paulian and Dommergues Citation1950). Ginther and Hébert (Citation1963) doubted this interpretation, particularly because of the considerable effort that must have gone into constructing the numerous terraces within the cirque. They dug a trench at the Grande Grotte without being precise on its exact location and extent and reached a reddish soil at a depth of 40–50 cm. No archaeological objects were recovered from the trench. However, they did find a few pottery sherds on the slopes of the cirque and Vérin (Citation1986) mentions that one of these was considered by John Kirkman (who excavated at Gedi on the East African coast in the 1940s and 1950s; Kirkman Citation1954) to be part of a Chinese jar of the sixteenth century AD.

An inspection of high-resolution satellite images of Teniky in 2019 revealed that the extent of the archeological site was much larger than previously known. The images showed the presence of a large number of terraces and linear and rectangular structures on a north-south trending hill west of the Sahanafo River (, Zone 2) some 2.5 km from the cirque. These new findings, and the fact that no detailed archaeological excavations had ever been carried out at Teniky, motivated us to carry out an archaeological survey of this intruiging site in an attempt to constrain the timing of its occupation and to better understand the rock-cut architecture in order to make inferences about the origin and culture of those who once lived at the site.

Here, we present results of field surveys and excavations carried out at Teniky in 2021 and 2022. During field prospecting in 2021 we discovered a new site with rock-cut niches under a rock overhang on the western flank of the north-south trending hill west of the Sahanafo River hidden from view by dense vegetation (GPS co-ordinates: 22°17’53.9”S, 045°17’10.9”E; ). The niches have been cut into the rock over a length of about 25 m. Remnants of walls consisting of carved sandstone blocks were visible in front of the main niche, with the blocks having a similar composition and dimensions as those making up the sandstone walls at the Grande Grotte. In 2022, we continued surface prospecting and carried out excavations in front of the newly discovered niches.

For descriptive purposes, we distinguish eight zones of archaeological interest at Teniky (). The rock-cut niches discovered in 2021 are located within Zone 1, which forms part of the north-south trending hill with terraces and dry stone walls (Zone 2). Zone 3 is situated just east of the Sahanafo River and includes dry stone walls in a forested area of ∼80 × 80 m, a 120-m-long east-northeast trending dry stone wall in its northern part and a 750-m-long north-northeast trending dry stone wall in its eastern part. Zone 4 and 5 include sandstone quarries respectively discovered on the western and eastern side of the Sahanafo River. Zone 6 contains a nearly closed circular, dry stone wall with a diameter of approximately 150 m visible on satellite images, whereas Zone 7 comprises terraces and dry stone walls on a ridge overlooking the Sahanafo Valley, some 600 m north of the cirque. Finally, Zone 8 covers the cirque with the archaeological structures, including rock-cut architecture, terraces and carved sandstone walls, known since at least the late nineteenth century.

During our field surveys, we observed six stone tombs, mostly on ridges at higher elevations (the purple squares in ). The approximately square stone tombs are about 1 to 1.5 m high and sometimes several stone tombs occur next to each other, with a total surface area up to about 10 m² (inset in ). All the tombs showed signs of destruction and looting. According to our local guides, these tombs are associated with the Sakalava ethnic group, which inhabited the region before being driven out by another ethnic group, the Bara, probably in the nineteenth century. Mid-twentieth-century maps of the Isalo region show the presence of a village called Sahanafo (or Sahanafa) on the east bank of the Sahanafo River. When the Isalo National Park was established in 1962, the inhabitants of Sahanafo were relocated outside the park’s limits and the village was abandoned. The village’s exact location is unknown, but it was probably within or close to Zone 3. Remains of dry stone walls were observed in the field at or near its presumed location (Zone 3). However, it remains unclear when and by whom these walls were constructed and whether they do indeed form part of the former village of Sahanafo.

We first describe the known archaeological structures at the cirque in Zone 8. This is followed by the results of our excavations at the rock-niches in Zone 1 and the results of surface prospecting in Zones 2 through 7. Subsequently, we provide an interpretation of our observations and their significance in the discussion and conclusions to this paper.

