' Semicreolization ' ? – The restructured Portuguese of the Tongas of São Tomé , a consequence of L 1 acquisition in a special contact situation

This paper discusses aspects of language transmission among the Tongas of São Tomé, the descendants of indentured Africans, on the Monte Café plantation where restructured Portuguese developed alongside an Umbundu-based koiné. It considers the sociohistorical context of language acquisition and transmission, and the role of Portuguese L2 influence in primary linguistic data (PLD) for L1 acquisition in this speech community. In the Tonga Portuguese of a first generation born to African parents, these processes gave rise to broad restructuring relative to (i) agreement rules (number, gender, and subject-verb), (ii) verb and tense and aspect marking, (iii) the signalling of definite and indefinite reference, and (iv) negation. These restructurings constitute a variable set resembling, in form and function, structures found in varieties of Creole Portuguese. However, the consequences in Tonga Portuguese are only slight in comparison, so the language appears to have been partially creolized. The motivation for restructuring is viewed both from DeGraff’s perspective of a Universal Grammar approach to acquisition and creolization, stressing the quality of PLD and its role in yielding unmarked structures, and from the perspective of Bantu substrate influence. Finally, discussion turns to Holm’s notion of ‘semi-creole’ and the validity of classifying Tonga Portuguese in these terms.


Introduction
This is an initial notice on the Portuguese of the Tongas of São Tomé (Africa), descendents of African plantation workers indentured during the Portuguese colonial system. 1 Principal concerns are the historical sociolinguistic context of 'Portuguese' language transmission among the Tonga community of the Monte Café plantation and reflexes of this process in morphosyntactic restructuring.Part of a larger project concerned with the development of Tonga Portuguese across three generations of speakers (cf.Baxter 2001), the present discussion concentrates on the speech of 10 of the oldest speakers interviewed in 1998, aged between 73 and 91 years, and born on Monte Café. 2 1.1.Background to indentured labour on the São Tomé plantations While slaves indigenous to São Tomé, speakers of São Tomé Creole Portuguese, were liberated in 1854 (Neves, 1929:55), along with children of foreign slaves, the liberation of foreign slaves was to come only 22 years later, in 1876.
After abolition, many of these foreigners returned to work for their former owners.Neves (1929:59) states that there were some 6000 slaves freed in 1876, and notes that half of this number was still working on the plantations by 1881, and only a third thereof in 1895.However, the bulk of plantation workers by this stage were new indentured workers.In the early 1850s, the Portuguese began to import indentured (and sometimes slave) labour from Angola (Nascimento & Gomes Dias, 1989:52).Later, they drew on English and French colonies as well as the Portuguese colonies on the African continent, Angola, Guiné, and Ajudá (in Dahomey, modern Benin).Eventually, Mozambique and the Cape Verde islands, also became labour sources.Table 1 summarizes worker numbers and origins for the period 1876-1879 (Neves, 1929:63) For the subsequent period, until 1901, detailed statistics have not been located.Neves (1929:63-64) gives only partial information, noting the predomi-1 I wish to express my gratitude to the two anonymous reviewers whose comments and criticisms greatly improved this paper.Responsibility for the final version is, of course, entirely mine. 2 The larger project, supported by a grant from the Australian Research Council, involves the quantified study of 1.5 hours of recorded interviews from each of 26 speakers, stratified by sex and in three age-groups.The project aims at studying the development of specific areas of morphosyntax across three generations by means of quantificational analysis.However, the present paper, as a <first-notice>, merely aims to give a general overview of key aspects of the sociolinguistics and structure of Tonga Portuguese.Detailed morphosyntactic analysis of the variable phenomena outlined herein is a goal of future research.
nance of Angolans (7,419 in 1881, and from 1885-1892, 10,411), and the presence of workers from Ajudá (716 in the years 1886-1887).He does, however, present detailed statistics for 1901-1928, and asserts that the Mozambicans began to arrive only in 1901 (Neves 1929:67).This information is summarized in Table 2: 4 The indenturing system continued until the Portuguese revolution in 1974, Cape Verde and Mozambique providing the bulk of workers after the 1940s.For the purposes of this paper, there are two notable aspects of this system.Firstly, prior to 1909, no Angolans were repatriated, although workers of other origins were (Cadbury, 1969).Secondly, while the overwhelming majority of indentured workers were eventually repatriated in the 20thC, their São Tomé--born offspring, referred to as Tongas, were obliged to remain on the island. 5s such, they became the defacto property of the plantations.By the 1940s, they were sufficiently numerous that, partly under government pressure and partly through the need for space for incoming workers, special small housing complexes, aldeamentos, were created for them on the plantations. 6

