The tense-mood-aspect systems of the languages of Suriname 1

This chapter deals with tense, mood, and aspect (TMA) marking in the languages of Suriname, focusing on the stability of forms, meanings, and structural patterns. Despite its prominent position in the creolization debate and occasional mentions in the literature on linguistic areas, studies on TMA in (non creolization) contact settings in Suriname are relatively few. TMA has been studied in detail in the world’s languages, however, in terms of: – typology (Dahl 1985, 2000; Boland 2006; Dahl and Velupillai 2011a, 2011b, 2011c, 2011d; Dryer 2011; Velupillai 2012); – creolization (Singler 1990; Bakker et al. 1994; Winford 2001; Velupillai 2015: 391–403); and – historical development & grammaticalization (Bybee et al. 1994)


Introduction
This chapter deals with tense, mood, and aspect (TMA) marking in the languages of Suriname, focusing on the stability of forms, meanings, and structural patterns. Despite its prominent position in the creolization debate and occasional mentions in the literature on linguistic areas, studies on TMA in (non creolization) contact settings in Suriname are relatively few. TMA has been studied in detail in the world's languages, however, in terms of: -typology (Dahl 1985(Dahl , 2000Boland 2006;Dahl and Velupillai 2011a, 2011b, 2011c, 2011dDryer 2011;Velupillai 2012); -creolization (Singler 1990;Bakker et al. 1994;Winford 2001;Velupillai 2015: 391-403); and -historical development & grammaticalization (Bybee et al. 1994) In a relatively short period of time, Suriname has seen numerous, often radical, linguistic developments due to its many languages, pervasive multilingualism, and array of contact scenarios. We will investigate the vulnerability of features to contact induced changes in the TMA systems of the Surinamese creoles, Surinamese Dutch, Sarnami, and Surinamese Javanese.
Although some linguists believe that any type of borrowing or structural influence is possible in principle in a bilingual setting (e.g. Thomason and Kaufman 1988), much work has been done in attempting to determine which forms are most borrowable (e.g. van Hout and Muysken 1994) and which structures are most stable (e.g. Cysouw et al. 2008). Despite this, there is still little agreement about borrowability and stability hierarchies.
Borrowability (the likelihood that a language will take a form from another language) and stability (resistance to change) are fundamentally different since one necessarily involves language contact and the other does not. Borrowing and stability are different in that studies on the former tend to deal with external processes involving some component of language contact while the latter tend to target internal developments. Sakel (2007) has helped refine our perception of borrowability by systematically distinguishing borrowing of forms and copying of patterns in contact settings: the distinction between matter (MAT) and pattern (PAT) borrowing.

Borrowing and borrowability
We will first consider MAT borrowing. Several general borrowing hierarchies have been proposed for concrete lexical or morphological forms (e.g. Haugen 1950;Weinreich 1953;Muysken 1981;Thomason and Kaufman 1988;Matras 2007). Muysken (1) and Matras (2), for example, propose hierarchies based on case study data. Muysken's study investigates Spanish borrowings in Quechua, while Matras' findings are based on a sample of 27 focus languages in broad geographic and typological distribution. Though they arrive at somewhat different outcomes, there are some important similarities, e.g. nouns are most borrowable while bound functional elements tend to be more difficult to borrow. Other methods also provide inconsistent results for general borrowability scales.
(1) Van Hout and Muysken (1994: 60) noun, name < adverb, complementizer, conjunction, exclamative, negation, preposition < adjective, auxiliary, copula, verb < numeral, quantifier, wh-word < demonstrative, determiner, preposition+determiner, possessive pronoun, personal pronoun, pronominal clitic (2) Matras (2007: 61-62) nouns < conjunctions < verbs < discourse markers < adjectives < interjections < adverbs < other particles, adpositions < numerals < pronouns < derivational affixes < inflectional affixes Other work, e.g. Tadmor et al. (2010), shows similar trends, where nouns are more borrowable than verbs and content words more borrowable than function words. They also demonstrate that grammatical categories are not the only factor determining borrowability; certain semantic fields are more frequently borrowed than others. Others (Pagel et al. 2007;Pagel 2009: 411) suggest that the frequency with which words are used predict their resistance to change. Perhaps because TMA is expressed through a variety of means in different languages -e.g. super-segmental markers such as tone, adverbs, clitics, inflectional morphology -and within individual languages, it plays a marginal role in the proposed borrowing hierarchies. On the whole, it would figure relatively low on these hierarchies.
Based on Matras and Sakel's (2007a) 27-language sample, Matras (2007) proposes a number of micro-hierarchies for matter and pattern replication. Of particular relevance here are those relating to TMA. Modality is more susceptible to contact induced change than aspect, which is more susceptible than future tense, etc. A further hierarchy was also posited for modal categories. The < arrows indicate the direction of the implicational pattern in the data.

Stability of structures
Recently, linguists have taken an interest in stability of linguistic structure with the idea that certain features are more resistant to change. Structural features would thereby provide insights into language evolution at a greater time depth than possible through the comparative method (cf. Dunn et al. 2005;Dunn et al. 2008). While others (e.g. Greenhill et al. 2010) contend that structure is more susceptible to change than vocabulary, a number of efforts have been made to determine whether there are universally stable linguistic structures (and if so, what are they?), and/or whether stability of particular features depends on language families. Since the methodologies and results of these studies do not always lend themselves to user-friendly comparison, Dediu and Cysouw (2013) have reviewed a number of such studies and made their outcomes comparable through statistical conversion. Each of the works they included applied measures to data from the World Atlas of Linguistic Structures and despite that all used the same source of data, none concluded with the same stability hierarchy. Those WALS features pertaining to TMA have been extracted from the studies in Dediu and Cysouw (2013) and are presented in (4)-(8), from most to least stable.
(4) Cysouw et al. (2008) Congruence test: position of tense-aspect affixes < past tense < morphological imperative < future tense < perfect < epistemic possibility < perfective / imperfective < overlap b/w situational and epistemic modal marking < imperative-hortative system < prohibitive < optative < situational possibility Coherence method: optative < imperative-hortative system < morphological imperative < situational possibility < position of tense-aspect affixes < future tense < perfective / imperfective < epistemic possibility < perfect < prohibitive < overlap b/w situational and epistemic modal marking < past tense Rank method: optative < future tense < perfective / imperfective < perfect < imperative-hortative system < position of tense-aspect affixes < situational possibility < morphological imperative < past tense < overlap b/w situational and epistemic modal marking < epistemic possibility < prohibitive (5) Dediu (2011) optative < perfective / imperfective < future tense < past tense < overlap b/w situational and epistemic modal marking < morphological imperative < perfect (6) Parkvall (2008) P1 (all families contained in WALS): position of tense-aspect affixes < past tense < optative < morphological imperative < prohibitive < imperative-hortative system < future tense < situational possibility < overlap b/w situational and epistemic modal marking < perfective / imperfective < epistemic possibility < perfect P2 (subset of "most widely accepted families" in WALS): position of tense-aspect affixes < past tense < morphological imperative < perfective / imperfective < prohibitive < situational possibility < future tense < imperative-hortative system < epistemic possibility < optative < perfect < overlap b/w situational and epistemic modal marking (7) Wichmann and Holman (2009) optative < past tense < position of Tense-aspect affixes < perfective / imperfective < situational possibility < epistemic possibility < future tense < morphological imperative < prohibitive < perfect < imperative-hortative system < overlap b/w situational and epistemic modal marking (8) Maslova (2002Maslova ( , 2004 optative < perfective / imperfective < position of Tense-aspect affixes < future tense < past tense < situational possibility < imperative-hortative system < morphological imperative < epistemic possibility < perfect < prohibitive < overlap b/w situational and epistemic modal marking With just a glance at these hierarchies, one will notice that there are both important trends and contradictions among them. Dediu and Cysouw used a principal component analysis to rank shared features according to their stability and relative consistency in each of the methods they investigated combined. The TMA features they mention can be ranked as follows: (9) Dediu and Cysouw (2013) optative < past tense < perfective / imperfective < future tense < perfect < overlap b/w situational and epistemic modal marking Each of the eight methods applied to the same (sub)set of data has produced a unique result and Dediu and Cysouw's (2013) analysis of the eight combined outcomes provides yet a ninth ranking of the features. This suggests that a universal scale of feature stability either does not exist, or has not been satisfactorily demonstrated by quantitative methods. None of the methods mentioned in this section have accounted for social factors, such as frequency of use of particular linguistic features or the broad socio-cultural setting in which speakers of a particular language find themselves. Thus another possibility is that universals of feature stability are (partially) determined by the social setting of the languages' speakers.