Archaeological structures within the cirque (Zone 8)

Satellite images of the cirque show the presence of man-made terraces constructed on the slopes of the cirque and covering an area of about 20 hectares (; Zone 8 in ). The terraces vary in size and shape, with the largest having a surface area of about 1200 m². Field prospecting reveals that most of these terraces have one or more dry stone walls on the downward valley side. A major northeast-southwest trending dry stone wall (up to 1 m wide and 2 m high) follows the ridge of the cirque on its northwestern side, with a similar wall running north-northwest/south-southeast on a hill slope further south (). Below, we describe and illustrate the main archaeological structures within the cirque (), most of which have previously been mentioned by Paulian and Dommergues (Citation1950), Ginther and Hébert (Citation1963) and Ramilisinona and Rakotoarisoa (1973).

Excavations at the rock shelter in Zone 1

In late 2022 we carried out excavations at the rock shelter in Zone 1 discovered the year before (). The shelter is curved in plan view and eight quadrangular niches (N1–8) have been cut in the cliff wall (b). A ninth niche (N9), with a near-circular opening, is located a few metres away, towards the west (inset in b). In addition to the niches, we found further archaeological structures at the foot of the rock shelter, including (i) two sandstone walls (W1 and W2) in the northeast part; (ii) an area of approximately 4 × 3 m in front of the sandstone walls, partly outlined by rock slabs placed vertically in the ground (ST14; c); and (iii) a stone basin (ST18) about 3 m away from the cliff, close to the rock-cut niches N2-N8 in the southwestern part of the rock overhang.

After clearing the site of dense vegetation, excavations were undertaken close to the niches. We started with trenches in front of niche N1, where most structures seemed to be concentrated: SD001 (surface area: ∼3 × 2 m), SD002 (6 × 1.5 m), SD003 (1 × 1 m) and SD004 (3.5 × 2.5 m). In the areas covered by SD001 and SD002, we dug two deeper trenches, SD009 and SD008, to investigate the stratigraphy at a larger depth. In front of niches N2-N8, we dug two small trenches: SD005 (2 × 1.5 m) and SD007 (1 × 1.5 m).

Surface prospecting

In addition to the excavations in Zone 1, we carried out surface prospecting in the Sahanafo Valley and on the hills on either side of it to obtain a first inventory of the wider archaeological landscape. Satellite images were of great help in guiding our surface prospecting.

Discussion

Our archaeological survey documents a rich archaeological landscape at Teniky covering an area much larger than previously known. Archaeological structures are found in the Sahanafo valley and on hills on either side, covering nearly 8 km². New finds include rock-cut niches and stone walls at a rock shelter (Zone 1), where detailed excavations were carried out. Radiocarbon ages on charcoal indicate that the stone walls (W1 and W2) at the rock shelter were constructed close to the turn of the first and second millennia AD.

The walls at the rock shelter in Zone 1 are carefully constructed with squared sandstone blocks extracted from nearby quarries in the Sahanafo Valley. They delimit a slightly elevated space before a large rock-cut niche (N1) with a raised bench and carved pillar-like columns. Seven further rectangular niches (N2–N8) with a raised bench occur a few metres away. Considering their dimensions and the considerable effort that must have gone into building the walls and carving the niches (N1–N8), it seems unlikely that they served as storage rooms or dwellings but rather that they had a ritual function. A ninth smaller and nearly circular niche is slightly set apart from the rock shelter. Its small size and position, coupled with the presence of grooves around the inner rim of the opening indicating that it could be closed off, suggest that this niche probably had a different ritual function than the others. The morphology and character of a nearly square stone basin, a few metres away from the niches, suggest a symbolic, ritual use rather than a functional one. Although poorly preserved, aligned rock slabs were placed vertically in the ground directly in front of the stone walls, probably initially delimiting a small, roughly 10 m² rectangular space that has since been partly eroded. Its spatial proximity to the rock shelter’s walls and niches suggests that the function of the space is directly related to the latter. Given the overall limited horizontal space in front of the rock shelter with a steep slope just below and the near complete absence of any material culture or structure indicating a domestic use (only one ceramic sherd was found), it is unlikely to represent a former living site. Instead, we consider it likely that the entire assemblage of structures observed at the rock shelter in Zone 1 was used for ritual activities.