Language acquisition and transmission among the Tongas
Within the confines of the plantations, labourers worked in restricted situations where Portuguese, and not São Tomé Creole, was the language of 4 At least until the second decade of the 20thC, and especially in the period leading up to the beginning of the 20th century, the contractural conditions, and the living and work conditions for these people amounted to a continuation of slavery.In fact, prior to 1909, many of the Angolans on São Tomé were people who had been sold into slavery in Angola, as documented by Swan (1909).These facts were confirmed in the testimonies of our oldest informants.When word of this 'modern slavery' spread to Europe at the turn of the 20th century, an international scandal ensued with European cocao buyers boycotting São Tomé cocoa.This, in turn, led to a substantial improvement in general contracting conditions and the institution of a repatriation system which began only after 1908!Thus, between 1908-1928, 40,880 workers were repatriated, 57% being Angolans, 37% Mozambicans, 5.8% Capeverdeans.Nevertheless, conditions on the plantations must have still been very harsh.Cadbury (19695 Cadbury ( [1910]]), who conducted an investigation of the conditions of the indentured workers on São Tomé, noted that children of the workers were not repatriated.This fact was also echoed in the testimonies of our older informants. 6While technically, all off-spring of indentured workers are referred to as Tongas, the use of this term on the plantations today often refers only to the descendents of continental Africans (i.e.not Capeverdeans), a fact which reflects the segregation of the workers of different origins during the Portuguese colonial period.
communication between overseers and workers. 7On particular plantations, the proportions of workers of different origins varied, to the extent that different plantations were known by the particular African languages used (Rougé,  1992).Workers were housed separately in sanzalas (lit.slave quarters), and assigned to work-groups, according to country of origin.Moreover, at least until the mid-1940s, they were confined to the precincts of the plantations. 8In 7 The overseers were European.This is reflected in the observations of Cadbury (1969  [1910]) and, for the plantation considered in this paper, in Monte Café (1895) and in the testimonies of elderly informants.During the 20thC, however, some overseers were Tongas.Furthermore, following the liberation of the local-born slaves in the 19thC, it appears to have been a colonial policy that the local creole population was not to be employed on the plantations.This point is reflected in Cadbury (1969  [1910]) and in the testimonies of our oldest informants, including Tongas who had worked as overseers.In the 19th century, following emancipation, the local creoles had refused to work on the plantations (Neves, 1929), and in the 20th century, until the late 1950s, they were forbidden entry to all plantations.The fact that the Tongas did not have contact with São Tomé creole in the 20th century, and may have had only minimal contact with it in the 19th century, since plantations founded in the second part of the 19th century used foreign slaves, is a very important point when considering the origins of Tonga Portuguese.It essentially eliminates São Tomé creole as a contributor to the development of Tonga Portuguese, unless one envisages a long diachronic process extending back to the beginning of the 19th century, at a time when São Tomé creole would have been spoken by plantation slaves.However, the plantation considered in this paper, Monte Café, was founded precisely in 1854 (Mello, 1864:5), the very year of emancipation of local-born slaves, and appears to have used imported contract workers from the beginning. 8This statement is based on the testimonies of our oldest informants, including Tongas who were overseers.
One of the reviewers of this paper has raised two important questions regarding this 'isolation'.Firstly, given what is known of traditional slave plantation societies, would there not have been frequent contacts between plantations and between plantations and towns?And, secondly, if such contacts existed, would they not plea for a larger potential formative role of the local creole in the genesis of Tonga Portuguese?Depending on the location of the particular plantation in relation to towns and areas where the local creoles resided, and the size of the plantation, the chances of contact with locals would have varied somewhat.Certainly, Tongas who were connected more with the administration of the plantations would have had greater access to off-plantation contexts.Nevertheless, according to the oldest Tongas interviewed, including Tongas who were former workforce overseers, on three plantations (Monte Café, Rio D'Ouro and Praia das Conchas), in the pre-war period, the average plantation worker required a pass to exit the plantation, and passes were only issued in special circumstances.The worker's day was strictly regimented.It began with a role-call, and ended with a curfew.Work hours were strictly overseen.Plantations had their own shops and bars for the workers and, until the 1940s, the workforce had only one half-day off, on Sunday afternoon.Furthermore, many plantations had their own jetties, so that a good deal of produce was shipped semi-directly, and not transported overland to the capital.Thus the need for a workforce presence off--plantation was minimal.
While the possibility of contact with the creole population may have been significant on smaller plantations established close to creole settlements, the case of other the sample of informants interviewed, only speakers in the youngest age-group claimed to have any knowledge of São Tomé Creole, and this was the result of post-1950 contact outside the plantation.
The continental Africans seldom arrived with a knowledge of Portuguese.On the plantations, while they continued to speak their own languages, some of which were to function as lingua francas, they acquired L2 Portuguese through contact.Here it is important to note the testimony of the oldest informants interviewed.They clearly consider Tonga Portuguese to be simply língua de branco 'whiteman's language', primarily the language used with the Europeans, or with fellow workers who didn't speak Umbundo.
The regular flow of Africans meant that adult-acquired L2 Portuguese was constantly present as the dominant variety of Portuguese spoken by the plantation workers.Their descendents, in addition to maintaining the (often koineized) ancestral languages to varying degrees (Rougé, 1992), acquired Portuguese on the basis of Primary Linguistic Data (henceforth PLD) which must have been heavily influenced by adult-acquired L2 Portuguese (and by Tonga L1 Portuguese, itself heavily influenced by workforce L2 Portuguese).Thus, the variety of Portuguese spoken by the Tongas diverged considerably from that of their Portuguese overseers.This language was used in work contexts, and for inter-group communication, and it gradually became used for intra-group communication alongside the ancestral languages.It was not until the 1950s that Tonga children began to be exposed to some formal instruction in Portuguese, curtailing further independent development of Tonga Portuguese and leading to the use of models progressively closer to European Portuguese.

