1.1 Linguistics

The Dutch merchant Frederick de Houtman van Gouda was the first to notice that the Malagasy natives speak a language “very similar to Malay” [2]. A decade later the Portuguese Jesuit Luis Mariano arrived to the identical conclusion noticing a similarity between the Malagasy speech and the Malay languages [3]. Subsequently, the work of Herman Neubronner van der Tuuk [4] established beyond doubt the relationship between Malagasy and other Indonesian languages (but he wrongly proposed a close relationship with Toba Batak).

Finally, the incontestable link to a precise Indonesian language is due to the Norwegian missionary Otto Christian Dahl (1903-1995) who begun his apostolic mission in Madagascar in 1929 and later, in 1935, embarked on linguistic studies (a short bibliography in [5]). His missionary vocation and his interests in linguistics immediately found common ground [6] and some years later he published his fundamental work [7] where he firmly established a striking kinship between Maanyan, spoken in the South-East of Kalimantan, and Malagasy (see also [8] and [9]). His main conclusion was summarized by himself as: “Le Maanjan et le malgache sont des languages très etroitement apparentées. Parmi les langues indonésiennes connues actuellement aucune n’a autant de ressemblances avec le malgache que le Maanyan” [7]. This statement after 70 years is still undisputed.

In his 1951 paper Dahl also proposed, on the basis of historical and linguistic evidence, that the date of the landing event was 400 CE. His dating was shortly confirmed by Isidore Dyen who introduced the lexicostatistics grouping and dating methods into the Malagasy debate [10].

In [11], for the first time, it was considered that Malagasy should not be considered a single language but a constellation of dialects well different each other. Subsequently, by considering Swadesh 100 words list of sixteen different varieties, the authors of [12] were able to perform a lexicostatistical and glottochronological study concluding that the most diverse dialects were those in the North and that the Island was first colonized around two thousand years ago. These possibly wrong conclusions were a consequence of the fact that the research was performed without the help of the modern algorithms for building trees, the analysis of their data by recent tools brings to different conclusions [13]. The Vérin et al. dating for the landing event was confirmed in [14] on the basis of archaeological findings, but in this second paper it is admitted that the remains cannot be attributed with certitude to the ancestors of modern Malagasy.

The discovery of the Maanyan connection correctly collocated Malagasy at its place among East-Barito languages (so called after A. B. Hudson [15]), but it did not resolve the relation with the other Austronesian languages. Dahl himself addressed this problem [8] and occasionally traced Malagasy words to Ngaju [7, 9], but the first systematic study of non-Maanyan connections was performed by Alexander Adelaar. This author first considered the relations with Malay [16], and later also Javanese, Ngaju and Malay variants as Banjarese [17–19].

A careful timing of borrowings, which is derived from the known history of the Indian Ocean, allowed Adelaar to state that the founding event occurred between 600 CE and 700 CE. In [16] he wrote “Dahl (1951) used the presence (in his counting) of 30 Sanskrit loanwords in Malagasy as evidence that the migrations to East Africa must have taken place after the introduction of Indian influence in Indonesia. The oldest written evidence of Indian presence in this area is a Sanskrit inscription from around 400 AD found in Kutai, South East Borneo. Dahl therefore proposed the 5th century AD as the most likely migration period. In my 1989 article I interpret the Sanskrit influence in a different way. All but one of the Sanksrit loanwords in Malagasy have corresponding forms in Malay and/or Javanese. Moreover, many Sanskrit loanwords show the same phonological adaptations as their Malay and Javanese counterparts; I therefore conclude that the loanwords in question were actually borrowed via Malay or Javanese. As a consequence, the migration date should not be correlated to the beginning of Indianisation in the archipelago, but more specifically to the emergence of Sanskrit influence on Malay. This influence was the manifestation of an Indian Malay civilization, which was evidenced for the first time in the emergence of the maritime polity of Srivijaya in the seventh century AD in South Sumatra”.