The presence of a 40 × 15-m-large raised platform bordered by dry stone walls at the highest point of the north-south trending hill west of the Sahanafo Valley, and the find of a small sandstone basin within it, suggest that this site was probably also used for occasional, ritual activities, perhaps linked to those presumed to have taken place at the rock shelter in Zone 1.

Considering the dimensions, location and character of the archaeological structures in the cirque east of the Sahanafo Valley (Zone 8), we infer that most of the rock-cut structures there also had a ritual function. It is unlikely that the small rock-cut chamber (the Petite Grotte) with its beautifully carved pillars and benches represents a former storage room, while its small size and height of only 1.2 m makes it unsuitable to live in. Instead, it is more likely that the chamber had a ritual function. Most rock-cut niches in the cirque are considered to have had a ritual function, including the semi-circular and rectangular niches and the niche with a raised bench inside the rock shelter (the Grande Grotte). Rock-cut niches and rock-cut boulders with recesses around the openings may have had a different function than those without recesses. Furthermore, we note the presence of a terrace with a square stone basin and vertical sandstone slabs placed vertically in the ground. Its location immediately adjacent to the cliff wall with niches suggests that it could also have had a ritual use.

The sandstone blocks of the walls in Zone 1 are similar in size and composition to those used in constructing the sandstone walls delimiting the rock overhang at the Grande Grotte in the cirque (Zone 8). Although we have no absolute ages for the latter walls, it seems reasonable to assume that they are from the same period. Besides the radiocabron ages on charcoal from Zone 1 indicating a human presence approximately 1000 years ago, evidence for the former presence of an ancient population at Teniky is also provided by the finds of imported Chinese and Southeast Asian sherds dated to between the eleventh and fourteenth centuries AD (Bing Zhao, pers. comm.). Man-made terraces at Teniky cover a total area of approximately 30 hectares on the hills and ridges west and east of the Sahanafo River, suggesting that Teniky was a fair-sized settlement. Several terraces are directly at the foot of the cliffs, where rock-cut structures have been carved. The presence of large rock-cut boulders and stone basins on several terraces suggests that the levelling of the terraces was coeval with the carving of the rock-cut structures. Thus, in summary, we find evidence that most of the archaeological structures at Teniky, including the stone walls, quarries, terraces, niches, rock-cut boulders and stone basins, form part of a coherent archaeological assemblage made by a specific group of people who lived there about a 1000 years ago.

Teniky is a unique archaeological site in Madagascar as no similar rock-cut architecture is known from anywhere else on the island or from the adjacent East African coast. Given the limited excavations carried out so far and considering the relatively few pieces of ceramics found, we cannot yet say much about the duration of the human occupation at Teniky. We can, however, rule out the possibility that shipwrecked Portuguese sailors made the structures at Teniky in the sixteenth century as previously proposed because Portuguese ships did not sail the Indian Ocean until 1498. Although we did not find one single piece of evidence for a Portuguese presence, it is nevertheless still possible that Portuguese mariners stayed for some time at Teniky during an attempt to traverse the island.

The question thus arises as to who was responsible for the rock-cut architecture and terrace construction at Teniky, and why these people chose to live in such an isolated and secluded environment over 200 km from the nearest coast? The presence of early second-millennium Chinese and Southeast Asian ceramics indicates their participation in Indian Ocean trade networks, but does not give information on their origin. Does the entire assemblage of archaeological structures at Teniky warrant a first attempt at inferring the ethnicity and/or religious affiliation of those who once lived there?

