Deep Linguistic Prehistory
with particular reference to Andamanese
by Niclas Burenhult
Table of Contents
1.1. Johanna Nichols: Linguistic diversity in space and time
2.1. Head or dependent marking?
3.1. Correlations of types and categories
Reprinted from Working Papers, 45.1996, pp. 5-24, Lund University, Department of Linguistics, Lund, Sweden
Reproduced with permission.
In 1992, American linguist Johanna Nichols introduced a new method of detecting typological patterns at great time depths, based on the morpho-logical analysis and cross-linguistic comparisons of several structural types and grammatical categories (Nichols J. 1992. Linguistic Diversity in Space and Time. Chicago/London: The University of Chicago Press). She claimed that her method reveals patterns that may go back as far as the initial modem human colonization of the globe, and she set up a preliminary model of early linguistic spread. Has Nichols taken a ground-breaking step towards a greater understanding of our distant linguistic past? And how can we test this?
Towards the end of her book, Nichols J. 1992:263-65 calls for an analysis of 'critical' languages which are in a unique position to fill the gaps in her study and thus essential to our understanding of global linguistic prehistory. Using Nichols' method as a testing model, this article highlights one such critical language group the Andamanese language family, spoken by the indigenous Negrito population on the Andaman Islands, in the Bay of Bengal in an effort to shed further light on the distant linguistic past of our species.
1.1. Johanna Nichols: Linguistic diversity in space and time
Nichols' 1992 study involves a statistical survey of four salient morpho-logical patterns labelled structural types (head/dependent marking, morpho-logical complexity, clause alignment and word order) and a number of other phenomena called grammatical categories (inclusive/exclusive opposition, noun classes, numeral classifiers, alienable/inalienable possession, number and valence-affecting operations).
Nichols notes several geographical discrepancies, notably between east and west, or, more specifically, between the Old World and 'colonized' areas (the Pacific and the New World), and she offers a far-historical interpretation of her findings. On the basis of the east-west discrepancies, she posits an ancient typological split between the linguistic populations of the Old World and those of the Pacific, with the Pacific then functioning as a secondary center of spread and source of circum-Pacific linguistic colonization (Nichols J. 1992:228-229). She suggests that these patterns could only go back to the initial global spread of our species, starting in Africa some 100,000 years ago (Nichols J. 1992:258-259, 274-275).
1.2 .The Andamanese, their islands, and their language
Situated in the Bay of Bengal between Sumatra and Burma, the Andaman Islands is a group of some 200 islands with a total area of 6,340 square kilometers. The main group North, Middle and South Andaman are separated only by narrow straits and are collectively known as Great Andaman. Other islands include, for instance, North Sentinel Island and Little Andaman. The islands are mountainous and covered by dense tropical forest. Politically, the Andaman Islands and the neighboring Nicobar Islands constitute a union territory of the Republic of India, but geographically they form part of Insular Southeast Asia.
The indigenous population of the Andaman s belongs to the Negrito stock, a group of dark-skinned and short-statured peoples also found in the Malay Peninsula and the Philippines, who are often considered to represent the original population of Southeast Asia. At the time of European contact, the Andamanese were divided into 13 separate tribes of hunter-gatherers. Following European colonization of the Andaman s in the late 1700s, the indigenous population dwindled rapidly as a result of infectious diseases, and today only three tribes remain the Onge, the Jarawa and the Sentinelese together numbering only a few hundred individuals. The Onge and some Jarawa groups have friendly relations with outsiders, but most Jarawa and the hostile Sentinelese still aggressively oppose any foreign encroachment.
The Andamanese language family has been classified into two subgroups, Great and Little Andamanese. The Great Andamanese subgroup, spoken in the Great Andaman archipelago, consists of ten extinct languages: Bea, Bale, Puchikwar, Juwoi, Kol, Bo, Cari, Kede, Jeru and Kora (spelling and classification follow Manoharan S. 1983, 1986). The Great Andamanese languages became extinct in the 19th and 20th centuries, but a creolized form of Great Andamanese (labelled Present Great Andamanese by Manoharan S. 1986:27, 1989, and here referred to as PGA). predominantly based on Jeru, is still spoken by a handful of individuals resettled on Strait Island.
The Little Andamanese subgroup includes three languages: Onge, Jarawa and Sentinelese. Onge is spoken by less than 100 individuals on Little Andaman, and Jarawa speakers (approx. 300) still inhabit the western parts of Great Andaman. Sentinelese is spoken by some 150 individuals on isolated North Sentinel Island (Singh R. 1975:72). The classification of Sentinelese among the Little Andamanese languages is, however, pure guesswork and has not been established scientifically, as no linguistic studies have been carried out on the island (Manoharan S. S. 1983:83).