The acquisition of Portuguese on Monte Café plantation
On the Monte Café plantation the Tongas are predominantly of Angolan ancestry, and retain an Umbundo-based koiné. 9Angolan workers played a significant role in earlier stages of this plantation.Already at the time of the foundation of Monte Café, in 1854, one of its owners secured authorization to import workers from Angola, a practice he was to continue into the 1870s plantations, such as Monte Café, appears to have been quite different.Monte Café is on the shoulder of a mountain in an area which, in the 19 th C, would have been considerably remote, and even today lies beyond the areas of creole settlement.Thus, unless foreign-born slaves resident prior to 1854 (and who had worked in contexts where Sãotomense was spoken) had been employed on Monte Café in the foundation phase, together with new foreign contract workers, I believe that the potential formative role of the local creole is discounted. 9In the 1940s an aldeamento was created for the (largely) Angolan ancestry Tongas.
This helped create a focus for this community within the plantation.To some extent it probably helped preserve the Umbundo-based koiné, as well as their distinctive Portuguese.(Nascimento & Gomes Dias, 1989:52, 68.).Thus, it may be assumed that the 300 workers reported on Monte Café in 1864 in documents of the Provincial Government (Mello, 1864:5)   (Curadoria, 1895-1896).Table 3 presents partial statistics for Monte Café in the 1930s and 1940s. 10In 1935, the Mozambicans predominated slightly (= 52% of foreign workers) over the Angolans (= 46% of the foreign workers).However there are still large numbers of Angolans in the 1940s.What were the circumstances under which children born on Monte Café learned their first language(s)?In general terms, from the above statistics, for example in 1935, when 78% of the adults were continental Africans, it is evident that the majority of the workforce spoke African languages and, since Portuguese was the language of communication between overseers and the workforce, they must also have spoken varieties of L2 Portuguese.For the oldest Tongas interviewed, born in the first two decades of the 1900s, the home linguistic situations were of several types, according to the ethnolinguistic backgrounds of the parents.In several instances both parents were African.In most cases they were Angolan, yet in others they were mixed (the mother being 10 From these documents, two important details emerge.Firstly, the Capeverdeans only become a significant force on Monte Café in the late 1940s, and it may be assumed they were even less significant prior to the 1930s, according to our informants and the general statistics presented by Neves (1929).Secondly, from the Monte Café contract register for the 1940s and from interviews with old Tongas, there were very few Mozambican women.Thus, the Mozambicans registered in 1946 are all men, whereas the Angolans registered that year are men (265) and women (63).This trend may account for the dominant Angolan cultural ancestry of the Monte Café Tongas.
During his evaluation of the labour situation on São Tomé, Cadbury (1969 [1910]) had noted the presence of both Angolan men and women, in contrast to the dominant male groups of other origins.
Angolan and the father being of another source, for example from Ajudá or from Mozambique).
Where both parents were Angolan, the first language acquired was the parent's L1, Portuguese being learned largely in contact with other adults and children, especially in the field (as mothers took their young children to work with them), in the quintal ' the yard of the plantation administrative area' (where children of roughly 8 years age were assigned yard duties with an overseer, generally an African or a Tonga11 ) and in the context of the sanzala.In all three contexts children would have heard Tonga L2 and L1 Portuguese, and they could also have heard L1 European Portuguese from overseers in the field and in the quintal.In this latter context children were actually obliged to speak Portuguese.In the field and in the sanzala context they would have also heard and used Umbundo.
Children of African parents of different origins acquired the Angolan mother's language (which in some cases was also used by the father), and they acquired 'Portuguese' based on the L2 version often spoken between parents, and also through their exposure to Portuguese in the field and quintal contexts.Of the first generation Tongas interviewed, none acquired African languages other than Umbundo.
Finally, children born to Tonga couples, or to couples where one parent was Tonga and the other African, appear to have acquired both the Umbundo koiné and Portuguese simultaneously.In this case one principal source for PLD for the acquisition of Portuguese was Tonga L1.However, as in the case of children of African couples, the field, quintal and sanzala contexts would also be relevant to language acquisition both of Portuguese and of Umbundo.
The situations just described suggest that whereas some children's acquisition of Portuguese was essentially simultaneous with their acquisition of the Umbundo koiné, for others it must have amounted largely to pre-pubescent L2A.For the purposes of the present paper, I shall treat both paths as approximating to L1 acquisition.Figure 1 summarizes the context of language contact and acquisition in the 'indentured' workforce on the Monte Café plantation, stressing the role of varieties of L2 indentured worker Portuguese as a model both for L1 and L2 acquisition:

Sources of Primary Linguistic Data
Product of acquisitional process

Aspects of restructuring in Tonga Portuguese
This section presents a preliminary overview of aspects of restructuring in Tonga Portuguese in order to give notice of the potential importance of this variety to discussions of semicreolization and contact situations involving European Portuguese.Two clarifications regarding the methodology are warranted.Firstly, as this is a <first-notice> paper, and since it would be most unwise to embark on an extensive morphosyntactic analysis of variable data without first having conducted a detailed variation study (as, for example, that of Baxter, Lucchesi & Guimarães, 1997), the following discussion will naturally be general and highly restrained.I will adopt a cautious qualitative approach at this stage, while keeping quantificational treatment of the topics discussed here as future research goals (Baxter, 2001).Secondly, I shall compare restructured Tonga Portuguese with Modern European Portuguese (henceforth MEP).This is done in order to evaluate the developmental distance of Tonga Portuguese from MEP for wider comparative purposes.I assume that MEP, in the foundation period of the plantation (post-1854), and constantly present through Portuguese overseers and administrators throughout the colonial period, provided one model for the acquisition of Portuguese by the Tongas. 12In the foundation period, MEP would have been a prime contributor. 13he speech of the oldest informants displays a great deal of variation at all levels.Within this variable system there is a common set of restructurings which differentiate the language from Portuguese, many of which are reminiscent of the grammars of Creole Portuguese languages and of dialects of Brazilian Portuguese, especially Afro-Brazilian varieties (Mello, Baxter, Holm &  Megenney, 1998).Restructuring has taken place principally in those areas of Portuguese morphosyntax which rely heavily on inflectional morphology, yet it has also noticeably altered the morphosyntax of independent Portuguese morphemes.Both processes can be attributed to the role and nature of previous generation Tonga L2 Portuguese (and Tonga L1 Portuguese, itself fed by Tonga L2 Portuguese) as PLD to L1 Portuguese acquisition for the oldest age group recorded. 14The well known difficulty for adults to acquire inflectional morphology, at least in the early stages of L2 acquisition, and their reliance on relexified structures (DeGraff, 1999:494-98), would have been very much apparent in adult Tonga L2 Portuguese, and it has had serious consequences for 12 The Portuguese used by the European Portuguese on Monte Café was not necessarily metropolitan (see Baxter, 2001), and may have been deliberately simplified, but its basis was still MEP. 13Of course the principal linguistic models involved in the dynamic contact situation envisaged in Figure 1 are Tonga Portuguese models. 14The broad acquisitional perspective sketched here has been around, in one form or another, since the mid 1970s, and is very well-known in the research literature (cf.Andersen, 1983).In this paper, I follow a very standard view of language acquisition whereby PLD provides the basic information for the setting of parameters (Chomsky,  1981).On Monte Café, as in cases of creolization, the PLD of the oldest generation of Tongas interviewed were not of a uniform L1 origin.Rather, they were strongly (influenced by) Tonga L2 Portuguese and by Tonga L1 Portuguese (itself influenced by L2 Tonga Portuguese through PLD) and contained a good deal of variable and inconsistent material, including a narrower range of structures and relexified structures, all typical of early-stage L2 acquisition (cf.DeGraff, 1999:494-498).It is largely on the basis of this kind of PLD that the L1 learners of Tonga Portuguese set their parameters.The idea that the form of the grammars of creole languages may be related to the nature of PLD has long been a central topic in creole linguistics, and has been used to feed different theoretical perspectives, for example creole universalist (Bickerton, 1974, 1981, 1999) and substratist (cf.Lumsden, 1999).What the reader needs to bear in mind is that this process in the community in question was strongly cyclic, driven by the sociolinguistics of contractural labour, creating a situation in which Portuguese was constantly acquired as a second language by adult contract workers.Following the sociodemographic models of plantation societies proposed by Baker & Corne (1982) and Bickerton (1988), I assume that adult L2 models competed with L1 models both through PLD to subsequent generations and also significantly in the speech community.
the acquisition and development of Tonga L1 Portuguese. 15I shall deal first with inflection, both in the NP and in the VP.