Apart from the possible — but contested — early presence of foragers (Dewar et al. Citation2013; Douglass et al. 2019; Wright et al. Citation2022), archaeological excavations suggest that humans did not arrive on the island of Madagascar until the mid- to late first millennium AD (Anderson et al. Citation2018; Mitchell Citation2019, Citation2020, Citation2022). At some stage early migrants moved into the island’s interior and settled at Teniky in the tenth/eleventh centuries. Although it cannot be fully excluded that they developed their own unique culture and traditions after arriving on the island, the relatively short time interval between arrival on the coast and inland migration makes it more likely that they introduced their culture and traditions from elsewhere. In the following, we present a preliminary hypothesis on the origin and beliefs of the people who once lived at Teniky, but at the same time should like to clarify that our hypothesis must be validated or rejected based on further field studies.

When comparing the approximately 1000-year-old rock-cut niches at Teniky with similar structures elsewhere in the Indian Ocean, the closest stylistic parallels are found in Iran, where a multitude of archaeological sites show niches cut in steep rock walls dating to the first millennium AD or earlier (e.g. Grenet Citation1984; Boucharlat Citation1991; Cereti and Gondet Citation2015; Farjamirad Citation2015). Notably, in the Fars region in southern Iran hundreds of niches of different sizes and shapes have been cut in steep cliff walls, for example at the city of Siraf (). Siraf was a major and prosperous port on the Persian Gulf during the mid/late first millennium. Whitehouse (1971) considers that the port was already of importance in Sasanian times (i.e. before AD 656), although this has recently been questioned by Priestman (Citation2022). In any case, both authors agree that Siraf was an important port city in early Islamic times in the ninth and tenth centuries, with ships actively participating in the Indian Ocean trade network and sailing as far as India, Southeast Asia and East Africa (Whitehouse 1971; Whitehouse and Williamson Citation1973). Extensive archaeological excavations were carried out at Siraf by the British Institute of Persian Studies between 1966 and 1973 (Whitehouse Citation1968, Citation1969, Citation1970, 1971, Citation1972, Citation1974; Whitehouse et al. Citation2009) and, although these studies mostly focused on a Sasanian fort and Islamic monumental houses and buildings near the shore, cemeteries in the hills behind the city were also investigated (Whitehouse Citation1972, Citation1974; Whitehouse et al. Citation2009). Whitehouse mentions the presence of Islamic tombs on the southward-dipping slopes behind the city of Siraf, but also describes the presence of numerous niches (Whitehouse refers to them as ‘chambers’) cut in vertical cliff walls on the slopes and in steep gorges traversing the hills (). These niches are up to 2 m wide and 1.5 m high, sometimes including a rock-cut bench, and are considered by Whitehouse (1971, Citation1972) too small and inaccessible to have been dwellings. Several rock-cut niches contain a scatter of unburnt human bones and Whitehouse (1971, Citation1972) interprets these as tombs. Although single niches occur, several are usually closely grouped (e.g. a–e). Besides the large niches, Whitehouse (1971, Citation1972) also describes and illustrates smaller, rectangular niches (Whitehouse Citation1972: Plate Ia; f–i), sometimes set in a recessed panel retaining plaster, and interprets them as formerly closed-off ossuaries of a Zoroastrian community, intended to receive the bones of the deceased after exposure of the corpse.

Zoroastrianism is one of the oldest monotheistic religions in the world, practised without interruption to the present day, and was the dominant state religion of the Persian Sasanian Empire (AD 224–656). Zoroastrian funeral rites do not allow direct burial in the ground (Boyce Citation1979). Instead, corpses are left in places of display above the ground, generally called dakhmas in Pahlavi, often on hilltops or mountains, with excarnation of the corpses usually done by animals, such as dogs and vultures. After excarnation, the bones are dried and placed in ossuaries (bone receptacles, called astōdans in Pahlavi) to avoid polluting the earth. After the conquest of the Sasanian Empire by the Arabs in the mid-seventh century, Islam was imposed, but for a long time Zoroastrianism and other religions coexisted with it (Boyce Citation1979; Choksy 1987; Morony Citation2013). The gradual conversion of the majority of Persians to Islam occurred first in the cities and then in the countryside between the eighth and tenth centuries, with a peak in conversion in the ninth (Choksy 1987). However, Buzurg ibn Shahriyar (Citation1981) mentions that Zoroastrians still lived in the city of Siraf in the tenth century. The conversion of Persian Zoroastrians to Islam was, however, never fully accomplished and a small group of Zoroastrians still lives in Iran today. Historical documents, archaeological excavations and genetic studies indicate that some Zoroastrians left Iran and settled in western India in the late eighth century (Nanji and Dhalla Citation2008; Chaubey et al. Citation2017; Dalal and Mitra-Dalal Citation2021; Kumar et al. Citation2023), where they have become known as Parsis, i.e. the people of ‘Pars’ or ‘Fars’, their region of origin in southern Iran.