The present study
The present study specifically targets stability / borrowability of TMA systems in situations of intense language contact by investigating the transfer of forms and patterns surrounding realization of TMA in a sample of Surinamese languages: the creole languages of Suriname (Anglo-creole), Surinamese Dutch (Germanic), Sarnami (Indic), and Surinamese Javanese (Austronesian). Our investigation is mainly diachronic, in that we intend to trace the development of the various TMA systems, though we also use a large set of synchronic data to supplement our findings, and in some cases, propose possible changes in progress.
For some languages, diachronic data are available, e.g. Sranantongo and Saramaccan, while for the other Surinamese creoles, we have to rely on reconstructions based on synchronic linguistic data, socio-historical data, and what we know about early Sranantongo (from which all Surinamese creoles appear to descend). In the other cases, Dutch, Sarnami, and Surinamese Javanese, a diachronic component can be inferred from comparison with closely related / ancestral varieties of the Surinamese variety, i.e. European Dutch for Surinamese Dutch; Avadhi, Bhojpuri, and other overseas Hindi varieties for Sarnami; Javanese as spoken on Java for Surinamese Javanese.
Suriname affords us a great opportunity to investigate a number of typologically different languages as well as several genetically related languages (creoles). Our sample covers the spectrum of contact scenarios: maintenance, shift, stable bilingualism, and creole formation within a single multilingual society. Conveniently, the TMA systems have been somewhat of an obsession of creolists. Similarities in creole TMA systems had been initially noted as evidence supporting the monogenesis hypothesis. But TMA (sub)systems were later also provided as evidence for several other theories of creole genesis (e.g. superstratist, substratist, bioprogram). We see creole formation as an ongoing complex layering of contact processes, involving elements from both substrate and superstrate as input, as well as language internal developments. The creoles continue to develop under pressure from contact with other creole and non-creole languages. Therefore, it is worth systematically tracing the TMA developments in the creole languages for comparison with other languages with which they interact.
The other three languages in our sample will help us to determine the extent to which the structure independent factors determine borrowability / stability of TMA forms and patterns. While the specifics of social, attitudinal, and practical aspects surrounding the Surinamese creoles, Dutch, Sarnami, and Surinamese Javanese differ, each language is an integral part of Surinamese society. And since universals are hardly agreed upon (see above), parallels we see across the languages of Suriname are more likely indicative of a strong influence of the languages' setting rather than universal tendencies or purely internally driven developments.
In the following section 2, we will provide general definitions of tense, mood, and aspect. Section 3 will describe the development of TMA in the Surinamese creoles. Surinamese Dutch TMA will be contrasted with European Dutch in section 4, followed by developments of Sarnami and Surinamese Javanese in sections 5 and 6. We will then summarize and compare the developments in order to see (a) where changes have occurred in TMA systems and (b) if those developments can be attributed to contact or internal development.

Definitions
Despite relatively agreed upon theoretical definitions, TMA categories are not always neatly separable in practice. Markers of TMA often overlap. For example, multiple categories can be conveyed with a single form, or meanings of one category can be conveyed by a marker of another depending on context. Tense and aspect are particularly linked in this respect; they both express types of temporal relationships. Temporal meaning can also be inferred from modal categories. TMA markers (or lack thereof) are often combined to derive additional meanings. Further, lexical semantics, Aktionsart, stativity/dynamicity, discourse context, and others all play a role in the conventionalized expression of temporal relations and speakers' perception and intention. Nonetheless, we are able to differentiate the concepts of tense, mood, and aspect and in some cases exemplify them in a less blurry fashion than is evident in everyday spoken language. The following subsections provide basic definitions of TMA.

Tense
Tense refers to one way in which languages conventionalize the expression of an event in time. As Müller (2013: 29) puts it, tense 'is a representation of the relationship between three points in time'. These three points, first coined by Reichenbach (Boland 2006), are the points of speech (S), event (E), and reference (R). In some cases, the point of reference coincides with one of the other points, or according to Comrie (1985), is absent, which is called absolute tense. Examples (10)-(12), taken from McWhorter and Good (2012: 118-121), illustrate a simple relationship between S and E, past, present, and future, respectively. Languages differ in the means by which they mark tense. English for example has a tendency to use inflectional suffixes to mark past, and unmarked verbs are interpreted in the present. In the Surinamese creoles however, unmarked dynamic verbs are interpreted as past events and past is marked on stative verbs by a preverbal auxiliary. Similarly, languages differ in the number of grammaticalized strategies for marking  Müller (2013: 46-57), for example, discusses a number of South American indigenous languages that morphologically mark several levels of temporal remoteness (e.g. in the past, just now, weeks/months ago, years ago).

Modality
Of the three TMA categories, modality is by far the most difficult and disagreed upon category of TMA. Most basically, modality is a "grammaticalization of speakers' (subjective) attitudes and opinions." (Palmer 1986: 16). Modality "presents a statement about the truth or realization of a state or event. It refers to the attitude of the speaker or of one of the persons involved in the situation described (Bakker et al. 1994: 247). Beyond these most basic types of definitions, the specifics of modality become muddled with a multitude of strategies and sets of terminologies from different disciplines which tend to only partially overlap and outright conflict with each other.
Boland (2006) presents modality, quite clearly, as an interplay between three parameters: sense, source, and scope. Sense consists of a linear continuum, and though she acknowledges that the number of distinctions made in the continuum depends on the language, Boland (2006: 69) lists those four distinguished by English: possibility, disposition, weak necessity, and necessity. The source describes the origin of modality. She describes three sources: epistemic (having to do with knowledge), participant-internal (where modality is ascribed to some internal characteristic of the participant), and participant-external (where modality is ascribed to external characteristics of the participant) (Boland (2006: 72). The latter can also be divided into two sub types: deontic, i.e. necessity or possibility of acts performed by morally responsible agents, and non-deontic. The interaction of sense and source are detailed in Tab. 1.
According to Boland (2006: 74) the various combinations of sense and source account for the majority of modal distinctions, but in some cases, a third parameter -scope -is useful for understanding more fine grained semantic distinctions. Scope refers to which part of the utterance a modal governs. There are three possibilities of scope: the predicate (where "the description of the relation or property predicated of the argument(s) is modified" thereby defining the relationship between the participant and the state of affairs in which it is involved, also called participant-oriented or inherent modality), the predication ("the event is situated in the real or imaginary world" also called event-oriented or objective modality), or the proposition ("the truth of the propositional content is evaluated" also called proposition-oriented or subjective modality). Scope also combines with sense and source, though there is some disagreement on the possibilities. One point of contention is whether there is a one to one correspondence between scope and source -i.e. epistemic modality would be proposition-oriented. Further not all logical combinations of scope and source are possible, e.g. participant-internal modality can only be participant-oriented.