Agreement rules
The three categorical agreement rules (henceforth Agr.) of Portuguese have not been fixed in the grammar of the oldest age group.In this sense there has been a collateral restructuring in the direction of a system devoid of agreement.

Agr. in the NP
Number.For the oldest Tongas recorded, we find that NP number agreement is most infrequent.In the few instances where number inflection does occur it is marked on the determiner, not on the nucleus (as in (1)a).As expected, we find that other mechanisms express number; numerals or quantifiers (as in (1)b), especially the quantifier tudo 'all'.However, NPs totally devoid of number indication and dependent on context, as in example (1c), are common: 15 One of the reviewers of this paper raises the question that, if children in bilingual contexts can discriminate two or more languages, then why can't they do it here?That is, I assume, why couldn't Tonga children discriminate between Tonga L1 Portuguese and Tonga L2 Portuguese and choose the L1 PLD over the L2 PLD?This is a very interesting, and potentially complex, issue to which I can only respond briefly within the confines of this paper.Firstly, having PLD from L2 and emergent L1 versions of the same language, where the emergent L1 version was historically (and cyclically) derived from L2 versions, and where L2 speakers were numerically dominant, is not necessarily a clear case of usual bilingual acquisition.Secondly, even if children can choose PLD originating in L1 over PLD originating in adult--acquired L2, where L1 and L2 are versions of the same language, surely there are at least three important conditioning factors: (i) the degree of relative exposure to L1 and L2 versions of that same language during acquisition; (ii) the degree of structural difference between the L1 and L2 versions; (iii) the proportion of adult L2 speakers and L1 speakers of that language in the speech community concerned.
In the Tonga speech community, it is a sociohistorical fact that new generations developed their Portuguese increasingly on the basis of Tonga L1 Portuguese PLD, by virtue of children being born to L1 Tonga Portuguese speaking parents.Research on the use of the noun phrase number agreement rule by twenty-four Tongas in three age-groups strongly supports this (Baxter, 2001).The probability that a Tonga with only Tonga parents would attach a plural morpheme is pr.63 (p.= .029).However, a Tonga who had at least one L2 Tonga Portuguese parent disfavoured the use of this same agreement rule (pr.0.36, p = .029).This research suggests that, either the choice of L1 PLD over L2 PLD is irrelevant, or such a choice can be overturned by factors such as (i), (ii) and (iii) above.Further research into other areas of Tonga Portuguese grammar, using quantificational methods, should help throw further light on this issue.
(1) a. issos casa R-f.223 this-NEUT-PL+house[FEM] 'these houses (lit.these house)' b. tuto casa já fechô 16 A-f.950all-NEUT+house[FEM]+already+close-PRET.3SG 'all the houses were already shut (lit.all house already shut)' c. isso criança que papa17 ta vere 18  Gender.The fate of NP gender inflection, whether within the NP or between subject and predicate NP, is similar.Instances of agreement, as in (2)d, are infrequent.Portuguese gender bearing determiners have become fixed forms, based either on masculine or feminine gender forms.Thus, for example, the (principal) first person singular possessive determiner is minha/mia, derived from the Portuguese feminine gender form.The demonstratives isso and aquele, invariant for gender in Tonga Portuguese, derive from Portuguese forms which require masculine gender agreement.

Agr in the VP
The categorical rule of subject person/number Agr. on the verb in Portuguese, has similarly not been fixed, as we see in example set (3), although here too there is an extremely low level of agreement in the first person (sg and pl) of certain common verbs, such as the verb "know" in example (3d): (3) a. vocês vai R-f3.1119 you-PL+go-PRES.3SG'you go' b. ami ponhó alá di fogo 21 A-f.1269I+put-PRET.3SG+adjacent+of+fire 'I put it next to the fire' c. nosso fechô porta tudo 22 S-f.3.314we+close-PRET.3SG+door[FEM]+all-NEUT'We closed all the doors' d. é qui no sê di naçá dele 23 S-3.16 be-PRES.3SG+that+NEG+know-PRES.1SG+of+ nation [FEM]+of-he 'it is (just) that I don't know his ethnic group (lit. it is that I don't know of nation of he)' The result of this restructuring is that the verb acquired in L1 Tonga Portuguese of the oldest speakers corresponds generally to the third person singular or regular base, a form commonly found in Creole Portuguese languages. 24In Tonga Portuguese, this tendency is so strong that inflected forms of irregular verbs are often based on the root form of the verb, rather than on the irregular third person singular.Thus, for example, we find fazeu 'did' (Standard: fez) and fazeva 'did/used to do, would do' (Standard Portuguese: fazia), rather than the standard irregular third person forms.Furthermore, there are two very common verbs based on the root of the subjunctive/imperative of irregular Portuguese verbs: venha 'come' and ponha 'put', and these also occur inflected for tense, for example venhô 'came' and ponhô 'put', where the standard Portuguese forms would be veio and pôs. 25 The prominence of the root form of Portuguese verbs in L1 Tonga Portuguese points very strongly to the role of L2 Tonga Portuguese (of a previous generation) as input to PLD for this particular first generation of Tonga L1 Portuguese speakers.Here we can appreciate the influence of salience and frequency of target forms acquired in L2, and massive reduction of inflectional morphology.The example of the verbs based on subjunctives is especially interesting as the verbs in question would have been used frequently by Portuguese (and Tonga) overseers in subjunctive forms in embedded and direct orders addressed to workers in the context of plantation labour.
Summing up here, it might be said that the loss of the three Agr rules, NP number and gender, and subject-verb agreement, is certainly typical of varieties of Creole Portuguese, such as Capeverdean or Sãotomense.However, at the same time, these characteristics have been closely observed in L2 contact Portuguese among Amerindians in central Brazil (Emmerich, 1993; Lucchesi,  1999).Indeed, in Brazil's past, this type of phenomenon, via PLD strongly influenced by adult-acquired L2 Portuguese among Africans and Amerindians, may have been responsible for the introduction of variable agreement into Brazilian Portuguese (Mello et al, 1998).So, while the possibility of some influence from Cape Verde Creole Portuguese (henceforth CVC) and São Tomé Creole Portuguese (henceforth STC) in models of L2 Portuguese can't be totally excluded, it seems reasonable to argue that the acquisition of L2 Portuguese in the fluid contact situation on Monte Café would have been a sufficient motive alone to bring about models for this restructuring relative to Portuguese inflections. 26