The relationship between the Great and Little Andamanese subgroups has been the subject of some discussion. Few cognates link the two subgroups together, and Greenberg J.H. 1971:810 doubts their genetic relationship. However, it was pointed out already by Radcliffe-Brown A.R. 19 14:40, 1933:497 and later by Manoharan S. 1983:86 that great morphological resemblances point to a common genetic origin.
Although the Andamanese language family is generally regarded as an isolate, attempts have been made to classify it into broader groupings. The most serious of these is Greenberg J.H.'s 1971 Indo-Pacific hypothesis, which links Andamanese to the non-Austronesian languages of Melanesia and to Tasmanian on the basis of 35 cognates and a few grammatical features, but this hypothesis has not gained widespread support.
The earliest account of Andamanese language is that of Colebrooke R. 1795. A great deal of research was carried out in the late 1800s by British government servants, predominantly by Man, Temple and Portman . Man E.H. 1883:49-56, for instance, gives a fair amount of data on Andamanese in general, and Portman M.V. 1898 makes a detailed description of five languages on southern Great Andaman. Useful linguistic notes were also made by anthropologists working in the Andamans in the early 1900s, notably Radcliffe-Brown A.R. 1933:495-504.
Since 1950, linguistic research in the Andaman Islands has been carried out by linguists from the Anthropological Survey of India (Zide N. and Pandya V. 1989:640). Field investigations on Onge have been conducted by Ganguly P. 1972 , Nigam R.C. 1969 and Dasgupta D. and Sharma S.R. 1982. A brief description of Jarawa has been collocated by Nair V.S. 1979. Attempts at subgrouping and analyzing the Andamanese language family as a whole have been made by Manoharan S. 1983, 1986, who has also conducted extensive field work among the few remaining speakers of Present Great Andamanese (Manoharan S. 1989).
2. Structural types and grammatical categories
In this section, attempts are made to identify the structural types and grammatical categories present in the Andamanese languages. However, before doing this, it is necessary to make some introductory clarifications.
Andamanese nouns and verbs can be classified as either 'dependent' or 'independent'. This terminology is unfortunate for our purposes, since 'dependent' is used in quite another sense in this discussion, and for the sake of clarity the terms 'inalienable' and 'alienable' will be used here. Inalienable nouns, which usually denote body parts, kins and parts of a whole, cannot appear without a possessive prefix or a cliticized noun. Similarly, inalienable verbs always incorporate a pronoun or noun indicating subject or object relationship. Alienable nouns and verbs, on the other hand, are free forms. Although systems vary, the basic idea appears to be the same in all Andamanese languages. As will be noted, the split pattern results in differences in the morphological marking of syntactic relations. (See Radcliffe-Brown A.R. 1933:497-98; Ganguly P. 1972:3-4; Nair V.S. 1979:22-23; and Dasgupta D. and Sharma S.R. 1982:10-12).
2.1. Head or dependent marking?
The following sections deal with the morphological marking of syntactic relations in NPs, PPs and clauses. In the examples given, the head of the constituent is boldfaced and markers are italicized.
Morphological marking in NPs. When combined with alienable nouns, nominal possessors in Onge take a genitive suffix, as illustrated by the following example (Dasgupta D. and Sharma S.R. 1982:15):
'the Onge's cloth'
In Jeru and some other languages of North Andaman the genitive marker of nominal possessors can also stand by itself (Radcliffe-Brown 1933:503, 504):
In Onge, pronominal possessors in combination with alienable forms take the same genitive suffixes as the nominal possessors (Dasgupta D. and Sharma S.R. 1982:18). The languages of southern Great Andaman (including Bea, Bale, Puchikwar, Juwoi and Kol), on the other hand, appear to have possessive pronouns without a distinguishable genitive suffix (Portman M.V. 1898:131). Pronominal possessor in Onge (Dasgupta D. and Sharma S.R. 1982:18):
In combination with inalienable forms, possessors take on a different character. Nominal possessors are cliticized to the head noun, and pronominal possessors turn into possessive prefixes, as shown by the following examples from Onge (Ganguly P. 1972:3-4; Dasgupta D. and Sharma S.R. 1982:10-13):
the dog's head'
In Onge, modifying adjectives are incorporated into the head noun, either directly after it or following an article (Dasgupta D. and Sharma S.R. 1982:38-39). This type of NP may therefore be regarded as head-marked.