Notwithstanding the large variety in shape and size of the rock-cut niches in both Teniky and Siraf, the following similarities can be noted between the two sites: (i) larger niches are often closely aligned next to one another with a relatively narrow division between adjacent niches, giving rise to a pillar shape in frontal view; (ii) the openings of the larger niches are either rectangular or arched with a flat base; (iii) there is always a vertical distance between the base of the niche and the ground, i.e. none of the niches reach the ground; and (iv) some of the smaller rock-cut niches show recesses, suggesting that they could be closed. Apart from the stylistic resemblances of niches at Teniky and Siraf, the fact that at both places the openings in the cliffs never reach the ground is of importance, as this follows Zoroastrian funerary rites, according to which the exposed body must not touch the ground and must be protected from the natural elements to prevent pollution. Besides the abovementioned similarities, there are also differences: as far as we can tell from the literature, no nicely carved stone walls have been reported close to rock-cut niches in southern Iran and neither is there evidence for vertically placed slabs in the ground in front of the niches. However, this might be related to the fact that only limited archaeological research has been done on the niches at Siraf and other sites in southern Iran.

Despite the fact that no traces of human bones have yet been found in any of the niches at Teniky, we tentatively interpret the assemblage of rock-cut structures, stone walls and stone basins as part of a former necropolis the characteristics of which suggest a link to Zoroastrian funeral ceremonies. The large rectangular rock-cut niches may have been dakhmas where the bodies of the dead were exposed. The smaller, circular niches with recesses might then represent astōdans in which the bones of the dead were placed with the openings being closed off by a wooden or stone slab to protect them from the rain and thus prevent them from polluting the earth. The Petite Grotte and the large niche in the Grande Grotte are tentatively interpreted as chamber tombs, where bones stored in astōdans could have been placed on the raised benches along the sides. The two large worked rock boulders found on the slope of the cirque possibly represent larger sarcophagus-like astōdans used to dry or store bones. The openings on all four sides of the boulders have recesses, suggesting that these openings could have been closed in case of precipitation. Differences in the style, size and shape of the inferred astōdans might reflect differences in the social status of the dead.

It is not only the rock-cut architecture at Teniky that points to Zoroastrian rituals, but also the presence of stone basins and tables with rectangular or arc-shaped recesses and openings. These show stylistic similarities with the stone basins, tables and platforms used in Zoroastrian ritual ceremonies (Darmesteter Citation1892; Boyd and Kotwal Citation1983), for example to hold water or fire, both agents of ritual purity. Darmesteter (Citation1892) and Boyd and Kotwal (1893) describe and illustrate the Urvisgah (or Urwīsgāh), a part of the Zoroastrian fire temple complex, where liturgical ceremonies are performed. Plate IV in Darmesteter (Citation1892) shows the Urvisgah of the Maneckji Seth Fire Temple in Mumbai, which was constructed in the first half of the eighteenth century (). The plate shows low stone tables for holding utensils, square stone platforms for the priests to sit on cross-legged and slightly higher stone tables for holding a fire vase. In addition, the plate portrays a low square table with a circular depression, possibly for holding water used during purification ceremonies.