Adelaar also remarked that “Malay loanwords in Malagasy often pertain to a maritime environment (which includes names of winds and directional terms)” and that the Maanyan speakers live along the rivers of Kalimantan, hence whithout having the necessary skills for long-distance maritime navigation. He proposed as a possible explanation that the Maanyan speakers constituted the crew of expeditions led by Malay sailors.

This is not the end of the story, since Adelaar also provided some additional evidence that a number of words in Samihim (a Barito language), which do not occur in Maanyan, have cognates in Malagasy [20]. Moreover, in [21] he highlighted the relevance of lexical borrowings from South Sulawesi languages, while rejecting the hypothesis of borrowings from Philippine proposed in [22].

To complete this unexpected complex picture of the lexical sources of Malagasy, Adelaar also described the influences of Bantu languages [23]. Indeed, this topic was already addressed in 1954 by Dahl, which wrote about the influence of Comorian Bantu over Malagasy especially in phonology [24], and later by Blench and Walsh in [25, 26] and [27], which focused on domesticated animals and mammals names. Philippe Beaujard [28] argued that the first migrants Indonesian came into contact with Bantu speakers only after their arrival in Madagascar, probably at the end of the first millennium CE.

Some topics discussed above were also considered in [13] (see also [29]) through the application of new quantitative methodologies inspired by, but nevertheless different from, classical lexicostatistics and glottochronology. All approaches in these papers converge to the conclusion that Malagasy dialects are classified into two main geographical subfamilies: South-West and Center-North-East. Moreover, a date of landing was determined around 650 CE, in agreement with the proposition made many years before by Adelaar [16]. Finally, by means of a technique which is based on the calculation of differences in linguistic Diversity proposed in [30] and which is a quantitative implementation of a well known argument in linguistics and biology [31, 32], it was argued that the landing took place on the South-East coast of the Island. The South-Eastern landing was also proposed by various historians and anthropologists (see, for example, [33]).

Adelaar in [34] refutes the hypothesis in [13], and argues that the South-East landing “.. is not supported, as it concerns later migrations, taking place half a millennium after the first SEB speakers had arrived. If anything, these migrations must have a skewing effect on Serva et al.’s outcomes and on Sapir’s principle, rather than showing the place where SEB speakers first arrived and from where they dispersed all over Madagascar.”.

On the basis of both linguistic data and Geo-maritime arguments, other scholars arrived to the conclusion that the colonization started from North. This Northern hypothesis is clearly motivated in the passionate writings of Rory Van Tuyl [35].

1.2 Genetics

Although recent work in linguistics has highlighted the non-Barito contribution to the Malagasy, there is still unanimous consensus about its collocation among East-Barito languages. This view seems in contrast with recent findings of genetic research.

In [36] it is reported that the mtDNA Polynesian motif, which can be found in Malagasy individuals at a macroscopic rate, is found at a low rate in Borneo, not at all in the Barito river area. At higher rate it is found in Sulawesi and coastal Papua New Guinea, and it is predominant in Polynesia. Nevertheless, it is unlikely that this motif came to Madagascar from Polynesia by some direct genetic contact because Polynesian Y chromosome haplogroups have not been found among Malagasy paternal lineages [37]; more likely it came from Borneo, but not directly from Maanyan speakers ancestors.

In [38] and [39] it is shown that both maternal and paternal lineages of Malagasy point to multiple regional sources in Indonesia, with a focus on Southern and Eastern Borneo, but also Sulawesi and the Lesser Sunda islands. Surprisingly, the Malagasy do not exhibit a clear genetic connection with the Maanyan, despite the obvious linguistic association. The suggested explanation is that the “settlement may have been mediated by ancient sea nomad movements”. This opinion is also shared by [40] which also proposes that “the distribution of a musical instrument such as the valiha among the highland Merina, points to some sort of direct Sulawesi connection”. This may be true, but we would not attribute as much value to shared musical instruments as to linguistic evidence.

Finally, [41] found strong “support for an origin of the Asian ancestry of Malagasy among the Banjar. This group emerged from the long-standing presence of a Malay Empire trading post in South-East Borneo, which favored admixture between the Malay and an autochthonous Borneo group, the Maanyan”. The present day Banjar people speak a Malay language, but the authors argue that the ancestors of the Banjar spoke a language close to the reconstructed Proto-Malagasy, in turn, forcefully close to nowadays Maanyan.