'a big snake'
Information on modifying adjectives in the Great Andamanese languages, including PGA, is scanty, but examples from Bea (Portman M.V. 1898:118, 120) may indicate that adjectives stand by themselves and that neither head nor dependent exhibit any form of marking.
Morphological marking in PPs. In all Andamanese languages, direction, location, instrument etc. is expressed by a wide range of suffixes on nouns and pronouns. In the literature, these are generally called postpositions, but presumably they are more correctly referred to as case (or case-like) suffixes (as recognized by Basu D.N. 1952:64 and Manoharan S. 1989:77-80). Case suffixes in Onge (reanalyzed example from Dasgupta D. and Sharma S.R. 982:20-22, 53-54):
'from the forest'
and Bea (reanalyzed example from Radcliffe-Brown 1933:503):
So, instead of being an independent word and head of the constituent, the adposition is suffixed onto its noun. Therefore, the Andamanese languages cannot be said to have a phrase which can be counted as a PP. Nichols J.1992:59 refers to these constructions as dependent-marked Ns rather than PPs.
Morphological marking in clauses. It has proved almost impossible to find suitable examples of clauses in the material available, but the following conclusions can be drawn from the data on Onge (Dasgupta D. and Sharma S.R. 1982:40-69): in combination with alienable verbs, nominal subjects and direct objects appear to be unmarked, whereas indirect objects presumably take a case suffix. The following clause (involving an instrument instead of an indirect object) will have to illustrate this pattern (Dasgupta D. and Sharma S.R. 1982:53-54):
'the women dig tubers with iron hoes'
Pronominal subjects are unmarked, but both direct and indirect objects, if pronouns, receive an object suffix (Dasgupta D. and Sharma S.R. 1982:24. 64):
'you gave me a match-box'
'she painted me with clay'
If the verb is inalienable, it is prefixed by a nominal or pronominal subject in intransitive clauses, and by a nominal or pronominal object in transitive clauses. Intransitive clauses with inalienable verb (Dasgupta D. and Sharma S.R. 1982:25, 63):
'the dog barked'
Transitive clauses with inalienable verb (Dasgupta D. and Sharma S.R. 1982:24, 63):
'I pierced the fish'
'I scold him'
I have found no examples of indirect objects in connection with inalienable verbs. However, phrases designating location, instrument etc. are not cliticized to the verb and take the usual case suffixes, as in the following example (Dasgupta D. and Sharma S.R. 1982:63):
'the foreigner shot the pig there with the gun'
Judging from examples provided by Manoharan S. 1989:105-8, a similar system exists in PGA. In combination with alienable verbs, nominal and pronominal subjects and nominal direct objects occur independently and are unmarked (it is not clear whether pronominal direct objects take an object suffix, as in Onge). Inalienable verbs must have a pronoun prefix or a noun clitic.
Portman 's 1898 and Radcliffe-Brown's 1933:501, 504 data, although meagre, indicate a similar morphosyntactic structure in the extinct Great Andamanese languages, but no definite conclusions should be drawn from this material.
Conclusions. We have to admit that not all of the examples given above live up to the constituent requirements stipulated by Nichols J. 1992:46-47. We also have to face the fact that data is lacking from the majority of the Andamanese languages, and that the material on Onge is by far the most exhaustive. Indeed, there is a risk that our heavy dependence on Onge may result in a skewed picture of the morphosyntactic structure in Andamanese in general. As noted above, however, we can be fairly sure that the basic pattern is the same in all Andamanese languages, and Onge will be treated below as a typical representative.
Two important conclusions can be drawn. First, the Andamanese languages exhibit a very clear split in head and dependent marking in NPs and clauses. NPs with alienable nouns, and clauses with alienable verbs, are dependent-marked, whereas NPs with inalienable nouns, and clauses with inalienable verbs, are principally head-marked. The head marking pattern is thus restricted to a bound set of nouns and verbs, but, following Nichols' 1992:60-61 criteria for counting markers, such a marking pattern is still major and salient and should be counted. Second, the Andamanese languages lack true PPs, since they have a set of case-like suffixes instead of adpositions. Therefore they have dependent-marked Ns instead of PPs (Nichols J, 1992:59).