Only a few primary sources discuss the East African coast at the turn of the first and second millennia AD and even fewer make a reference to what is most probably the island of Madagascar. The latter include Buzurg Ibn Shahriyār, a tenth-century Persian sailor and writer, who collected narratives from sailors based in port towns on the Persian Gulf, including Siraf (Trimingham Citation1975; Buzurg Ibn Shahriyār Citation1981). His narratives suggest that Persian contacts with Madagascar existed in the mid-tenth century or earlier. Although Persian trade goods such as ceramics from southern Iran dating to the ninth to twelfth centuries have been found at archaeological sites in Madagascar, for example at Kingany (Anderson Citation2021), it remains difficult to infer a Persian presence from material culture only. The site of Teniky might represent an exception. If the rock-cut architecture and associated stone basins at Teniky are the work of a Zoroastrian community that settled in the Isalo Massif about a thousand years ago, this would strongly point to a former Persian presence in southern Madagascar.

At the moment we can only speculate as to when and where the group of people who ultimately moved to Teniky first landed on the coast of Madagascar and whether they also settled at other coastal or inland sites. The nearest potential first-millennium coastal site is Sarodrano on the shores of the Mozambique Channel, 200 km away from Teniky (). Here, excavations were undertaken in 1966 by Battistini and Vérin (1975) and a single radiocarbon date on charcoal (GAK-928) yielded an uncalibrated age of 1460 ± 90 BP, corresponding to cal. AD 418–841 using Oxcal 4.4 and SHCal20. Unfortunately, however, the site that yielded the date was destroyed by a tropical cyclone in 1970 (Battistini and Vérin 1975).

The inland sites of Asambalahy and Rezoky () are only 60 and 80 km to the west and west-northwest of Teniky, respectively, and were excavated by Vérin (1971). They yielded imported Persian and Chinese ceramics, as well as iron slag. Vérin (1971) proposed that they were occupied between the fourteenth and seventeenth centuries. However, Dewar and Wright (Citation1993) proposed a revision of the dates for these sites, suggesting that they were at least as old as the thirteenth century. Regional surveys and absolute age dating of Asambalahy and Rezoky are needed to determine whether any links might have existed in the past between them and nearby Teniky.

The former inhabitants of Teniky lived along the Sahanafo River, which joins the Malio, a tributary of the Isahena River, with the latter joining the Mangoky River near the small town of Beroroha, seventy kilometres north-northwest of Teniky. The Mangoky is the largest river in Madagascar and flows from the central highlands in a westerly direction to enter the Mozambique Channel fifty kilometres north of Morombe (). A large part of the Mangoky is navigable by small boats such as pirogues and might have been used in the past for inland reconnaissance and human migrations. From an archaeological point of view the Mangoky’s banks and the areas adjacent to them are underexplored, with regional surveys needed to determine whether any early inland settlements can be located. In this context, it is worth noting that excavations at a cave with rock art in the Makay Massif (), north of the Mangoky River, have yielded a radiocarbon carbon age on charcoal of 860 ± 30 BP (GifA-11109/SacA-24631; Gonthier et al. Citation2020a, Citation2020b), which corresponds to a calibrated age in the range cal. AD 1180–1276 (using OxCal 4.4 and SHCal20). However, it is unclear whether any material culture was directly associated with the excavated site. It would be interesting to compare the cave paintings from the Makay Massif with those documented by Rasolondrainy (Citation2012) at the rock shelter of Ampasimaiky in the southern part of the Isalo Massif, 120 km south of Teniky (). Rasolondrainy (Citation2012) proposed that some of the signs painted on the walls of the rock shelter resemble Libyco-Berber writings or inscriptions and inferred a pre-ninth-century AD human presence.

Dry stone walls are a prominent feature at Teniky, in particular in the Sahanafo Valley, where a stone-walled enclosure is situated close to the Sahanafo River with rectilinear stone walls north and east of it (Zone 3, ). Impressive stone-walled enclosures (called mandas) are also known from coeval sites in southern Madagascar such as Andranasoa and Andaro (). They are dated between the tenth and thirteenth centuries on the basis of radiocarbon dating and the presence of imported sgraffiato and Chinese ceramics (Parker Pearson et al. Citation2010). It is clear that further comparative studies between Teniky and the manda sites are necessary to determine whether they show similar archaeological features and whether their inhabitants shared a common culture.