For what concerns the African component of Malagasy genetic makeup, it is well known that it is approximately of the same size of the Asian component [42]. Nevertheless, in [43] it was shown that “the distribution of ancestral components was ethnic and sex biased, with the Asian ancestry appearing more conserved in the female than in the male gene pool and in inland than in coastal groups”. Similar conclusions concerning the African-Indonesian admixture were reached in [44], leading to important consequences: “The distribution of African and Asian ancestry across the island reveals that the admixture was sex biased and happened heterogeneously across Madagascar, suggesting independent colonization of Madagascar from Africa and Asia rather than settlement by an already admixed population. In addition, there are geographic influences on the present genomic diversity, independent of the admixture, showing that a few centuries is sufficient to produce detectable genetic structure in human populations.” On the other side, it should be also considered that the sex bias could be seen in the light of kinship structure, which is matrilineal. The same matrilineal (and not patrilineal) preponderance of Austronesian DNA is seen in the East of Indonesia [45].

Very recent research [46] found evidence of a weak but well detectable Malagasy genetic flow in Somalia and Yemen, which supports the existence of African-Malagasy contacts after the main Asian colonization event of the Island.

Finally, an extremely interesting contribution to the understanding of the colonization of Madagascar comes from [47], where it is proposed a “scenario in which Madagascar was settled approximately 1200 years ago by a very small group of women (approx. 30), most of Indonesian descent (approx. 93%). This highly restricted founding population raises the possibility that Madagascar was settled not as a large-scale planned colonization event from Indonesia, but rather through a small, perhaps even unintended, transoceanic crossing.”

1.3 Archaeology

One of the questions which still await a definitive answer is: there were someone around before the arrival of the ancestors of nowadays Malagasy?

A clear answer will hardly be provided by linguistics, it may be that some hint will come from genetics, but in principle only archaeology allows going back sufficiently deep into the past, unfortunately this discipline provides contrasting evidence.

In [48], it is reported that a sediment core from Lake Kavitaha, central Madagascar, provides a stratigraphic record showing a marked rise in charcoal about 1300 year BP followed by a decline in pollen of woody taxa, culminating in a change to grass-dominated pollen spectra within about 4 centuries. In a much more recent paper ([49]), it is shown that starting from 890 CE carbon stable isotopic data indicate a rapid, complete transformation from an open forest landscape to grassland system as a result of the impact that early inhabitants had on the environment. Other recent research ([50]) suggests that grassland formations on Madagascar are natural although it is admitted that it is possible that human disturbance may have resulted in a much larger modern extent of grassland than in pre-human settlement of Madagascar.

Similar dating is found out in [51]: “At Nosy Mangabe we have an apparently continuous occupation dating to the late eighth century […]. At present, there is no direct evidence for occupation of the interior during the first millennium CE”.

All dating in the above mentioned papers are compatible with a landing event in seventh century as proposed by prevalent linguistic research.

Nevertheless, in other research ([52]) it is claimed that the minimum age for initial human presence on the island may be set at approximately 2000 BP, on the basis of AMS 14C dates for human-modified femora of extinct dwarf hippos from South-West Madagascar. In more recent research [53], dating is further pushed to the past: “Multiple lines of evidence point to the earliest human presence at ca. 2300 years BP”, a dating which is also embraced in [54]. Moreover, in [55] it is reported that Lakaton’i Anja near Antsiranana, in the North, yielded several stratified assemblages that indicate occupation from at least 4000 years BP.

This race to the past seems to progress unabated, in fact, the authors of [56] report human-modified bones dating before 10500 BP predating all other evidence. This dating suggests prolonged human-faunal coexistence with limited bio-diversity loss. In [57], the authors underline that recent contributions yielded estimates of initial settlement which may differ as much as 9000 years. Their assessment is that the presence of people on Madagascar dates at least 2000 years BP, but also that an Early Holocene arrival cannot be rejected.