Counting markers and determining type. The tabulation of head and dependent marking involves a count of the number of 'points' (i.e. affix, clitic or particle slots) which are head-marking (H), dependent-marking (D) or detached (F) for each of the three constituents (NP, PP and S) described above. Constituents with noun dependents and constituents with pronoun dependents are counted separately and totaled for each constituent type (see Nichols J. 1992:56-62, 97-98).
The Onge noun phrase thus contributes two dependent-marking points (the genitive suffix on noun and pronoun dependents in combination with alienable heads) and three head-marking points (the pronoun prefix and noun clitic on inalienable heads, and the cliticized modifying adjective) to the Onge total. True adpositional phrases do not exist, so the PP constituent contributes no points. The clause presumably contributes three dependent-marking points (suffixes on noun direct objects and pronoun direct and indirect objects) and two head-marking points (the direct objects in the form of pronoun prefix and noun clitic on inalienable verbs). The NP+S total is thus five dependent-marking points and five head-marking points.
In determining the head-dependent 'type' of the language, the proportion of dependent-marking points in NP+S is computed as D/(D4+H+F) (Nichols J. 1992:59-60). According to Nichols' 1992:97-98 specifications, languages can be counted as head-marking if they have a proportion of 0.0 0.3, as double/split-marking if they have a proportion of 0.4 0.6, and as dependent-marking if they have a proportion of 0.7 1.0. In Onge, the proportion of D-points is 0.5 and hence the language falls neatly into place within the double/split-marking section of the scale, as would be expected.
As noted above, the information on the Great Andamanese languages is far too limited for a similar analysis. It could be argued, however, that some Great Andamanese languages would contribute fewer points overall (note the lack of marking in some NPs in Bea, for example), but this would probably not result in any dramatic shifts in the type scale. It will be assumed here that all Andamanese languages belong to the split-marking type.
2.2 .Morphological complexity
Morphological complexity, in Nichols' terms (Nichols J. 1992:64-65, 87-88, 98), is calculated simply by adding up the D, H and F points for NP and S. Languages totaling between 1 and 5 points show low complexity, those totaling between 6 and 10 points show moderate complexity, and those totaling between 11 and 15 points show high complexity. Onge totals 10 points and consequently shows moderate, although near-high, complexity. It is possible that some of the Great Andamanese languages show less complexity than Onge (due to the lack of marking in some NPs mentioned above), but they would probably still be moderately complex.
Nichols J. 1992:65-66 lists six alignment categories, based on the morphological distinction or nondistinction of subject of transitive (A), direct object (O) and subject of intransitive (S): neutral, accusative, ergative, three-way, stative-active and hierarchical. In determining the alignment type of a language it is necessary to tabulate an alignment category for each part of speech (Nichols J. 1992:88-91). Hence we need to identify morphological alignment marking in pronouns, nouns and verbs.
Judging from the examples above, nouns and verbs in Onge clauses with alienable verbs have no inflectional oppositions identifying A, O and S and thus exhibit a neutral alignment pattern. However, pronouns show an accusative pattern, because pronoun 0 appears to receive a distinct marking, while pronoun A and S are unmarked.
Clauses with inalienable verbs, on the other hand, present a different pattern. Here, S and O (both nouns and pronouns) are incorporated into the verb whereas A is independent and distinct. Clearly, such a pattern could be referred to as ergative-like, but it does not involve inflectional marking and should therefore not be considered here (Nichols J. 1992:65). Instead, A, O and S should be regarded as unmarked and neutral.
In determining the dominant alignment type of a language. Nichols J. 1992:92 excludes the neutral pattern unless the language "has absolutely no relevant morphology". The sole non-neutral type in Onge accusative is therefore the dominant type. It is not clear whether PGA and the extinct Great Andamanese languages have object inflection in pronouns. Their dominant alignment type is accusative if they do, and neutral if they don't.
2.4. Word order
The basic word order is verb-final in all Andamanese languages. Onge (Dasgupta D. and Sharma S.R. 1982:40) and Jarawa (Nair V.S. 1979:23-24) have SOV, and so does PGA (Manoharan S. 1989:105). Judging from Radcliffe-Brown A.R. 1933:504 and examples from e.g. Portman M.V. 1887:7, 1898:126-31, the same was true for the now extinct Great Andamanese languages.
2.5. Inclusive/exclusive distinction
Manoharan S. 1986:28-29, 1989:67-68 distinguishes an inclusive/exclusive opposition in the first person plural pronoun of PGA. However, no inclusive/exclusive distinction has been observed in any other Andamanese language, including the extinct Great Andamanese languages, on which PGA is based, and one might suggest that the inclusive/exclusive distinction in PGA is a recent phenomenon and not a feature typical of Andamanese in general.