Restructuring in the area of aspect and tense
Here we find that Portuguese inflectional morphology representing aspect and tense, and the distinction of finite/non-finite, have also not been acquired categorically.Nevertheless, in this case there has been a greater degree of fixing of these inflections than in the case of agreement morphology.The oldest the other based on the third person singular of the indicative, vai.The presence in Tonga Portuguese of a verb based on the root form ponha finds an interesting parallel in Brazilian rural dialects, where the verb ponhar is quite common. 26Recently, Naro & Scherre (1998)  speakers in our sample display a variable verb system in which inflectional forms approximating to the grammar of European Portuguese coexist with non standard non-inflecting forms in the same apparent functions.Let's first look at what has been acquired matching Portuguese.There are only three inflecting tense-aspect forms: which are the present, example (4), and two past forms (i) the simple preterite, for 'past perfectives', as in example (5); and, (ii) the imperfect, for 'past imperfectives', as in example (6).Progressives are represented by an analytic construction which we shall discuss shortly.Other Portuguese finite verb inflections have not been acquired. 27So, for example, there is no subjunctive, and there is also no synthetic future or conditional, although these would have been largely absent in the European Portuguese model, as they are infrequent in informal speech.
(4) é isso memo qui nosso fala aqui S-3.66 be-PRES.3SG+this-NEUT+same-MASC+that+wesay-PRES.3SG+here 'it is just this that we say here' (5) quando chegô aqui ele pariu ami A-f3.1249 when+arrive-PRET.3SG+here+he+givebirth-PRET.3SG+I'when (she) arrived here she gave birth to me' (6) gende també qui fazava tu tembo, festa, nõ ta aqui, morreu Af3.1087 people+also+who+do-IMP.3SG+all+time+party+NEG+be-PRES.3SG+die-PRET.3SG'also people who used to do everything those days, parties, aren't here, (they) died' Now, what about the non-standard forms in finite functions?A verb in the form of the Portuguese third person singular present indicative, or root form, may express either past 'perfective' as in (7), or past 'imperfective', as in (8). 27All of the characteristics just listed find parallels in the Afro-Brazilian Portuguese dialect of Helvécia (Baxter, 1997).It is hoped that future comparative research involving Tonga and Helvécia Portuguese data may help shed light on the genesis of the latter dialect. (
In the speech of the oldest informants, 'non-standard' forms in past time contexts are frequent.The following texts (examples ( 9) through ( 11)) give a further idea of the range of non-standard verb forms found in past time contexts: (9) Saiu ãla mesimo de cidade delé.….
28 While ni in this example could have been triggered by the interviewer's use of the same item, examples (9), reduced to n' in this case, ( 10) and ( 16) provide other examples of the locative-directional preposition ni, which corresponds to 'in', 'to' and 'from'.In European Portuguese these functions are represented distinctively by locative em (and its contractions with the definite article na, no, nas, nos, etc.), directional a or para, and source de.However, in the speech of the oldest speakers of Tonga Portuguese all three functions may be expressed by ni, with de also expressing source.
The preposition ni also occurs in São Tomé creole, there it only has the functions of locative and directional, and the source function is indicated by di.However, motivation for the conflation of the three functionally distinct forms of Portuguese into one form may be readily found in Umbundo.Here there are three prepositions which overlap slightly: the locative preposition ku/ko, which may express direction and location, and the preposition mu, which expresses location (within), and direction (both into and out of an enclosed space), and the preposition mo which expresses location (on).The semantic feature of location common to all three associates them strongly with Portuguese em (and contracted forms), whereas the feature of direction common to two of them helps to associate Portuguese de and em in the case of Umbundo mo.The preceding four examples reflect a strong tendency in the speech of the oldest informants.There is a mixture of standard inflected forms, together with non-standard forms.On the one hand, broadly 'perfective' functions are represented by standard Portuguese preterite inflections.This is the case of the active verbs saiu, moreu and landjô, in example ( 8), and the stative verb tendeu, with a telic reading, in example ( 9).However, in some cases a non-standard verb, that is a root form, may represent a (past) 'perfective' specific event.Such is the case of the active (aspectually telic) verb in nosso fecha 'we shut (the doors)', in example ( 9).On the other hand, the remaining non-standard verbs represent broadly (past) 'imperfective' functions Thus, in example ( 8), the stative verb te doesn't carry standard imperfect inflection, although it is in a past context.Rather, its form approximates to that of the Portuguese present indicative.Similarly, the stative verbs sabe, in example (9), and é and chama in example (10), also do not carry the standard Portuguese imperfect tense-aspect inflection.All these stative verbs occur in a form approximating to the Portuguese present indicative or root form.Similarly, the stative and active verbs in example ( 11), all of which may be interpreted as representing habitual past actions, are all uninflected for the standard Portuguese imperfect which would be required in this context, and appear in the root form.