All these further dating cannot correspond to the Indonesian colonization performed by the ancestors of modern Malagasy but it could go back to previous inhabitants of Madagascar. This argument will be examined in more details in Section 3.

Some cautious rethinking about human presence on the island can be found in [58], where the authors underline the nonexistence of butchery marks prior of about 1200 years BP. Moreover, their close analysis of the Lakaton’i Anja chronology suggests the site dates 1500 years BP or more recent, while older dating is due to extensive bioturbation at the site [59]. According to the authors, these findings indicate initial human colonization of Madagascar 1350–1100 years BP.

An even more decided criticism to early presence of humans in Madagascar can be found in [60], where the author concludes that there is no compelling evidence that people were present on Madagascar before the mid-first millennium CE. All claims and proofs contrasting this conclusion are re-examined in detail and rejected on the basis of general criteria of archaeological validation.

This last paper [60] and the previous two [58, 59] lead again to a dating which fits the linguistics one for the main colonization event.

1.4 Preamble summary and the contributions of this paper

We have seen that much has been understood about the colonization of Madagascar, and also that much has yet to be unveiled.

Consensus seems to converge on the hypothesis of an arrival in the second half of first millennium CE, probably in seventh century. In this paper we confirm this dating by a new but simple argument. We assume people arrived by direct navigation from Indonesia, while Africans’ contribution to the Malagasy genetic pool and, at a much lesser extent, to the Malagasy language, took place at a later time.

The colonizers from Indonesia were a quite small and possibly heterogeneous group of people, mostly from South-East Borneo. Likely a large part of them were the common ancestors of Malagasy and Banjar people of Borneo. Banjar speak now a Malay variety, but it is argued that their ancestors spoke an East-Baritio language related to the modern Maanyan language, which is the closer language to Malagasy.

After arrival, the language started a process of diversification which led to modern varieties. The relationships among these varieties probably reflect the historical process of internal colonization. In this paper we produce cladograms by standard tools as UPGMA and NJ, but also we propose a new method of classification which better accounts for the fact that in a strongly inter-related network of varieties a simple genealogical description of language relations is inadequate.

We also propose that the colonization started from the South-East of the Island adapting a simple argument from linguistics (and biology) which identifies the location of maximum linguistic Diversity with the homeland of a family of languages. To reach this goal we improved the definition of Diversity given in [13], in order that it is defined in all geographical locations and not only in those places where varieties were collected.

The South-Eastern landing is compatible with a direct navigation from Indonesia, so that admixture with East-African may have taken place after the main Asian colonization event [28].

A very animated discussion among scholars is the eventuality that Madagascar was inhabited before it was colonized by the ancestors of modern Malagasy. The opinions are very different, ranging from proposals of an early occupation thousands of years BP to vigorous rejection of any human presence before the main colonization event in the second half of first millennium. It has been also argued that hunter-gatherer populations could show traces of an old component in their language different from modern Malagasy language. We don’t find any trace of this component in Mikea hunter-gatherers, who are often suggested being living descendants of the pre-existing inhabitants; more likely they represent a former neolithic people which went back to hunting and gathering for historical reasons. Nevertheless, we cannot exclude that an early occupation occurred leaving still undetected linguistic traces or, eventually, leaving no traces.

3.1 Classification of Malagasy dialects

The information concerning the vertical transmission of vocabulary from Proto-Malagasy to the contemporary dialects can be extracted from the matrix of genealogical distances by a phylogenetic approach. There are various possible choices for the algorithm for the reconstruction of the family tree, we show the tree generated by Unweighted Pair Group Method Average (UPGMA) and by Neighbor Joining (NJ).

In Fig 2 we report the UPGMA output. The leaves of the tree are identified by the name of the dialect which is followed by the name of the town/village where it was collected and whose location can be appreciated in Fig 1. The absolute time-scale (i.e., the value of τ in (3)) is calibrated by the results of next subsection, which indicates the date 650 CE for the start of diversification of Malagasy varieties. Nevertheless, the timescale it is irrelevant for discussion in this section since the shape of a tree remains unchanged when τ is modified.

The UPGMA tree shows a main partition of Malagasy dialects into two main branches (Center-North-East and South-West), this partition coincides with the one in [13] but it is at variance with the main partition of a previous study [12]. The difference is a consequence of the fact that the second paper [12] was written without the help of the modern algorithms for building trees, the treatment of their data by recent tools gives the same main partition which is found in this paper (see [13] for a discussion of this point).

In turn, each of the two branches splits into two sub-branches. The South-West branch, whose leaves are associated to yellow, blue and violet, splits in a large blue-violet sub-branch and a smaller yellow one for Antandroy dialects which show a remarkable isolation. It should be considered that part of the divergence of Antandroy dialects can be explained by the common/polite dichotomy concerning words for body parts. In fact, some southern variants, including Antandroy, have an ordinary word and a respectful one for body parts, but only in the Androy region the respectful words are of daily use (see [74] for an exhaustive investigation into the origins of Malagasy terms for body parts). This is why our policy of choosing the most employed word for a given meaning resulted in some respectful words recorded in the Antandroy lists. If by brute force we substitute the polite forms with the common ones, the Antandroy isolation decreases and the three variants get closer to the Mahafaly and Antanalana variants. Nevertheless, while reduced, the isolation remains remarkable.

The Antaisaka and Zafisoro variants (violet) cluster with the blue group, but we decided to assign a different color because they are slightly isolated from the other blue dialects, possibly because they are transitional toward Central and Eastern dialects.

The Center-North-East branch, whose leaves are associated to green, red and orange, also splits in two sub-branches, a green one for Eastern and Central variants and a red one for Northern dialects. Sakalava of Mahajanga (orange) is transitional, being a Northern dialect (Sakalava Boina) it is influenced by Central varieties (especially Merina, the official language) and also Western dialects (Sakalava Menabe).

There is a strict correspondence between the UPGMA cladogram and geography, this can be easily perceived comparing colors in Figs 1 and 2.

The cladogram resulting from NJ algorithm is reported in Fig 3. NJ substantially gives the same cladistic results, the only difference being that the Antandroy variants remain even more isolated from the others and that Sakalava of Mahajanga more decidedly clusters with Northern dialects.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. NJ tree obtained from the genealogical distances.

The classification is the same of the UPGMA tree, the only difference being that the Antandroy variants remain even more isolated from the others and that the Sakalava of Mahajanga more decidedly clusters with Northern dialects.

https://doi.org/10.1371/journal.pone.0240170.g003

Both UPGMA and NJ rely on the assumption that transmission is mostly vertical, so that a phylogenetic cladogram is fully justified. The main theoretical difference between the two algorithms is that UPGMA assumes that evolutionary rates are the same on all branches of the tree, while NJ allows differences. The question of which method is best suited to infer the phylogeny has been well studied and the general consensus is that NJ usually comes closer to the true cladistic (if it exists and it is mostly generated by philogenesys). Nevertheless, UPGMA more easily can be used to attribute dates since it assumes that the distance from the root is the same for all leaves, performing, in practice, a kind of average over different evolutionary rates. Luckily, in the present case no significant qualitative differences separate the two cladograms.

However, in Section 4 we show that relations among dialects are not necessarily tree-like, and we propose a totally different method for cladistic which privilege horizontal instead vertical transmission. Unexpectedly, the resulting cladogram is completely compatible with those in this section.

Finally, it is interesting to remark that the qualitative aspects of the cladograms are dominated by the least stable items, i.e., those items whose corresponding words undergo to a faster evolution (more often replaced or modified). This is not astonishing since many items are designated by identical or very similar words in all dialects so that they are totally useless for any kind of cladistic. By definition, these are the most stable items, while the least stable items are those designated by different words in different dialects so that they are very useful for cladistic.

We generated two UPGMA trees, the first (S1 Fig) from Swadesh lists with the 35 most stable terms, and the second (S2 Fig) from Swadesh lists with the 35 least stable terms. It can be observed that while the first tree is totally different from the one of Fig 2, the second is very similar.