2.6. Noun classes
Opinions differ as to the existence of noun classes in Andamanese. In the early years of research, much attention was directed to a set of formative prefixes which were added to the nouns denoting body parts and which could be extended to certain other nouns in the form of "ordinary prefixes" on verbs and adjectives (Portman M.V. 898:34-45, 60, 79-83). Portman listed at least seven such prefixes related to various parts of the human body and concluded that they indicated gender, or, rather, different genera. Radcliffe-Brown A.R. 1933:498-501 states that "they give expression to a number of rather indefinite categories." However, Manoharan S. 1983:30; 1989:61-64, who lists 11 formative prefixes in PGA, is doubtful about these interpretations and states that the formative prefixes change "the meaning of the primary concept into the specific meaning" and therefore do "not organize different words into one group."
It is extremely difficult to get a clear idea of this system of formative prefixes, and the lack of raw data prevents us from taking the analysis any further, but let us outline the system and see if we can draw conclusions about its status as noun categorizer.
Human body parts are subdivided into a number of categories, each category having a distinct prefix. Thus, body parts like head, brain and heart form one category; hand, wrist, knuckle, nail, foot and ankle form a second, etc. (The number of prefix categories ranges from five to eleven or more, but the system is basically the same within the whole language family). The prefixes are obligatory and intimately associated with inalienability and hence the use of pronominal prefixes. The body part prefixes can then be applied to adjectives and possibly also verbs that refer to any noun (not necessarily inalienables), and the choice of prefix category depends on the properties of the noun in question, such as shape or position. The noun itself (if alienable) appears not to receive a prefix.
Evidently, the human body serves as a foundation for a wider classification of nouns, but the exact criteria for placing a noun within a particular category are not clear. It is not even clear whether agreement is obligatory or not. Furthermore, Onge word lists reveal that inalienable verbs display constant formative prefixes. This would mean that verbs do not show agreement with the noun they refer to, but carry their formative prefix for a different reason. Clearly, Man y questions remain unanswered and more material is necessary if we are to get a more complete picture of the formative prefixes. However, it will be concluded here that the Andamanese languages do display a noun class system, albeit unclear and extremely fluid, but no definite conclusions will be drawn about agreement.
2.7. Numeral classifiers
The formative prefixes described above have been likened to numeral classifiers (Manoharan S. 1986:30). It is true that the formative prefixes, like classifiers, involve shape categories and that classification is fluid, but they are fewer in number (classifiers typically range between 20 and 200) and, more importantly, they do not appear to be associated with numerals (see Nichols J. 1992:132). Moreover, unlike the formative prefixes, classifiers are always free forms and never form a morphological unit with the noun (Dixon R.M.W. 1986:106). Taken together, these criteria indicate that we are dealing with noun classes and not with numeral classifiers.
2.8. Alienable and inalienable possession
The Little Andamanese languages display a very clear alienable/inalienable distinction. In Onge, for instance, kinship terms, names of body parts and words denoting parts of a whole cannot appear without a possessive prefix (Dasgupta D. and Sharma S.R. 1982:10-13; Ganguly P. 1972:3-4). Judging from the limited data, the same pattern appears to exist in Jarawa (Nair V.S. 1979:22-23), and a similar system is evident in PGA (Manoharan S. 1989:64-65, 78-79).
The information available on the extinct Great Andamanese languages is rather confusing but appears to reveal a similar picture. In his survey of the languages of southern Great Andaman, Portman M.V. 1898:37, 60-69 lists both "simple pronouns" and "abbreviated forms of simple pronouns" and states that the formative prefixes used with names of body parts and words referring to "the human race" in general are "capable of combination with the abbreviated forms of the Pronouns". Although Portman probably did not recognize their significance as markers of inalienable possession, the "abbreviated forms of pronouns" are clearly identical to the possessive prefixes found in Onge, Jarawa and PGA. As to the languages of northern Great Andaman, Radcliffe-Brown A.R. 1933:501 states that personal pronouns may be either words or prefixes, but he does not relate the distribution of the pronominal prefixes to words denoting body parts or other inalienables. Still, it seems safe to conclude that all Andamanese languages display an alienable/inalienable distinction, in which nouns denoting body parts, and probably also some kinship terms, cannot be expressed without a possessive prefix.