Note also that the auxiliary verb ta in the verb phrase ta passare, the final verb in example (9), does not carry the standard Portuguese imperfect inflection, which in Portuguese would be required on a progressive structure in a past context.This is a particularly interesting example, because it represents a strong tendency in the data from the oldest speakers.Here we can appreciate the potential to restructure the Portuguese progressive structure of estar + a + infinitive to yield a structure consisting of an uninflected aspect marker ta plus verb: ta + verb.29 oldest informants have generally eliminated the Portuguese preposition and, in some cases, the verb in this structure matches the 3rd person singular or base-form of the verb, as in example (13c).
The distribution of verbal inflection in examples ( 7)-( 11) is reminiscent of observations in the literature that, in L1 acquisition verbal morphology develops initially to encode aspect, and in L2 acquisition of Spanish and Portuguese, the preterite is acquired prior to the imperfect.Andersen (1990:58-9)  has related this finding to the effect of distributional bias in the model (in terms of frequency in input and the connection between the semantics of certain verbs and the preterite).Subsequently, Shirai & Andersen (1995) have proposed that this acquisitional order is also owed to the fact that that children restrict their use of T-A inflections to the prototype of the past category, to verbs with the features [+telic], [+punctual] and [+resultative].
Decreolization, it seems, operates in a similar way.Thus, Bickerton  (1975:150) noted that past inflection occurs first on [+punctual] strong verbs, in the Guayanese Creole continuum.From Bickerton's perspective (that new forms appear in old functions), if these inflected verbs in Tonga Portuguese are appearing at the first stage of decreolization, then they should be in Creole functions.And this may well be the case, because the preterite inflection occurs on verbs with punctual and telic type functions.But the problem is to know what forms they are replacing.Within the prototypical Creole system Bickerton proposed, the insertion of inflection in such cases would be replacing a system wherein past punctuals were not indicated by a marker and past anteriors were so marked.However, in the data considered, there is no evidence of such a system, and this seems odd when one considers that the oldest speakers in the data base were born in the third decade after the foundation of the plantation. 30n alternative interpretation, which seems more plausible, is that earlier models of Tonga Portuguese did not have tense marking as such.In other words, the variation in question is merely a more regularized continuation of a pre-existing tendency in Tonga L2 and L1 Portuguese, with both inflected and uninflected root-form 'perfectives' representing the same broad function with appropriate aksionsart verbs.
In the present case, it does not seem unreasonable to interpret the above tendency in the speech of the oldest informants is a reflection of the nature of PLD available to children the process of acquiring 'Portuguese' in this particular setting and which permitted a stronger fixing of preterite inflection for closed events.In particular, I suggest that models of L2 Portuguese available must have displayed this same tendency, the preterite, and showing a diffuse relationship between the Portuguese imperfect inflection and non-perfective functions, the latter perhaps being more clearly represented by the base form of the verb.An interesting overall implication of this perspective, is that the verb system of today's oldest speakers of Tonga Portuguese is more aspectually orientated than time orientated.
A valid final question is whether the T-A systems of STC and CVC could have influenced L2 Portuguese on Monte Café.In STC bare statives (i.e.verbs  marked for T-A) may occur with either a past non-habitual or habitual reading (Lorenzino, 1998).So, in this latter respect there is a parallel with Tonga Portuguese.Active verbs in STC also show some similarity with Tonga Portuguese, the bare verb having a past non-habitual reading.However, STC is different in that active with a past habitual reading must take a preverbal marker ka.
The situation in CVC is somewhat similar.The system described by Silva  (1990) has the following characteristics.On the one hand, bare classical statives such as tem 'have', and sabê 'know', don't have a past reading, although semistatives such as creditá 'believe' and squicê 'forget' do.However, the habitual reading in the last group requires a preverbal marker sta.On the other hand, bare actives may be interpreted as representing past single events, yet, a past habitual reading requires the habitual/iterative preverbal marker ta plus a past suffix -ba.
While both these Creoles may have some bare past statives, bare past actives have only non-habitual readings.So, if plantation workers of African mainland origin were in contact with these languages, the treatment of statives in the L2 input to L1 Tonga Portuguese could have been influenced by these Creoles.However, as stated earlier, in sections 1-3, there is good reason to believe that such contact, and especially that with STC, is not relevant.
A potentially more important influence may have come from Umbundo.Here too there is a parallel with Tonga Portuguese because statives with 'present tense' prefixes permit a past continuous reading.The same applies to active verbs.However, if the event of the verb is considered effectively closed, i.e. broadly speaking 'perfective', the verb is assigned an appropriate preterite prefix and suffix. 31This situation described in traditional grammars of Umbundo-Kimbundo holds to a large extent in Monte Café Umbundo koiné (although the repertoire of different preterites displays some erosion).It seems possible that the (T)-A system of Umbundo could have lent some support to a bias towards preterite inflection in Tonga L2 Portuguese.