Notice that this instance is reversed when one has to compare languages with a very remote common ancestor, in this case, in fact, only the extremely stable items are useful [71].

3.2 A date for the arrival of the ancestors of modern Malagasy people

The date of the landing of the Asian ancestors of nowadays Malagasy people or, more precisely, the date of the start of the diversification of Malagasy varieties, is fixed by choosing τ. For example, the choice of τ in Fig 2 determines the genealogical distance between the root and the leaves, this distance is by definition the time that separates all contemporary Malagasy dialects from their last common progenitor (Proto-Malagasy).

As already mentioned, in the idea of Swadesh τ should have been a universal constant but this is not the case for many reasons, as the incidence of horizontal transfers and the differences in replacement rates for different meaning or different language families. To overcome these difficulties Swadesh himself proposed for τ a value smaller than the inverse of the replacement rate but without a clear recipe for fixing its actual value.

In principle one can fix τ by a single known historical date. For example, Icelandic (language 1) and Norwegian (language 2) started to separate T(1, 2) = 1100 years ago; on the other side their lexical distance D(1, 2) or their overlap C(1, 2) = 1 − D(1, 2) can be easily computed from vocabulary by Swadesh approach; then, from the equation the value of τ remains determined because both T(1, 2) and D(1, 2) are known.

This method is not precise since the computed τ may be different for different pairs of languages in the same family and, in general, it is different for different families.

We propose here a strategy which is also based on history, but instead of considering a single pair of languages, we perform a kind of average over all the genealogical distances between each member of a family and their last common ancestor. The present approach is totally different and much more intuitive with respect to the one which was used in [13], nevertheless the result is the same: the landing of the Asian ancestors occurred around 650 CE.

To reach our goal we simply need a family whose last common ancestor is known and well temporally localized. For our purposes we found that a good choice was to consider the family composed by 46 Romance languages spoken in the geographically continuous territory represented the Italian peninsula, Iberian peninsula and Southern Gaul (see Fig 4 for the names of the languages). We chose this territory because all dates are known and because from the end of its Roman colonization to present, the language dynamics in this region was a mix of phylogenesys and horizontal transfers between geographically contiguous languages without main borrowings from non-Romance languages. In other words, a dynamics very similar to the Malagasy one.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. This UPGMA tree is obtained from the 1035 genealogical distances between all pairs which can formed by 46 Romance languages spoken in the Italian peninsula, Iberian peninsula and Southern Gaul.

According to our evaluation these Romance languages started to separate each other between 2121 and 2033 years BP. This range for the genealogical distance from the root to the leaves implies 5027 ≤ τ ≤ 5245. Notice that the main partition corresponds to the La Spezia–Rimini Line (also known as the Massa–Senigallia Line), which always seems quite mysterious for Italian native people.

https://doi.org/10.1371/journal.pone.0240170.g004

After having determined the overlap C(α, β) between all the 46 × 45/2 = 1035 pairs of languages in this family (from the dataset described in previous section), we determine the genealogical distances . Then, we built a UPGMA tree (Fig 4) and we get a value T = 0.4044 × τ for the genealogical distance from the root to the leaves. If one knows from history that these 46 Romance languages started to differentiate T years BP one immediately gets τ = T/0.4044.

In order to evaluate T, the relevant events are:

• 102—101 BCE—The Teutons and the Cimbri are defeated in the battles of Aquae Sextiae and Vercellae. The Roman dominion over Iberian Peninsula and Southern Gaul (Occitania) is undisputed and definitely consolidated. More the 90% of the territory is under Roman control.
• 19 BCE—The Astures and the Cantabri surrender to Rome, ending the Cantabric Wars. The Iberian peninsula is totally and definitively under the Roman rule.
• 24—14 BCE—The army sent by Caesar Augustus beats the last Ligurian tribes who still resisted Roman domination. The Roman conquest of Southern Gaul (Occitania) is completed.

According to these events, the 46 Romance languages that we considered started to separate between 2121 and 2033 years BP. From τ = T/0.4044 we get the range 2033/0.4044 = 5027 ≤ τ ≤ 2121/0.4044 = 5245.