Aspect
While tense locates an event in time, aspect specifies the temporal structure of an event itself (Comrie 1976: 3). Like tense, languages differ in the number of aspectual categories they distinguish and the means by which aspect is conveyed. Commonly, two macro categories are often utilized by typologists: perfective and imperfective. The former conveys an event as a whole, while the latter "pays essential attention to the internal structure" of an event (Comrie 1976: 16 In the Surinamese creoles, unmarked verbs are read with perfective aspect (15). Preverbal e marks imperfective aspect in Sranan and Eastern Maroon creole (16). The Surinamese creoles will be discussed in more detail in the following section. In some languages, the imperfective can be further split into subcategories, such as habitual, continuous, or iterative (Comrie 1976: 25), though these distinctions will not play a role in our investigation.

Methodology
This article relies on linguistic data gathered in Suriname in 2010-2012 by Robert Borges, Kofi Yakpo, and Stanley Hanenberg as part of the ERC project "Traces of Contact" at Radboud University Nijmegen. Additional control data was collected by Kofi Yakpo in New Delhi, India in 2010 with speakers of Hindi, Awadhi, Bhojpuri, Maithili, and Magahi. Indonesian Javanese control data was collected in 2012 for comparison with Surinamese Javanese by Riski Lestiono, a collaborator of the ERC project and himself a native speaker of Indonesian Javanese, and Sophie Villerius during a Fieldwork Methods course in Leiden University. All language examples in this paper that come without a bibliographical reference are from our field data.
The data consists of elicited material gathered through the use of visual stimuli such as pictures, picture books (e.g. Mayer 1969), and a variety of video clips assembled as a standard elicitation kit for the Traces of Contact research group. We also collected more naturalistic data through semi-structured interviews and (un)guided conversations. Data was collected from various parts of the coastal area and the interior, from members of the different linguistic communities, speakers from ages fifteen to ninety years, and is somewhat gender-balanced.

The creole languages
There are seven creole languages that developed in Suriname that are still spoken there today. Sranan, which presumably was formed in the latter half of the seventeenth century, is traditionally the language of the Creole population (i.e. Afro-Surinamese non-Maroon) and is currently used by most Surinamese as one of the main languages (together with Dutch) of interethnic communication. The other six languages are spoken by Maroons, Afro-Surinamese people whose ancestors fled plantation slavery and formed independent communities outside the plantation area. As these communities became somewhat isolated from each other, they became differentiated due to unique linguistic developments. These Maroon languages can be further divided into two groups (a) the Eastern Maroon languages: Ndyuka, Aluku, Kwinti, and Pamaka, and (b) the Central Maroon languages: Saramaccan and Matawai (Smith 2002: 141). There is a high degree of mutual intelligibility within groups (a) and (b), but (b) has a significantly higher proportion of Portuguese functional and lexical elements, impeding intergroup intelligibility. Group (b) is therefore the most distinct from the other Maroon languages and Sranan. Unlike Sranantongo, the Maroon languages have been largely used as in-group languages, though it seems that recently a leveled Eastern Maroon variety is gaining ground as a lingua franca in the urban environment (Migge and Léglise 2011).

TMA in the Suriname creoles
The core concepts expressed by Surinamese creole TMA systems are mostly marked with preverbal markers (or their absence). The main focus of this section will be on the inventory of grammaticalized morphemes that mark concepts of TMA, though a number of auxiliary verbs, adverbs and adverbial clauses are also employed for encoding additional TMA concepts. Discourse position and context also play a large role in determining TMA interpretation in the Surinamese creole languages (Huttar andHuttar 1994: 489-493, Winford andMigge 2007: 76).
Examples of categories expressed in Surinamese creoles are past and future tense, perfective and imperfective aspect, and epistemic, deontic, and dynamic modality. The Surinamese creoles largely distinguish the same TMA categories and to some extent draw from the same set of forms to convey these categories, though the distribution of the forms in each language differs to some extent. Since by definition, creole languages are composed of forms and structures from multiple source languages as well as some degree of restructuring and innovation, the various components of Surinamese creole TMA systems will be presented alongside current views on their development.

Development of tense in the Surinamese creoles
Surinamese creoles employ a relative tense system, with reference to the speech act or other reference point. In the modern creoles, there are two tense markers, be(n) / bi (< English been) which locates an event prior to a particular reference point, and o for future tense (< English go). A difference in meaning is apparent between stative and non-stative verbs with past marking. Stative verbs marked with be(n) / bi give a simple past reading, while non-stative are interpreted with a past reference in their unmarked form. Non-statives marked with be(n) / bi convey a past-before-past (pluperfect) meaning. The stative / non-stative distinction does not play a role in the interpretation of verbs marked with o; these are always interpreted with future meaning. The expression of tense is not limited to the use of pre-verbal markers, but can also be achieved through the use of temporal adverbs or time adverb clauses, as well as particular aspect (e) and modal (sa) markers (Huttar and Huttar 1994: 489;van den Berg 2007: 185, 188, 191, 196;Winford and Migge 2007: 77-79).
Both modern tense markers are derived from English forms and are attested in other Anglo creoles. The distribution of these forms does not, however, coincide with either English or West African substrate languages. Past marking appears earlier in the sources of Sranantongo (early eighteenth century), while o is not attested as a future marker until the late eighteenth century (Winford 2006: 105, van den Berg 2007Winford and Migge 2007: 95;Migge and Goury 2008: 322).
Possibly resulting from the relatively late grammaticalization of o, there were several forms used in Early Sranan to express future time reference, as shown in Tab. 2. There are two hypotheses for the variation and development of future forms in Early Sranan. The first is that sa (<Du zal 'shall' or English 'shall') existed first as a future marker, but began to develop modal meanings in the late eighteenth century, after which the go / de go construction became the preferred construction for indicating future tense (van den Berg 2007: 188-199;Migge and Goury 2008: 326). Though neither 'movement toward a goal' grammaticalizing into a future morpheme, nor future markers developing into modal markers are cross linguistically rare developments (c.f. Bybee et al. 1994), this hypothesis does not completely account for the order in which these morphemes developed or the current distribution of sa across modal categories in the various Surinamese creoles (Migge and Goury 2008: 326).
According to the other hypothesis, by the late eighteenth century, sa and (de) go may have already developed into sociological salient variables associated with different varieties of Sranan, the Bakratongo (more European) and the Nengretongo variety, (more African), respectively (van den Berg 2007: 199;Migge and Goury 2008: 326). Migge and Goury (2008: 327) cite the high proportion of African slaves to Europeans (24: 1 in 1783) as the probable reason that the Nengretongo variant expanded to the Bakratongo variety and later reduced phonologically to o. The further development of sa as a modal marker will be discussed below.  morphemes used to mark each category in several Surinamese creoles. The most variation can be found in the realm of potential. Other categories are marked with relative consistency.

Modal category Early Sranan Saramaccan Ndyuka Pamaka Sranan
The most variation among the Surinamese creoles is found within the potential category. These sub-categories are marked with the forms sa, man, kan, and poi, though the distributions across the different languages are not consistent, reflecting unique developments of each language. Migge and Winford (2009: 129) argue that the potential categories of Maroon creoles are largely modeled on Gbe patterns, while Sranantongo exhibits additional internal developments and effects of contact with Dutch.
A number of complex developments have led to the makeup of the modern modal system in the Surinamese creoles. In the early sources, modality seems to have been largely modeled on Dutch (and possibly English). The meanings of early forms correspond to their etyma, though increased contact with Dutch and substrate languages forced developments of Sranantongo and the Maroon creoles in different directions. As sa was losing ground to (g)o as a tense marker in Suriname and acquiring modal meaning, Dutch had a stronger influence on Sranantongo which led to the Sranantongo modal system being modeled on the Dutch one. Modern Sranan sa and Dutch zullen share a similar range of modal meanings (Migge and Winford 2009: 146-148).
There are several reasons to suggest substrate influence on the Maroon creoles' use of sa. Firstly, western Gbe languages have a potential future marker lá / á, which also invokes a range of modal meanings that correspond to potential categories in the Maroon creoles (Migge and Winford 2009: 149). Secondly, the Gbe system of potential modality indexes the same categories as the Maroon creoles and several Gbe varieties (Aja, Xwela, Xwla) formally distinguish positive and negative potential categories, which would explain the use of man (Pamaka) and poi (Ndyuka, < Portuguese pode '3sg can') in negative contexts (Migge and Winford 2009: 150). Essegbey et al. (2013) also point out the formal distinction between inherent and acquired ability, marked with imperfective e and modal sa respectively in the Maroon creoles, which reflects patterns found in several Gbe languages.
The marker kan, on the other hand, appears to have been modeled on the usage patterns of Dutch kunnen, indicating root possibility, ability, and permission in both eighteenth century and modern Sranan (Migge and Winford 2009: 141-142). Kunnen is also used to indicate epistemic possibility in Dutch, and Migge and Winford (2009: 142) suggest that the marginal use of Sranan kan in such contexts is a recent development. Dutch is also clearly the source of mag, indicating permission, though its use is quite marginal in eighteenth century Grammaticalization also played a role in the modal system of Sranan and Pamaka, as evidenced by the status of man as an auxiliary, which developed from its use as a noun (van den Berg 2001: 249-252; van den Berg and Arends 2004: 25-28).
Our data from 2010-11 suggest that the modal system of Ndyuka (and possibly other Maroon creoles) continues to develop. We have noticed differential marking of modal categories between urban and rural dwelling Ndyuka speakers, with the former tending to align themselves with Sranantongo patterns. Likely due to an increasing number of Maroons in the city in the past decades, their increased exposure to Sranantongo, frequent interaction with Maroons from other ethnic groups and non-Maroons, and perhaps the inclination to establish an identity independent of their traditional ethnicity (see Migge 2007; Migge and Léglise 2013; Léglise and Migge 2015), Maroon languages have come under influence of each other and Sranantongo. Ndyuka speakers themselves are also aware of Sranantongo influence on their language. One informant explained that the closer you get to the coast, the more 'developed' the language is. Others describe the influence more defensively; coastal Ndyuka is moksi 'mixed' or basaa 'bastardized'. An urban dwelling informant describes the difference in terms of "modern" Ndyuka along the coast versus a more traditional variety in the interior.
Language attitudes aside, speakers are well aware that there is a difference between urban and rural varieties, though it is difficult for informants to pinpoint particular features, and some informants claimed to switch between varieties depending on context. Tab. 4 illustrates how the modal categories of Urban Ndyuka appear to have been influenced by Sranan. The following examples illustrate the phenomenon with respect to the expression of physical ability. The conventional form for expressing this modal category in rural Ndyuka is the preverbal particle sa, as shown in (21) 2 As noted above, man is also used in other EMC varieties and by some down river and Cottica Ndyuka. It is apparent that our data represent extreme points on a continuum; variation is the norm with respect to these forms and more research is needed to more accurately determine the patterns behind their usage. It is important to note here that the phenomenon presented cannot be described as a complete change; variation is the norm. Many of our urban informants were recorded using both rural Ndyuka forms next to urban forms, though this was not the case with our upriver speakers. With two geographic points of reference in our Ndyuka sample, Paramaribo and the upriver Tapanahoni, the data suggest that Sranantongo is the main source of urban features in Ndyuka; however contact with highly intelligible Eastern Maroon varieties should not be ignored. In fact, it is often difficult to determine the origin of a particular feature, such as the case of man in examples (18)-(20). Pamaka is not represented in our sample, though Migge and Goury (2008: 309) tell us that man is also employed for several modal categories in that language. Kan on the other hand appears to be an addition to the repertoire of urban Ndyuka originating from Sranantongo. While etymologically indeterminate features such as man might weaken our argument for an urban influence on Ndyuka, it should be noted that the important changes in traditional Maroon societies associated with coastal life in Paramaribo and urban centers along the Marowijne that have set the stage for the blurring of traditionally salient differences among Maroon varieties, as well as influence from Sranantongo (Migge and Léglise 2011, 2013).

Development of Aspect in the Surinamese creoles
Verbs that are not marked with an aspect marker are interpreted as perfective. Imperfective aspect is marked with preverbal e in Sranan and Eastern Maroon creole, and ta in Saramaccan. Completive aspect is marked in all creoles with a verb phrase final kaba (<Port. acabar 'to finish') which is homophonous to a main verb 'to finish'. Aspectual categories and their markers are detailed in Tab. 5. Winford and Migge (2007: 83) argue that the perfective interpretation of unmarked verbs is modeled primarily on the Gbe languages. The two language groups share a "more or less identical range of meanings and uses" with respect to unmarked verbs, including: property items and other stative verbs, non stative verbs with past reference, and non stative verbs with current relevance (Winford and Migge 2007: 81). Substrate influence also plays the primary role in the development of the completive marker. Like the Surinamese creoles, Gbe languages also have a completive category, conveyed with a verb phrase final marker derived from the verb 'to finish' (Winford 2006: 102;Winford and Migge 2007: 84-85). However, kaba is compatible with stative and non-stative situations, while the Gbe marker is only compatible with non-stative situations, suggesting some additional processes of grammaticalization in the Surinamese creoles (Winford and Migge 2007: 85; van den Berg and Aboh 2013).
Imperfective markers e and ta are derived from the locational copula de and the verb tan 'to stay, to wait', respectively. In the early texts these forms are used to mark progressive aspect, but only rarely habitual aspect, according to Winford and Migge (2006: 85), evidence which they use to suggest that imperfective aspect was a late categorical development dependent on the further grammaticalization of the progressive marker to an imperfective marker. However, van den Berg (2007: 200) states that de in Early Sranan covers several imperfective sub-categories: continuous, habitual, progressive, and ingressive. This suggests that, contrary to Winford and Migge's (2007) claim, imperfective was already a grammaticalized category early on, and was not modeled on the Gbe aspectual system which lacks a macro imperfective category as in the Surinamese creoles.

Discussion
A number of different structures in various substrate and superstrate languages, along with innovation and grammaticalization are responsible for the composition of the Surinamese creole TMA systems as they are today. Tab. 6 above summarizes the processes involved in the develop-ment of each TMA marker. Various aspects of all the Surinamese creole languages' TMA systems can be attributed to substrate influences, though this is more apparent in the Maroon creoles. English superstrate influence, contact with Dutch, and grammaticalization have also played a prominent role in the development of Surinamese creoles. Finally, we suggest that the creole languages are increasingly influencing each other's development, as indexed by urban Ndyuka modal marking, which patterns with Sranan in our data.

Surinamese Dutch
Surinamese Dutch (SD) is spoken both in Suriname and in the Netherlands, the colonizing country to which many Surinamese have migrated. It is a widely recognized ethnolect in the Netherlands (cf. Muysken 2013), and some of its features have led to ethnic stereotypes. It has also been described on a number of occasions, in part under the rubric of 'mistakes' of Surinamese children in the Dutch classroom. Charry (1983) is still the most sophisticated study focusing on phonological variation in this variety in the Dutch context, which requires much more investigation. De Kleine (2007) is an extensive morphosyntactic study of SD as spoken in urban Suriname. In Suriname, paradoxically, SD is not an ethnolect but an ethnically neutral national variety.
It should be noted that many of our elicitations in Suriname did not produce very informal speech. Some of the speakers recorded felt that they had to put on their best Dutch, i.e. as close to European Dutch (ED) as possible. In spite of this, the data reveal a surprising number of innovative features. Another issue is whether the SD recorded represents a stable variety in its own right or simply a gathering of second language speech samples. To some extent it is the latter, as some of the consultants recorded are clearly second language speakers. However, the fact that there are 15 analyzed samples makes it possible to see how wide Tab. 6: Developments of TMA in Surinamese creoles. spread a feature is across samples and how frequent within a sample. Some innovative properties are quite general, as can be seen from Tab. 8 in section 4.3 below, suggesting that they are entrenched within the SD speech community.

Category Marker Process
In 4.1 we present the basic outlines of the ED system, 4.2 contains the actual sketch of TMA in SD, and in 4.3, a more general perspective is introduced.

The TMA system of European Dutch
The TMA system of ED is not very rich in fully grammaticalized categories, but there are numerous auxiliaries and semi-auxiliaries, and some specialized constructions. The basic distinction in verbal morphology is that between past and non-past. In (24) an example is given with a regular (weak) verb and in (25)  The shape of the basic roots involved in these examples is the same; the status of a verb as 'weak' or 'strong' is not phonologically conditioned.
The basic temporal distinction being past/non-past, future tense reference is ordinarily marked with a simple non-past: (26) Morgen koop ik een fiets. tomorrow buy.1sg I a bike 'Tomorrow I will buy a bike.' There is a specialized construction to mark progressive aspect, aan het X-inf zijn 'be at X-ing': There is also a 'have' + past participle perfect, which is often used in ordinary past tense contexts, unlike its use in English: (28) Zij heeft gisteren schoen-en gekocht. she has yesterday shoe-pl buy.pp 'She bought shoes yesterday.' Furthermore, there is a whole range of auxiliaries and semi-auxiliaries. In Tab. 7, we present some of them with their main uses.

A sketch of TMA in Surinamese Dutch recordings
The SD data show a number of innovative features concerning the expression of TMA categories when compared to ED.
In the general area of Tense, there are some cases (in two samples) where past marking on a verb is innovative in comparison with ED. These include double marking (strong + weak past, as in (29)), and weak instead of strong marking, as in (30). The irregular past tense marking may be linked to a more general issue concerning tense organization in SD. De Kleine (2007) contains a detailed analysis of the TMA categories in SD as compared to ED. We will only mention a few of the points she makes. Regarding past tense marking, a very complex picture is given. De Kleine comments (2007: 69): 'The pattern that emerges from the data shows that the rules for past tense marking in SD, unlike ED, are governed by discourse rather than grammar.' And further on (2007: 75): 'It should be emphasized that there is a significant amount of variation regarding past tense marking in the data.' In the Frog Story recounting in (31) the speaker jumps from: [perfect] to [present]  This suggests that, contrary to ED, there is a tendency to view the auxiliary and the verb as a single cluster.
The overgeneralization of the aan het progressive construction is illustrated in (37), where the stative verb slapen 'sleep' is marked with aan het.
(37) Een muis is aan het slap-en. a mouse is at det sleep-inf 'A mouse is sleeping.' With other speakers we find overgeneralization of bezig zijn te X 'be busy to X' with the same meaning: In both cases the precise semantics of the predicate and the nature of the semigrammaticalized auxiliary are treated differently in SD from ED.
A feature shared with some varieties of ED is the use of generic doen 'do':

Conclusions and discussion
Tab. 8 gives an overview of our main findings. First the number of samples is mentioned, then whether the feature is also indicated in the earlier studies of Essed-Fruin (1983), de Kleine (2007), and de Bies (2008). Finally we indicate whether the feature is similar to a Sranan feature (+), or different from the relevant property of Sranan (−). If the feature does not correspond directly to Sranan, but has emerged through indirect Sranan influence, it is (+).
We can assume indirect Sranantongo influence in the case of the absence of verb final because Sranantongo has S Aux (=gaan/worden) V O constituent order. The overgeneralization of aan het V and bezig zijn could be linked to the general presence of progressive e in Sranan with activity verbs.

Variable in our corpus # Samples Essed-Fruin De Kleine De Bies Sranan
Absence of verb final after gaan and worden Essed-Fruin (1983: 122-137), in a very interesting early exploratory study, distinguishes three types of 'deviations' from ED within SD (see also Muysken, this volume): -Idiomatic deviations -Deviations from the conventional system -Deviations from the essential system The changes in word order and the irregular marking of past tense morphology reflect deviations from the conventional system, while changes in progressive aspect marking and the relation between past and present tense probably reflect more deep-seated deviations from the essential system.
De Kleine (2007)  The first case, transfer of grammatical functions, is the most frequent. A particular element already existing in ED, such as the use of aan het, is given a wider semantic range. This type of transfer leads to Syntactic camouflage (2007: 136), the term De Kleine cites from Spears (1982): an SD form which looks like an ED form has a different meaning.

Sarnami
Sarnami is a koiné resulting from the mixing of various northern Indian languages, among them (varieties of) Bhojpuri, Magahi, Maithili and Awadhi spoken in the present-day Indian federal states of Uttar Pradesh, Bihar, Jharkhand and West Bengal. Our data and sociolinguistic interviews show that Sarnami is used by all generations in the Indo-Surinamese community of Suriname in a pattern of trilingualism involving Sarnami, Sranantongo and Dutch. Entrenched multilingualism in the Indo-Surinamese community has led to contact induced change in Sarnami. The most obvious changes are lexical in nature: Sarnami has acquired numerous lexical items from Sranantongo and Dutch, and also features calques from these two languages (cf. Yakpo and Muysken, 2014). There are also contact induced changes in the grammar of the language, e.g. in constituent order (Yakpo, this volume). In the following, we document contact induced changes in the TMA system of Sarnami. The core TMA system is constituted by markers which instantiate central tense, aspect and mood categories. We will see that the most profound changes can be witnessed in the non-core TMA system. This is to be expected since the non-core system is by definition paradigmatically and syntagmatically less tightly organized and expresses more specialized (hence more peripheral) semantic notions. Given the medium to short time depth of contact between many of the languages of Suriname -this concerns first and foremost the immigrant languages of the Asian-descended communities -we would expect the impact of language contact to make itself felt first in the non-core system.

The TMA system of Sarnami and its contributing languages
Sarnami has a TMA system that is characteristic for the Indo-Aryan languages of northern India. The core TMA system makes use of bound morphology, i.e. verbal suffixes, with support from the auxiliary verb 'to be' for the expression of composite tense/aspect notions.
The template for the formation of composite tense/aspect categories is virtually identical in all languages. For example in (42), past progressive is instantiated in a construction in which the lexical verb (soc-'think') is marked for imperfective aspect by suffixation (-at) in order to express an ongoing situation. Simultaneously, a form of the verb 'to be' (rah-) is inflected for past tense and person-number, cf. (42), (43). The 'to be' auxiliary is, in turn, marked for past tense (-il In the non-core system, auxiliaries combine with lexical verbs in order to express less central aspectual and modal notions in auxiliary constructions. In all the languages including Sarnami, the group of auxiliaries encompasses items ranging from little grammaticalized lexical verbs to highly grammaticalized 'vector verbs ' (cf. Masica 1993: 266). Although an etymological relation between vector verbs and lexical verbs can in many cases be established, the former may express specialized aspectual and spatial notions with only remote semantic connections with their lexical counterparts. One such example is the verb lag-, whose source meaning is '(to) attach', as shown in the following example from Sadani Bhojpuri: (45) kona mẽ rʌtʌn kʌr gheir lagʌl ahe.
'In the corner a mass of jewels is heaped up (lit. 'is attached').' (Sadani Bhojpuri;Jordan-Horstmann 1969: 86, no gloss provided) In Sarnami and its contributing languages, lag-also occurs as an auxiliary verb in a dative experiencer construction. In its most abstract and grammaticalized function, lag-is employed as an aspectual auxiliary verb with the meaning '(to) begin'. In its auxiliary function, lag-is inflected like any other full lexical verb. Compare the following two examples from Sadani Bhojpuri and Sarnami respectively: In the following section, we concentrate our analysis on contact-induced developments in the non-core TMA system of Sarnami.

Contact-induced changes in the TMA system of Sarnami
In the non-core TMA system of Sarnami, auxiliary verbs combine with lexical verbs in order to express less central aspectual and modal notions in auxiliary constructions. Tab. 9 summarizes contact-induced developments in the non-core system of Sarnami, which will be discussed in further detail below.

Continuative
In our Sarnami corpus, continuative aspect is exclusively expressed via a construction involving non-native material. The adverb/particle doro 'through', of Sranan origin, may appear in combination with the Sarnami verb já/ga-'go' to express the notion 'go on, continue', as shown in (50) below. The construction therefore features a mixture of matter and pattern borrowing.
(50) en doro ga-il bajá-we. and cont go-pst play.music-inf 'And he continued playing music.' The element doro is also found to express a continuative reading on its own without addition of the verb já. An example for the use of this adverbial strategy can be found in (51).
(51) doro bajá-we hai. cont play.music-inf be.prs '(He) continues playing music.' The continuative construction in Sarnami exemplifies the complex multidirectional transfer processes that characterize the Surinamese linguistic area. For one part, the continuative construction is a direct calque from a corresponding Tab. 9: Developments in the non-core TMA system of Sarnami. construction in Sranan. In Sranantongo too, an adverbial element doro as well as a lexical verb go collocate to express a continuative reading, as can be seen in example (52). The syntactic differences between Sarnami and Sranantongo can be attributed to the corresponding differences in clause-linking strategies available to the two languages. In Sarnami (cf. (50) above) the lexical verb is an infinitival complement to the auxiliary construction. In Sranantongo (cf. (52) below) we instead find a serial verb construction in which the aspectual 'auxiliary' follows the lexical verb without any overt sign of subordination:

Non-native structure (Literal) translation Native equivalent (Literal) translation
(52) a bigi wan plèy go doro nanga a bal. det big one play go cont with det ball 'The big one continues playing with the ball.' In Sranan as well, the adverbial doro alone can express a continuative notion, as shown in (53). (53) a moysmoysi e leysi en buku doro. det mouse ipfv read 3sg book cont 'The mouse continues reading its book.' The continuative construction is however not native to Sranantongo either. It is a nativized borrowing, and originally a calque from the Dutch particle-verb collocation door-gaan, lit. 'through-go', best translated as 'continue'. Syntactically, the Dutch construction differs from the corresponding Sarnami and Sranan ones. In Dutch, the lexical verb specified by the continuative auxiliary is expressed as a clausal adjunct: (54) en hij gaat door met spel-en. and he goes through with play-inf 'And he continues playing.' Unsurprisingly, Dutch also features the use of door alone in constructions such as (55), in which the adverbial alone expresses a continuative notion.
(55) hij gooit dan die drum weg om dan rustig weer he throws then that drum away to then calm again door te lezen. through to read 'He then throws that drum away in order to then calmly continue reading.
In Dutch, as in other Germanic languages including English, there are scores of such complex predicates in which an adverb or particle collocates with a verb to render a large range of spatial, Aktionsart and idiomatic meanings (cf. e.g. Müller 2002). In contrast to Dutch, Sranantongo has but a handful of complex predicates of this kind, all of which are more or less nativized calques from Dutch. Sarnami does not have a native layer of verb-particle complex predicates either. We therefore interpret the presence of the continuative construction in Sranantongo and Sarnami as a transfer from Dutch. It is most likely that the Sarnami equivalent of this construction entered the language via Sranantongo. The evidence is the phonological shape of doro, which features a paragogic final vowel. Nevertheless, the function of the construction in Sarnami, and the distributional aspects described above certainly owe just as much to transfer from Dutch as from Sranan.

Ingressive
The Sranan verb bigin 'begin (to)' is more frequent as a marker of ingressive aspect than an equivalent native construction. This frequency is fairly evenly distributed among speakers, however with a clear preference for the non-native structure by speakers under 25 years, where the non-native construction figures in over 80% of all cases.
Example (56) shows how Sarnami speakers make use of bigin for the expression of ingressive aspect, which we define here as the entry into the situation described by the main verb. Taking a closer look at the construction, we remark the presence of the verb kare 'do'.

(56) aur bigin kar-il
o-ke kát-e ke. and begin do-pst dist-acc/dat cut-inf acc/dat 'And (she) has begun to cut it.' With the help of the generic verb kar-, non-native verbs and nouns may be integrated into a Sarnami clause. The use of generic verbs in such constructions has been well documented for Indo-Aryan languages (cf. e.g. McGregor 1995: 63). The extensive presence of (bilingual) compound verbs in the language family makes up for the scarcity of verb-deriving morphology. The non-native ingressive construction parallels a native one, see (57). This construction also involves the generic verb kar-, preceded by the (native) noun suru 'beginning': (57) tab suru kar-is eerste aurat-iyá doorstuur then beginning do-pst first woman-def pass.on kar-e ke. do-inf acc/dat 'Then the first woman began to pass (it) on, (and) the thing that was in the person's hand.' The comparison between the non-native ingressive construction in (56) and the native one in (57) however also shows that the former construction features a verb where the latter has a noun. In fact, we do not usually find native verbs in the (object) position before the generic verb kare. This is therefore a specific adaptation mechanism for loan verbs in Sarnami (cf. Appel and Muysken, 1987: 126-127;Muysken, 2000: 197-202, 208-211).
Another common way of expressing an ingressive notion in Sarnami is very different from the two above. It involves the use of the highly grammaticalized, largely desemanticized "vector verb" lag-, lit. 'touch, come in contact with ', cf. (58). The auxiliary is postposed to the verb it specifies and is found in a position in which other vector verbs in Sarnami and other Indic languages are found (see section 5.1 above).

Completive
The regular way of expressing a completive aspectual reading is via a native construction. Speakers may however opt for other, non-native means as well. The native construction involves the use of the vector verb cuk-, probably derived from the lexical verb cuke '(to) lack'. Although cuk-has no lexical meaning of its own, it is inflected like any other full verb. At the same time, the preceding lexical verb appears in the non-finite conjunctive participial form, which is only overtly expressed via the suffix -i in vowel-final verb stems. Compare examples (59) and (60).
(59) káṯ cuk-al dui pisi meṉ aur dhar de-il. cut compl-pst two piece loc:in and put 'give'-pst '(She) has finished cutting it into two pieces and has put it (down).' (60) ondertussen sab aurat-iya-n sab hánth meanwhile all woman-def-pl all hand milá-i cuk-al (…) join-conj.ptcp compl-pst 'Meanwhile all the women have finished shaking all hands (…)' There is however another non-native rendering of completive aspect in the data. This possibility is, however, not fully productive, since it is lexically restricted to calques of Dutch complex verbs that incorporate the terminative Aktionsart affix af-, lit. 'off'. The following example shows how the affix may be ingeniously pressed into its (lexical) aspectual function. The example features the Sarnami verb banáwe '(to) make' preceded by the Dutch terminative affix. The entire predicate is a lexically mixed calque of the Dutch predicate af-maken '(to) finish (off)': (61) en voor de rest ham apan oto kin-t-i and for the rest 1sg refl.poss car buy-ipfv-1sg ekád-go apan ghar af-bana-it-i some-clf refl.poss house term-make-ipfv-1sg 'And as for the rest, I'd buy my car, (with) some, I'd finish my house.'