Further south at the coast, Triangular Incised Ware (TIW) pottery has been found on the west and east bank of the Menarandra rivermouth (at Enijo, ; Parker-Pearson et al. Citation2010). This pottery style is from approximately AD 600–1000 along the southern Swahili coast of East Africa and indicates first-millennium interactions between Madagascar and East Africa (Parker-Pearson et al. Citation2010). So far, Enijo is the only site with TIW pottery in Madagascar and it remains unclear whether its occupants interacted with the manda sites or others such as Teniky further north.

At the moment it is not known whether Teniky had any connections with contemporaneous sites in northern Madagascar, such as Kingany (Anderson Citation2021), Mahilaka (Radimilahy 1981), Vohemar (Wright et al. Citation2022), Nosy Mangabe (Dewar and Wright Citation1993) or those in the Mananara Valley (Wright and Fanony Citation1992). More archaeological research, including detailed comparative analyses of local and imported pottery, is needed to determine the role that Teniky and its inhabitants played within Madagascar and the Indian Ocean.

Conclusions

Our archaeological excavations and field prospecting at Teniky reveal a much larger and more important archaeological landscape than previously known. This includes many newly found archaeological structures, including terraces, stone walls, stone basins and rock-cut structures in various sizes, shapes and forms constructed in the late first/early second millennia AD. No similar assemblage of archaeological structures is known anywhere else in Madagascar or East Africa. Our investigations of the rock-cut architecture in and near the cliff walls at Teniky suggest that most of them had a ritual function and we tentatively interpret them as part of a necropolis. The rock-cut niches at Teniky show similarities to those known from various sites throughout Iran, dated to the first millennium or older and related to Zoroastrian funerary practices. We cannot exclude the possibility that the archaeological structures at Teniky were the work of a group of people whose specific rites and beliefs developed and evolved after their arrival on the island and whose rock-cut structures show, by chance, formal similarities to Zoroastrian ones in Iran. However, we consider it more likely that the people who arrived on the coast of Madagascar and ultimately settled at Teniky introduced their rites and beliefs from outside the island and continued to practise them in a similar way while there. Independent of the origin, religion and funerary rites of Teniky’s former inhabitants, further archaeological studies are required to fully situate it within Madagascar and the western Indian Ocean. Questions that need to be addressed include: Where and when did the settlers first arrive on the coast of Madagascar? Did they come straight from their region of origin, or did they first settle in other parts of the Indian Ocean? Why did they move inland to the isolated inland site of Teniky? How did they live, how long did they stay and why was the site deserted? What was the relation of Teniky with contemporaneous sites on Madagascar and what was its role in the Indian Ocean trade network? We hope that future research may help resolve these questions.

Acknowledgments

We should like to thank the Ministry of Communications and Culture of Madagascar for granting us a research permit (309-2022/MCC/SG) to carry out excavations at Teniky. We express our gratitude to the late Dr Mamy A. Rakotoarijaona (Director General of Madagascar National Parks), Dr Harinaina Léon Razafindralaisa (Director General ad interim of Madagascar National Parks) and Nandrarana Jean-Jacques Rakotoarivelo (Director Isalo National Park) for their support of our research. Prof. Lucien M.A. Rakotozafy, Director of the Institute of Civilizations/Museum of Art and Archaeology (ICMAA) of the University of Antananarivo and Dr Bako Rasoarifetra (ICMAA) provided administrative and logistical help. We thank Prof. Dr Frank Preusser of the Universität Freiburg, Germany, for examining the samples taken for OSL dating, Dr Bing Zhao for giving the first age estimates for the ceramic sherds, Ali Aghanjazadeh for providing us with drone images and photographs of the rock-cut structures at Siraf and Hoda Moradi for acting as an intermediary. Finally, we should like to thank two anonymous reviewers for their constructive comments and suggestions. This work was supported by the Burgergemeinde Bern (grant numbers 2021-284 and 2022-736).