In determining whether a language exhibits plurality neutralization or not, Nichols J. 1992:146 limits her survey only to those languages in her sample that have nonzero marking of both dependent and head at the clause level. This is to avoid the problem of distinguishing between underdevelopment of number and more general lack of inflection. Onge clauses with inalienable heads do display both head and dependent marking, and presumably Andamanese is then qualified for inclusion in the survey of plurality neutralization.
Onge has three numbers singular, dual and plural marked by suffixes on the noun (Ganguly P. 1972:4; Dasgupta D. and Sharma S.R. 1982:13-14). These number suffixes are optional. No data is available on Jarawa. As for the Great Andamanese languages, Portman M.V. 1887:4 claims that number is absent from nouns, and Radcliffe-Brown A.R. 1933:503 makes no mention of number in his account of nominal suffixes in Bea. Moreover, Manoharan S. 1989:61 states that PGA nouns do not have number inflection, which would seem to confirm the notion that the Great Andamanese languages did not exhibit number distinction in nouns.
Number distinction (singular and plural) exists in the pronominal systems of all Andamanese languages studied. No plural pronouns have been observed in Jarawa, but this is probably due to the difficulties involved in collecting material (Nair V.S. 1979:23). In Onge (Ganguly P. 1972:5; Dasgupta D. and Sharma S.R. 1982:17), the singular and plural pronouns appear to be expressed by unrelated roots. In the Great Andamanese languages, however, personal plural pronouns sometimes appear to be derived from their singular counterparts (see examples in Portman M.V. 1898:60-61; Radcliffe-Brown A.R. 1933:501; Manoharan S. 1989:68).
Some Onge verbs sometimes take a plural suffix, identical to the nominal plural suffix -le, when the subject is in the plural. This is illustrated by the following example (Ganguly P. 1972:6):
ekw-akobele-te-lle-be-gi 'they came running'
In the Great Andamanese languages, however, verbal number markers appear not to exist, and the same is true for PGA (Manoharan S. 1989:83).
Conclusions. Although very common, noun number suffixes in Onge are optional. Judging from Nichols' 1992:145, 296-97 treatment of Djingili, such a pattern should be considered an example of plurality neutralization. Similarly, only some Onge verbs take a plural suffix. The distribution of this suffix is unclear, but it should be evident that verbs too display a certain amount of plurality neutralization. Onge pronouns, on the other hand, do not suffer any plurality neutralization, since singular and plural forms appear to be expressed by unrelated roots. In the Great Andamanese languages, plurality neutralization is very apparent in nouns and verbs, which lack number markers altogether, and to some extent in pronouns as well, but this is less clear (for a definition of plurality neutralization in pronouns, see Nichols J. 1992:151-52). Hence all Andamanese languages exhibit plurality neutralization of some kind, but it is also evident that there are considerable differences between the Great and Little Andamanese subgroups, the former being radically more neutralizing than the latter.
3.1. Correlations of types and categories
The data presented here does not seem to contradict the correlations made between structural types in Nichols' 1992:97-115 survey. For instance, her claim that double/split marking favors moderate or high complexity and verb-final word order, and that accusative alignment, moderate complexity and SOV order are associated universally, is supported by the data from Onge.
The marking of NPs in the Andamanese languages is consistent with Nichols' claim that inalienable possession is most often head-marked while alienable possession is dependent-marked (Nichols 1992:117). The Andamanese pattern is also consistent with the claim that noun classes are associated with double/split marking and, more generally, with accusative alignment (Nichols 1992:138). However, in this respect Andamanese behaves more like a noun class hotbed language than an outlier, despite the fact that it does not seem to belong to a noun class hotbed (see below).
3.2. Incorporating Andamanese into Nichols' model of language spread
Before placing Andamanese in a wider perspective, it is necessary to make some basic assumptions about its origin and prehistory. As was pointed out at the beginning of this article, the Andamanese languages are spoken by Negritos, dark-skinned and short-statured hunter-gatherers who are considered by most experts to be descendants of the aboriginal population, and perhaps the first modern human settlers, of Southeast Asia (Bellwood P.S. 1985). The languages show no apparent genetic affinity to other languages of Southeast Asia or, indeed, the rest of the world, and there is no evidence of outside influence in the form of borrowing or precolonial linguistic colonization. In sum, it seems reasonably safe to assume that Andamanese is the sole remaining linguistic representative of pre-Neolithic Southeast Asia, its roots perhaps going back as far as the initial colonization of Southeast Asia by modem human s an isolate that has remained largely unaffected by the vast linguistic spreads that have occurred elsewhere in Southeast Asia in Neolithic and post-Neolithic times, i.e., the expansion of the Austroasiatic, Austronesian, Sino-Tibetan and Tai stocks. Andamanese might therefore give us a glimpse of what pre-Neolithic Southeast Asia may have looked like.