Conative
Conative modality, hence the rendering of the equivalent of 'try (to)' in English is normally expressed via the use of a non-native verb in Sarnami. In all but one case recorded in the data, speakers opt for the (Dutch-derived) Sranan verb pruberi or the Dutch equivalent probeer. As in other cases involving Dutch verbs, speakers use an invariant verb form, namely the 3sg present tense form, when integrating the Dutch verb (rather than another form, e.g. the infinitive proberen). Here too, speakers make use of a bilingual compound verb featuring the generic verb kare. Compare the following example with Dutch probeer: (62) olifant-wá probeer kar-e stop kar-e geluid elephant-def try do-inf stop do-inf noise bana-i kar-ke, ma mus-wá ke make-conj.ptcp do-acc/dat but mouse-def acc/dat ná hinder ho-we hai. neg prevent be(come)-inf be.prs 'The elefant tries to stop (the mouse) by making noise, but the mouse is not prevented [from reading].' The Sranan verb pruberi appears in the same kind of construction, as shown in (63) below.
(63) pruberi kar-e hai uppar big-e ke try do-inf be.prs upperside throw-inf acc/dat maar punah se gir ga-il but again abl fall go-pst 'It tries to throw it up, but it has fallen again.' An equivalent native way of expressing conative modality involves the usual kare light verb construction. However, it features a nominal complement (here kosis 'effort') rather than a verbal one as is the case in the two preceding examples involving non-native elements. In altogether thirty instances of conative modality, only two involve the use of the native structure shown in (64).
(64) aurat-iya-n kosis kar-e hai mardan-wa-n woman-def-pl effort do-inf be.prs man-def-pl ke chu-we ke, pakaṟ-e ke acc/dat touch-inf acc/dat hold-inf who 'The women are trying to touch, to grab the men.'

Conclusion
Sarnami shows significant contact-induced developments in its non-core TMA system: Some important aspectual and modal notions are primarily expressed via constructions that contain elements borrowed from Sranantongo and Dutch. The status of Sarnami as an independent Indic variety is therefore not only confirmed through the innovations it has acquired in the process of koineization (see Yakpo, this volume). Sarnami also stands out in the degree to which the language makes use of Sranantongo and Dutch items in its lexicon and grammar. Further research will have to show how much paradigmatically more tightly organized parts of the grammar, including the core TMA system also show signs of transfer from Sranantongo and Dutch.
while Javanese is regarded more as a language spoken with (grand)parents, in restricted contexts.
This section will provide an explorative overview of TMA marking in the Javanese language as it is spoken in Suriname. By comparing Surinamese Javanese material (the heritage language) with Javanese material from Java (the 'baseline' language), changes which are possibly due to language contact will be identified, and when possible, an explanation will be given for how these changes could have come about.
Previous studies have already shown some divergence between Surinamese Javanese and Javanese as spoken on Java, e.g. in the use of speech styles (see below), which are less differentiated in Suriname than on Java (Wolfowitz 1991). It is important to be aware of the fact that Javanese as spoken on Java is not homogeneous and that it has traditionally been divided into three dialects: western, central and eastern Javanese. The dialect as spoken in Surakarta and Yogyakarta (central Java) has been generally accepted as Standard Javanese (Dudas 1976: iv). Available information suggests, however, that dialects differ mostly in phonology, rather than in morphosyntax or semantics of TMA.
Another notable characteristic of Javanese are the so-called speech styles, which are strictly observed among baseline Javanese speakers (Ras 1985). These range from informal to formal, with the most informal speech level being ngoko, used with friends and relatives. On the other side of the spectrum is the formal speech level krama, used in dialogues with highly placed individuals and strangers. In between these two speech levels there are even more fine-grained differentiations, depending on the position of the interlocutors. However, these speech styles only differ in lexicon (including some grammatical affixes), but not in syntax or morphology, and since ngoko is the most widely used style (virtually the only style used in Suriname), examples presented here will be taken from ngoko speech.

TMA in Java
Javanese verbs are not marked for person or number and the Javanese language is not rich in morphosyntactic marking for TMA-categories. According to Robson (1992: 64), most TMA-categories in Javanese are marked by auxiliary words, which occur in pre-verbal position. One of his arguments for considering these as auxiliary words, and not as clitics for example, is the fact that they can be separated from the verb, for example by a negative adverb as in the following example which involves the auxiliary bakal and the verb lungo 'go': (65) aku bakal ora lungo 4 1sg irr neg go 'I will not be going.' (Robson, 1992: 66) 4 In standard Javanese orthography used here for the most part, this should be written 'lunga', but Robson writes lungo.
Tab. 10: Overview of TMA-markers (auxiliary words and suffixes) in Javanese. Tab. 10 provides an overview of TMA-marking words and suffixes on the basis of different sources (Robson 1992, Adelaar 2011, Vander Klok 2008, Vander Klok 2010. It should be mentioned that the status of some of the markers is not agreed upon. The marker wis for example, is considered a tense-marker instead of an aspect-marker by Vander Klok (2008). A similar example is bakal, which might be a tense marker comparable to arep according to Vander Klok (2010: 2). These points will be elaborated below.

Tense
Javanese does not have a rich set of forms for tense marking. Tense is often inferred from the context in zero-marked phrases, e.g. with the use of an adverb of time as in (67). Without further morphemes or adverbs that give information on the time of the event, it is usually interpreted as present, as in (66). Future tense is marked with arep (68) which is also regularly used in the meaning of 'to want' conveying participant-oriented modality (volitional modality) rather than tense. Modal use of arep will be discussed in the following subsection.
(68) aku arep ng-oreng sego5 1sg fut av-fry rice 'I will fry the rice.' (Vander Klok 2010: 1) As Hengeveld (2011: 592) argues, volitional modal markers are indeed a potential source for (absolute future) tense markers, and it is therefore highly likely that the modal meaning of arep was the original meaning, and that this tense marker developed only later.