The distinction of finite/non-finite verb
As mentioned with respect to the progressive aspect structure ta + infinitive, the distinction between finite and non-finite verb is blurred in certain contexts.The usual of these are either in complement clauses introduced by pa/pra 'for, to', as in example (12), or in verb phrases involving an auxiliary, as in ( 13): (12) isso dinhero é pra compra comida [R-f3.1243]this The nature of this variation also suggests that this distinction may not have been functional in L2 Portuguese and L1 Tonga Portuguese which provided input to L1A by these speakers.While the source of this variation probably largely lies in the problem of fixing inflectional morphology, in the case of complement clauses, there is an additional complication in that different Portuguese verbs select different complement structures, some infinitival introduced by para and others finite introduced by , para que or que.The only consistent complementizer in the speech of the oldest informants is pa/pra, based on Portuguese para.It is possible, that in a context of diluted inflectional morphology, the confusion and conflation of complementizers would also further contribute to this variation in the verb.

Restructuring relative to Portuguese free grammatical morphemes
Here I shall briefly consider 2 areas where restructuring has taken place relative to free grammatical morphemes, specifically, the development of a double discontinuous negative structure and the signalling of reference in the NP.Both cases are potentially controversial as they may or may not represent evidence of continuity in Tonga Portuguese of influences arising from past contact with CVC and STC, in the case of the article system in the NP, and São Tomé Creole alone, where the negation system is concerned.However, it is my view that both restructurings may be readily attributed to the effect of substrate influence on the form of L2 varieties of Portuguese.

Restructuring of Negation
Tonga Portuguese has two principal negative structures: a simple preverbal negation, and a double discontinuous negative structure (DDNeg), both with very specific functions.The simple structure functions as a general assertion negator, as in examples ( 14 In (15)a, the DDNeg structure is used in response to a negated question.This resembles the function of double negation identified by Schwegler in work on Brazilian Portuguese and Afro-Colombian and Dominican Republic Spanish (Mello et al, 1998: 104-107; Schwegler, 1991, 1996).That is, the question posed contains a presupposition of the type [I would expect X to be the case], and the negative response contradicts that presupposition.Example (15)b represents what might be called an absolute negative, of the type [it is definitely not the case that X].

What is the source of this double discontinuous structure?
In European Portuguese all three circumstances would be handled by simple pre-verbal negation, with the optional addition of já in case (15)b.However, an important point here is that in European Portuguese the negative response to an affirmative or a negative question such as ( 16)a can be constructed with an extra negator placed immediately after the verb, as in ( 16)b., or with the extra negative placed at the beginning of the clause, as in ( 16 Hence, Portuguese could provide a partial model for the development of the Tonga structure in one of its functions.Note that the forms of the Tonga negators, não/nã/nõ + não/nã/nõ/na are close to the Portuguese negator não, given variable denasalization of vowels in Tonga Portuguese.

What could have triggered the restructuring of Portuguese negation in
Tonga Portuguese?
On the one hand, if there had been some kind of continuity of São Tomé Creole Portuguese influence on the Monte Café plantation last century, at a time when speakers of STC and African recent arrivals were in contact, then STC could have provided a model for DDNeg.STC has double discontinuous negation as its sole negative structure, and its form is na + fa.However, for reasons mentioned earlier, in sections 1 and 2, I think this unlikely, since the Monte Café plantation was founded in the year of abolition of indigenous slavery, and appears to have relied completely on imported African field labour at that time.
On the other hand, the African languages of the plantation workers would have furnished ample potential structures for relexification to yield DDNeg structures in L2 Portuguese.Thus, for example, Umbundo has two principal negative structures.One is preverbal, through a prefix. 33The other is double discontinuous and uniform, however, it is limited to nominal and adjectival predicates.This state of affairs is confirmed in the Monte Café Umbundo koiné in data elicited from informants, examples ( 17 Here we note that the second negator in the DDNeg structure is na, a frequent variant in the second position of this structure in Tonga Portuguese.Given denasalization in Tonga Portuguese, Umbundo na would have provided a ready association with denasalized Portuguese não in L2 varieties of Portuguese.Furthermore, the earlier presence of speakers of northern Angolan languages, such as Kikongo which has dialects with DDNeg consistently on verbs (Maia, 1961), could have reinforced the presence of DDNeg in L2 varieties.DDNeg would have taken root in Tonga Portuguese as a result of a confluence of partial models in Portuguese and in the substrate.

Restructuring of the article system
Whereas (European) Portuguese requires the presence of a definite article for definite, and generic reference, and an indefinite article for indefinite reference, Tonga Portuguese has restructured the Portuguese reference system of the NP.An NP may occur with a determiner (<demonstrative) when it indicates a first introduced instance of definite reference, as in (18.line 1], but on subsequent mentions the same NP may occur without a determiner, as in (18.line 2], gente.Also an NP bearing to generic reference occurs without an article, as in example (19).The definite article of Portuguese has simply not been acquired, and the indefinite article has been weakly acquired, usually in the Portuguese feminine gender form uma, as in (18.line 7]: (18) 34 1. [ S-f3.1664-1694]  Now this man left machete there.This machete has name.
Here we see that first mention of an NP in discourse requires uma, (18.lines 3 and 7) and in subsequent reference the same NP may occur without a referential determiner or, if it is topic prominent (e.g.reintroduced), as in the case of the NPs muyé and home, the central actors in this narrative, reintroduced, it may occur with a demonstrative, such as isso, as in (18) lines 7, 10 and 11.
While this type of system has parallels in the NP reference systems of the Creoles of Cape Verde and São Tomé described by Lucchesi (1992), in Tonga Portuguese, in the context of the Monte Café plantation, it would seem more appropriate to attribute the principal source of restructuring to the influence of Bantu substrate on the L2 Portuguese of the indentured workers.
The reference system in Umbundo is quite similar.There are no definite articles, although there is a complex 9-tiered noun classification system of prefixes.The NP shows definite reference, or generic reference, without the presence of markers.An NP of indefinite reference may use a numeral, as in example (20)a.However, a topic prominent NP, or an NP requiring referential clarification, may occur with a demonstrative (or other clarification, such as 'same'), as in example ( 20 João 3SG.PRET-give-PRET+Pedro+dog+(this+(same)) 'João gave the dog to Pedro'