So, following Nichols 1992, what could we expect a pre-Neolithic language at the southeasternmost extreme of the Old World to be like?
Should we expect it to behave like other languages of the Old World? Or, considering the fact that Southeast Asia was a stepping-stone for the settling of the Pacific, will it have more in common with the colonized areas? Let us see how the Andamanese languages behave in relation to other parts of the world (Nichols J. 1992:184-208).
The scale of distribution of head/dependent marking is continental in size, and the most important patterns to be noticed are the high frequency of dependent marking in Africa and Eurasia, the high frequency of double/split marking in Australia, and the high frequency of head marking in the New World. The Andamanese languages, being split marking, thus exhibit the kind of marking that is predominant in Australia.
The distribution of alignment involves a worldwide preference for the accusative, and questions of the distributional scale of accusative are therefore irrelevant. Ergative alignment is a cluster phenomenon, mainly in parts of the Old World and Australia, and stative-active is a macroareal phenomenon that is common in the New World and New Guinea. Since the accusative pattern is the universal favorite, it is not surprising to find it in the Andamanese languages.
The mean complexity level is highest in the Old World, and moderate and high complexity levels are about equally common in this macroarea. In the colonized areas, however, moderate complexity levels are by far the most common. Onge, and presumably Andamanese in general, thus falls within the complexity range that predominates in the colonized areas.
The scale of geographical distribution of word order is smaller than continental, and the universally favored order is SOV. It is the most common order in all three macroareas, and its presence in Andamanese is no surprise.
3.4. Grammatical categories
There is a clear global dine in the distribution of inclusive/exclusive opposition and a considerable discrepancy between the Old World, which has low frequencies (21%), and the colonized areas, which have high frequencies (62% in the Pacific and 54% in the New World). Andamanese, which presumably lacks inclusive/exclusive distinction, conforms to the pattern that predominates in the Old World.
Noun classes and numeral classifiers are so-called hotbed phenomena which are smaller than continental in scale. Noun classes are frequent in the Old World and the Pacific (typical hotbeds include Africa, Europe and northern Australia). Numeral classifiers are rare overall but cluster along the Pacific rim. There are no examples of Southeast Asian languages with noun classes in Nichols' sample. Andamanese therefore does not appear to be part of a hotbed but forms a clear outlier. However, Nichols' 1992:131-32 claim that outliers typically give evidence of distant or former connection with hotbeds may be relevant to Andamanese (see below).
Plurality neutralization frequencies form a clear global dine Frequencies are low in the Old World (18%) and high in colonized areas (79% in the Pacific and 72% in the New World). Andamanese conforms to the pattern that predominates in the colonized areas.
Inalienable possession also forms a global dine. Frequencies are low in the Old World (23%) and high in colonized areas (65% in the Pacific and 51% in the New World). Andamanese conforms to the pattern that predominates in the colonized areas.
3.5. The adpositional phrase
Yet another global dine. The Old World is highly consistent in having adpositional phrases (76%), whereas the colonized areas show less tendency to have them (33% in the Pacific, Austronesian languages excluded, and about 50% in the New World). Again, Andamanese conforms to the pattern that predominates in the colonized areas, particularly to that of the Pacific.
Two of the features above, alignment and word order, display universal preferences. Accusative alignment and SOV word order are the most common in all parts of the world. Consequently, their presence in Andamanese is not unexpected and they are therefore more or less irrelevant to this areal comparison.
In five of the features, which all exhibit considerable discrepancies between the Old World and the colonized areas, the Andamanese languages uniformly correspond with the pattern that predominates in the colonized areas. Only one feature, inclusive/exclusive opposition, connects Andamanese to an Old World pattern.
Noun-classifying Andamanese cannot be connected to a noun class hotbed, geographically or historically, and thus appears to form a clear noun class outlier. However, Andamanese behaves like a typical hotbed language in having double/split marking and accusative alignment (Nichols J. 1992:138), and one might speculate that it was once part of a now-vanished Southeast Asian noun class hotbed.
The patterns summarized above tell us two things. First, the assumptions made at the beginning of this section appear to find support in the data presented. On the whole, Andamanese behaves very differently from its host area and macroarea Hence there is no evidence of precolonial linguistic influence or colonization from surrounding Old World areas. The isolated status of Andamanese is thereby further attested.