In turn, this range of values for τ, when used for the UPGMA tree constructed from the genealogical distances of Malagasy variants (Fig 2), gives a value between 1338 (τ = 5027) and 1396 (τ = 5245) for the genealogical distance from the root to the leaves. This means a start for the diversification of Malagasy variants between 2019 − 1396 = 623 CE and 2019 − 1338 = 681 CE. In conclusion, it can be stated that the Asian ancestors of nowadays Malagasy people landed around 650 CE or shortly before.

3.3 Who was there before?

Malagasy mythology portrays a people, called the Vazimba, as the original inhabitants, preceding the arrival of the ancestors of modern Malagasy (but this term may also refer to the people ruled by the first kingdoms which established on the highlands, not a different people from nowadays Malagasy).

Archaeology tried to carry this belief on a scientific ground answering the question: when first humans reached Madagascar? We have seen that the answer is not univocal, with some authors claiming a very early presence dating thousands of years, and other authors which deny any presence before the second half of the first millennium CE.

Since linguistics and also genetics point for a date of arrival of the ancestors of modern Malagasy around the seventh century CE, the question for linguistics and genetics becomes: was the Island populated by a previously arrived people which possibly left tenuous traces in some Malagasy vocabularies or some unique genetic marks?

Some authors tried to find these linguistic traces. In [25] it is argued that “… low-density hunting-gathering populations probably did cross the Mozambique Channel and begin to exploit the Malagasy environment. Such populations would probably have been physically like the present-day Hadza of Tanzania rather than Khoesan speakers. It is likely that these survive in the present-day Mikea/Vazimba populations; although today they speak Malagasy dialects, there are clear cultural and linguistic traces of a distinct origin”.

Moreover, in [75] the authors write “Generally speaking, the Vazimba data is so exiguous and so scattered, that it is unlikely to constitute a significant record of a substrate vocabulary, as opposed to a set of idiosyncratic words, not untypical for a population of former foragers. However, […] it looks as if there may be genuine substrate vocabulary in Beosi and that this could reflect the speech of a forager group which migrated from the African mainland in pre-Austronesian times.”

Indeed, these papers fail in providing any evidence, moreover they superficially individuate the subject of their research “Scattered among the Malagasy live groups of hunter-gatherers variously known as the Mikea or Vazimba”. In reality Mikea are not a constellation of tribes scattered all around but a specific population living in the Mikea forest, North of Toliara. One of the authors of the present paper (M.S.) inhabited in the area for few months, moreover, many scholars dedicated long studies to this people (see for example [76, 77]).

Furthermore, the authors of [78] after having performed a genomic study over three population including Mikea declared “We were unable to detect in Mikea unique ancestry components that would have been absent among sampled populations. These results support the hypothesis that the Mikea originated from agricultural populations and have reverted to the forest. Despite the Mikea population deriving from the same genetic admixture as Vezo and Temoro, they have reverted to a hunter-gatherer lifestyle.”

In conclusion, evidence points to a phenomenon of cultural reversion to a hunter-gatherer life style of a small population [76], which is not the only case ever observed [79].

Analyzing the point with the help of the two trees that we have generated (Figs 2 and 3), the Mikea variety is well collocated in the Southern blue group. In the UPGMA tree it is scattered to form a separated branch, but in the NJ tree it finds its position close to the neighboring Vezo and Masikoro dialects. Since NJ is usually able to detect deeper relationships between languages than UPGMA, we argue that the relative isolation in UPGMA is due to the recent drift after Mikea embraced a Paleolithic lifestyle.

We also measured the average lexical distance of each language from the others, finding that this quantity is large for Mikea, but not larger than those we found for Antandroy dialects (S3 Fig).

Finally, in the analysis of next section where languages form two-dimensional patterns, again we find that Mikea variety clearly belongs to the Southern blue cluster, even if it is slightly scattered away from the center of it.

In conclusion, our opinion is that Mikea is a standard Malagasy variety. Differences, which are likely a consequence of recent isolation due to their peculiar lifestyle, seem to be the result of some modern inventions, probably a strategy to self-protect from external intrusions.