Conclusion
While Tonga Portuguese is clearly the result of a restricted situation of language contact in which MEP, and Tonga Portuguese, were the targets of acquisition in a classical (colonial) plantation context, the set of sociolinguistic factors conditioning the language contact conspired to create a language, Tonga Portuguese, which, structurally, lies between European Portuguese and the better known varieties of Creole Portuguese.
The prime factors behind this result would appear to be the constant importation of large numbers of Angolans, and the limited extended contact between plantation workers of different ethnolinguistic backgrounds.Among the plantation workers from Angola, this led to the use of an Umbundo-based koiné as a primary language of intragroup communication, and this language was transmitted generationally amongst the Tongas.The principal motivation for the acquisition of Portuguese as a L2 by adults was the work situation, not intragroup communication, although this did have a minor role.The constant presence of a large number of Africans of fairly homogeneous Bantu background speaking L2 Portuguese meant that, alongside the L1 acquisition of Umbundo, varieties of adult-acquired L2 Portuguese (and, eventually, Tonga L1 Portuguese strongly influenced by Tonga L2 PLD) served as input to PLD for Tonga offspring.And this Tonga Portuguese, as an adult-acquired L2, and as emergent L1, served as a means of communication with other groups, plantation workers from Mozambique, Ajudá, Gabon and Cape Verde.However, Tonga Portuguese did not achieve the degree of vernacularization that would have led to the emergence of a stable Creole.For the community under study, while in the 1940s its Portuguese L2 and Umbundo L1 elements were almost completely removed by repatriation, this event coincided with a liberalization of the plantation conditions, under provincial governor Sousa Gorgulho, which gave the Tongas freedom of movement and, hence, greater contact with other varieties of Portuguese, hastening changes in Tonga Portuguese towards metropolitan oriented varieties of Portuguese.Thus, the incipient Creole was side-tracked by recent decreolization.Nevertheless, a good many of the Tonga Portuguese features considered in this paper were passed on to the second generation, although variable inflectional morphology is less frequent than in the previous generation.
The greater amount of inflectional morphology found in Tonga Portuguese, when compared with varieties of Creole Portuguese such as that of CVC, or STC, may be attributed to the fact that the Tongas constituted a permanent ethnolinguistic group which slowly accumulated on the plantation and whose members often had greater contact with administrators.Indeed some of the overseers were drawn from the Tongas, as were carpenters assistants, and housemaids.Nevertheless, the data considered in this paper come from Tongas who were field labourers.
Tonga Portuguese shows the classical ingredients to creolization, of an input heavy with data from L2 acquired in a contact situation.Thus we find superstrate and substrate influences, and some independent developments, which point to processes of mixing in the L2 continuum.Notable are the clearer instances of transfer of structure from the homogeneous substrate: DDNeg and the NP reference system.However, at the same time, we find items which must have been selected because of the conditioning factors of saliency, and frequency (e.g. the consistency of preterite marking on verbs representing punctual events) and transparency (e.g. the promotion of the progressive structure).We also see evidence of reinforcement provided by particular contexts, such as in the case of the verbs based on imperative/embedded orders, such as ponhar.While I have not ruled out the possibility that two other Creoles, Cape Verdean and Sãotomense, could have contributed to the pool of models available for PLD, there appears to be ample evidence that the developments in question (Neg2, NP reference system, tá + Vinfinitive) could arise as independent developments.
Where does the Tonga Portuguese of the eldest speakers lie in relation to the concept of semicreolization put forward by Holm (1998)?Is it the product of (i) decreolization; (ii) partial restructuring; or (iii) the result of contact with a Creole?While (iii) must be borne in mind, in view of the form of the article system, which resembles that of both CVC and STC, and DDN, which resembles that of STC, I believe that these may be easily attributed to the pressure on the L2 interlanguage continuum from a homogeneous substrate sharing these features.Moreover, the sociohistorical facts appear to argue against a decisive role for STC.
Weighing up decreolization versus restructuring is more problematic than it seems initially.Was the language of previous generations even further removed from Portuguese than the language of today's oldest speakers?Or does language of today's oldest speakers represent fairly closely a restructuring which did not progress much further than this?Linguistic evidence alone in support of decreolization or restructuring is difficult to determine.In reality, there is nothing in the creole-like structures of Tonga Portuguese which will allow a clear decision one way or the other.Rather, it is the historical sociodemographic evidence which lends support to the restructuring case.
The majority of workers brought in to Monte Café in the 19 th century were Angolans, and those imported in the final decade were almost entirely from the Benguela region.In these circumstances, Bantu languages, possibly Umbundo, prevailed in intragroup communication.The Tongas, slowly accumulating as permanent bilingual residents of the plantation, were in a situation whereby some of their number came to act as 'intermediaries' between African and European.Thus, it would appear that, on the one hand, there was no need for the development of a new intragroup language because the workforce was, for a long time, largely linguistically homogeneous.On the other hand, there was always some degree of bridging between Tonga Portuguese and European Portuguese because of the status of the Tongas as <plantation property>, so the degree of divergence from Portuguese is 'relatively' minor, yet logically reflects the homogeneous substrate of the Tonga's Angolan forebears.
Verde Creole a  L1 Cape Verde Creole b) *Subscripts< a >and < b > represent different generations in the case of L1s, and different waves of contract workers in the case of L2s.

Figure 1 .
Figure 1.Language acquisition and transmission among the indentured workforce on the Monte Café plantation.

Table 3 .
Monte have presented some interesting evidence that varieties of Old Portuguese and some modern dialects of European Portuguese display variable NP and VP agreement rules.Thus, they argue, European Portuguese would have been a source for the introduction of variable agreement into Brazilian Portuguese.These findings might have some bearing on the nature of Tonga Portuguese and must be kept in mind, especially as we do not know what type of Portuguese the overseers spoke in earlier periods.Nevertheless, two former plantation workforce overseers, elderly Portuguese, do not show evidence of variable agreement.Whether or not there was variation in the European model, it seems reasonable to argue that the overwhelming numbers of L2 speakers of Portuguese in plantation situations would have promoted a system devoid of overt Agr, regardless of superstrate input.

gendé nõ sama elé sõ nã, sama "canoa".
If we add to this the fact that initial[d]would be [nd], we see that it would not be difficult to produce ni in L2 Portuguese influenced by Umbundo.Most varieties of Brazilian Portuguese use the preposition em in a contracted form with the definite article (na, no, nas, nos) to express both direction and location, while many rural dialects of Brazilian Portuguese use a preposition ni for the same functions.It is quite likely that these conflations of functions might also be related to a strong Portuguese L2 phase in the history of Brazilian Portuguese.

Ainda isso gente qui tomô n'ardea, tudo no tá, no ta aqui maze.
)c, both of which negate a perceived assumption on the part of the inquirer: