My identification of Arabic, European, and recent Swahili and Comorian loanwords, as well as the contexts in which they occur, is largely based on the work of previous scholars such as Dez (1964, 1965, 1967) and Dahl (1988), although I may differ in my analysis of the changes the words in question have undergone. These loanwords escaped the effects of the Bantu contact period and its thorough phonological influence on older vocabulary. They are easier to recognize because their shapes have not changed so much from their lending form and their phonotactics is very different from that of inherited Malagasy words. Moreover, they denote, as a rule, concepts that are adapted from the contexts of experience that their respective lending languages belong to. There are consequently few cases that have a doubtful borrowing status.

The treatment of Austronesian and Sanskrit loanwords I present here is based on my own analysis, although most Sanskrit loanwords had already been identified by previous scholars. The description of Bantu loanwords is largely based on Dahl (1988). These older loanwords are much more difficult to spot because they have undergone the changes that took place during the Bantu contact period. Their phonological shapes are thoroughly transformed and they are relatively well-adapted to the inherited Malagasy word structure. Their status as loanwords is therefore more difficult to demonstrate than in the case of Arabic, European, and recent Swahili and Comorian loanwords. There is also no fast rule to determine their status, as it is based on a combination of phonological, semantic, cultural and historical factors, which work out differently in individual cases. In a very general way, I tend to use rating 1 (very little evidence of borrowing) with loanwords from before the Bantu contact period in cases where I disagree with other scholars that a certain word is borrowed. I use rating 2 to 4 if a Malagasy word reflects a phonemic correspondence which is diagnostic for borrowing, and its meaning is similar to – or compatible with - that of the lending form. Such diagnostic correspondences are, for instance, Malagasy s, r, ts, or tr respectively corresponding to s, r, c and (alveolar or retroflex) d in the alleged Austronesian or Sanskrit lending language. These correspondences are irregular and indicate that the word in question could be borrowed. The more such phonemes occur within a given word, the more likely it is to be a loanword, and the higher rating I will be using. In the same way, I use rating 2 to 4 for words containing a b, p, g, or non-root-final k, if these phonemes correspond to b, p, g or k in Malay words with a compatible meaning. Such words may be loanwords borrowed from Malay after Bantu contact period. Finally, I use rating 2 in some cases where a word begins with ki- or tsi- and denotes a noun, as this suggests a Bantu origin. This is the case with kibo ‘stomach’, for which I found corresponding forms in neither other Austronesian languages, nor in Bantu languages. (If I had found a corresponding form with a compatible meaning in Swahili, Comorian or another East African Bantu language, I might have used rating 3 or 4).

Developmental stages preceding modern Malagasy:

Proto Austronesian 6000 BC

Proto Malayo-Polynesian 4000-3000 BC

Proto (West) Malayo-Polynesian 2000 BC ?

Proto East Barito 100-300 AD ?

Proto South East Barito 500-600 AD ?

pre-migration before 700 AD

Bantu or Post-Bantu period 800 AD or later ?

pre-Malagasy (before dialect divergence) 900 AD or later ?

Periods of borrowing:

Sumatra Malay, Banjar Malay, Javanese,

South Sulawesi, Sanskrit before 8th century AD

First Bantu influence 8th century (?)

Sumatra Malay between 8th and 16th centuries (very few loanwords)

Swahili, Comorian languages 12th to 20th century?; ongoing in regional areas?

Arabic 12th (?) to 19th century; ongoing in regional areas?

English mostly in 19th century but probably ongoing

French since late 19th century, ongoing