Second, Andamanese shows clear connections to the colonized areas. In three of the four features that display a global cline, as well as in continental features like head/dependent marking, Andamanese behaves like a typical 'eastern' or 'colony' language. The only exception to this rule is the inclusive/exclusive distinction, and the reasons for this discrepancy can only be hinted at. Andamanese languages may never have had an inclusive/ exclusive opposition, or they may have lost it at some stage in history (due to internal forces, if we exclude the possibility of Old World influence). Alternatively, the Andamanese languages do have the opposition, but it has not been noticed by researchers. This would explain its presence in PGA (indeed, Nichols J. 1992:209 claims that inclusive/exclusive opposition is the genetically most stable of the features surveyed). Only future research can shed light on this. So, if we view the interpretation of inclusive/exclusive distinction made above as uncertain and disregard it, all relevant grammatical features point to a striking uniformity between Andamanese and the colonized areas.
What bearing does this uniformity have on Nichols' view on early linguistic spread? If we posit that the Andamanese languages are typical representatives of the languages spoken in Southeast Asia at the time of the human crossing to Australia, New Guinea and adjacent insular areas, between 50,000 and 60,000 years ago, it could be argued that the ancient typological split Nichols 1992:228, 275 suggests to have occurred when Old World people set out to colonize new worlds actually took place long before that. In other words, the first major typological split took place somewhere west of Southeast Asia and not at the departure point of the first colonizers as indicated by Nichols. The colony patterns visible today would thus originate in the Old World, not in the Pacific. This view presupposes that Southeast Asia was typologically relatively homogeneous at the time of Pacific colonization, perhaps due to a random founder effect or an early intra-Old World stabilization of frequencies (for a discussion on stabilization of frequencies, see Nichols 1992:213.15) Subsequent Old World interaction and language succession would explain the near total obliteration of these extremely early Southeast Asian patterns.
This interpretation may have far-reaching implications on Nichols' macroareal division, because it suggests that parts of the Old World once behaved like the colonized areas and that the human leap to Australia and the Pacific was not a typologically very dramatic one. Her division of the world into an Old World homeland vs. colonized areas would thereby appear artificial. The scenario presented here is perhaps more in line with prehistoric reality, since the 'Eve' theory of modern human origin, to which most prehistorians adhere, treats sub-Saharan Africa as the true homeland of our species and all other parts of the world as colonized areas.
However, there is a possible alternative explanation to the similarities between Andamanese and the colonized areas. If Nichols is correct in suggesting a Pacific origin of the colony patterns, one might speculate that Southeast Asia has been subject to secondary, pre-Neolithic influence from the Pacific area, Andamnanese then being a witness to extensive east-to-west impulses, Clearly, such an interpretation would be more in line with Nichols' view of the Pacific as a secondary center of spread and source of circum-Pacilic colonization.
Incorporating a single language or language group into a model based on statistical data from a huge number of languages is a venturesome undertaking The strength of Nichols' survey lies in its sheer size, and any attempt to verify or falsify her model (or parts of it) on the basis of data from a single language can easily be brushed aside as statistically uninteresting. Similarities and dissimilarities to patterns in the model may simply be due to chance. Still, the isolated Andamanese languages, spoken by people who are believed to be descendants of the aboriginal population of Southeast Asia, can by all means be regarded as 'critical' and may, if added to Nichols' model, have a significant bearing upon her interpretation of linguistic prehistory.
As noted above, Andamanese exhibits striking correspondences to typical 'colony' patterns, that is, to typological patterns that predominate in the Pacific and the New World, and two different scenarios explaining these similarities were presented. Clearly, the first scenario is the one that is most compatible with the archaeological record. It is a well-established fact that Australia New Guinea was populated from Southeast Asia (White J.P. and O'Connell J.F. l982:42-46), but there is no prehistoric evidence of secondary impulses going in the opposite direction However, it would not be unreasonable to assume that such impulses have taken place at some stage in prehistory. It will therefore not be concluded here which scenario is themost likely one, only that the first scenario is more compatible with the archaeological record, and that the second is more in line with Nichols' linguistic model.
One thing is certain, whichever view we take on the directionality of pre-Neolithic linguistic spread or influence in Australasia: the Andamanese languages, little known and largely neglected, do have an interesting story to tell. And, as Nichols J. 1992:230 points out, the most striking aspect is the fact that modern linguistic evidence can have anything at all to say about early human prehistory.
Last changed 30 March 2006