Modality
As for the Javanese modals, according to Vander Klok (2008), these are organized along two axes: quantificational force (either universal -'must' or existential -'may') and type of modal base (either epistemic or deontic). The combination of these two axes results in four modal categories, which are each marked by a different auxiliary. The classification that follows is shown in Tab. 11.
Apart from these four modal markers described by Vander Klok, we distinguish two more markers of modality in this paper, namely bakal (irrealis) and isa ('can').
5 In standard orthography this should be sega. The marker isa can be translated as 'can/be able to' and refers to the acquired (physical) ability of the participant to engage in an event: (69) di-gawani kayu sing luwih gedé, nah saiki isa, uv-bring wood rel more big intj now can dèké njukuk kain abang iku 3sg take cloth red that 'A bigger piece of wood is brought, now he can, he takes the red piece of clothing.' In fact, this marker is often used to describe the participant's inability, combined with the negative adverb gak: (70) ana arèk loro teka n-yobak n-jupuk klambi iku exist child two from av-try av-take cloth that mencolot~mencolot lugur gak isa kayaka-né jump~red fall neg can seem-def 'There are two children who try to take the piece of clothing, they jump, they fall, it seems they can't.' Classification of the marker bakal is not straightforward. Vander Klok (2010: 2) considers bakal to be a tense marker, comparable to the future marker arep, with a difference in the expression of agency.
Whereas arep expresses intention of the speaker and therefore implies agency, she considers bakal to convey only a prediction by the speaker, without any implication of agency. We propose that bakal is a marker of irrealis modality rather than tense. Consider the following examples of the use of bakal in contrast to ate (the East-Javanese variant of arep): (71) deke ng-angep nèk wong iku ate lunga 3sg av-assume comp person that fut go 'He assumes that she will leave.' (72) deke ora yakin nèk wong iku bakal lunga 3sg neg sure comp person that irr go 'He doubts whether she will leave.' In these examples, the difference between ate/arep and bakal seems to be the degree of certainty the speaker has about the truth of the subordinate clause. When the speaker was asked to explain the difference between ate and bakal, it was confirmed that this was indeed the difference: when contrasting different sentences with ate and bakal, the speaker declared that in the latter case she felt as if it was less certain that the event in the subordinate clause were truly to take place. Since this marker expresses the speaker's commitment to the truth value of the proposition, bakal should be classified as proposition-oriented modality. As mentioned above, there is still some discussion about the status of the marker wis: in Vander Klok (2008) it is considered a marker of tense rather than aspect. Here we will analyze wis as a marker of perfect aspect, since the defining element of this marker appears to be the emphasis on the result of the verbal action, and the relevance thereof in the present. Comrie (1976: 12) considers this present relevance of the past situation one of the defining characteristics of perfect aspect. Consider the following example, one of the most prototypical uses of wis, which is used in response to the offering of food (politely refusing) (82) wis wareg PRF full 'I am already full.' (so I don't need to eat anymore)

TMA in Surinamese Javanese
In this section, we will explore the TMA-system of Surinamese Javanese, and investigate possible changes it has undergone in comparison to the TMA-system of Standard Javanese.

Tense
The future tense is marked by arep as in the baseline variety. The following metalinguistic comment from one of the heritage language speakers illustrates the way the speaker understands the usage of this marker: (83) als we zeggen, ik ga naar de bank: arep nèng bank if we say 1sg fut loc the bank fut loc bank 'If we say, I go to the bank, "arep nèng bank."' It appears that arep also conveys aspectual meanings in the heritage variety. Consider the following examples, which are descriptions of video clips. In the video described in example (84), a man washes his hands, but the event begins after the beginning of the video.
(84) wong lanang arep wisuh tangan-é karo banyu, person male fut wash hand-def with water di-lapi tangan-é karo anduk uv-wipe hand-def with towel 'A man is going to wash his hand with water, the hand is wiped with a towel.' The same goes for the video as described in example (85), which starts with the image of a woman standing, after which the woman starts moving out through the window.
(85) wong wèdok arep metu tekâ jendélâ person woman fut go.out from window 'A woman is going to go out through the window.' In the context of the videos, we propose that the most natural translation for this marker here would be 'is going to', since it refers to a more immediate future, which actually starts happening during the time of the utterance. Therefore, this is not truly a future tense marker, but more of an aspectual marker, specifically prospective aspect.
Speakers of the heritage variety appear to mark future/prospective with arep more frequently and in contexts where the baseline speakers employ other strategies. Although the use of arep as a prospective is not an innovation in Surinamese Javanese (as confirmed by Indonesian speakers), it is not encountered with this meaning in the baseline corpus. The preference to use arep in these contexts appears to be a feature of the heritage variety. The apparent overgeneralization of arep could be due to influence of (Surinamese) Dutch and/or Sranan, which both categorically and formally differentiate immediate versus more uncertain future: gaan 'go' versus zullen 'shall' in Dutch, and sa and o in Sranantongo. Some caution is needed here, however, since the frequent use of arep may also be the result of the way the data were gathered in Suriname. We leave this for further research. In Suriname, speakers described the videos more simultaneously as they were viewing them, while the Indonesian speakers more often appeared to wait until after the video had finished.
Similar to the baseline speakers, Surinamese speakers do not mark verbs for tense, which is rather inferred from the context or the use of temporal adverbs: However, since this type of construction occurs only once in the heritage corpus, it appears that this is simply a case of code-switching (considering it is preceded by the Dutch pronoun ze).

Modality
The heritage variety uses a different form, inter, than the baseline variety to mark acquired (physical) ability. The form inter is not a loan from Dutch or Sranan, but originates from the Javanese word pinter, which literally means 'clever, skilled' and which expresses acquired ability, as in the utterance anakku pinter maca 'my child can / knows how to read' (lit. 'my child is clever in reading'). In the Surinamese corpus, the form inter (where the loss of the initial has arguably been caused by analogy with the alternation between bisa and isa in Indonesian Javanese), subsequently seems to have undergone extension from a marker of participant-oriented (acquired) ability towards a marker of event-oriented (participant-external) ability, as in (89), a PAT change.
(89) njukuk planga eindelijk inter n-jukuk kaos-é take board finally can av-take shirt-def 'He takes a board, and can finally take the shirt.' It is often used in combination with the negative adverb ora: (90) terus arep di-jukuk maar ora inter then fut uv-take but neg can 'Then it is going to be taken but he cannot.' Although the form inter seems to have completely replaced the form isa from baseline Javanese, there are no differences in its syntactic realization. Further research is needed to shed light as to how and when this semantic extension came about, and to what extent language contact played a role in this development.
Irrealis is expressed infrequently in both the heritage and baseline corpora by bakal.
(91) naar het schijnt sing bakal pâdâ~pâdâ sing kâyâ to it seems rel irr same~red rel similar.to volgende weekend next weekend 'As it seems, it will be the same, which is similar to next weekend.' In this context, bakal is used to express a more uncertain future, since it is a complement of the verb schijnt 'seems'. In the following example, bakal is used with the determiner suffix -e, as a sort of nominalizing procedure, but still expressing the irrealis category: Of the epistemic modals, mungkin and mesti, the first does not occur in either corpora. The other epistemic, mesti, occurs just three times, of which only one example appears in the heritage language. In the example, mesti combined with the determiner suffix e, where it can be translated as 'perhaps' (Vruggink 2001), and thus seems not to express epistemic modality: (97) aku durung ng-erti mesti-né dolan~dolan karo 1sg not.yet av-know certainly-def play~red with kanca-ku dolan~dolan [inaudible] friend-1sg.poss play~red *** 'I do not yet know if I perhaps play with my friend or play ***.' One certain contact-induced development in the category of modality is the borrowing of modal verbs from Sranantongo. In this case, the Sranan verb pruberi or proberi 'to try' (< Dutch prober) occurs in position of the native Javanese verb jajal 'to try'. The word order of the construction remains the same (auxiliary + verb/ complement) in the baseline language (98) and the heritage variety (99) It is difficult to establish whether these phrases should truly be interpreted as progressive constructions, or rather be considered presentational constructions with a relative clause without a relative pronoun. In the latter interpretation, examples in (102) could be translated as follows: 'There is a person who writes a letter' and 'There is a person who cuts an orange'. This last possibility might be supported by the fact that when the existential does co-occur with the relative pronoun sing, this is only used to in some sense delimit the subject of the clause, as in the following example: (103) ènèk wong lanang loro, sing siji ng-ekèk-i tas exist person male two rel one av-give-appl bag karo liya-né with other-def 'There are two men, of which one gives a bag to the other.' However, as argued by Hengeveld (1992: 265), the existential construction can indeed very well have a progressive interpretation. Hengeveld argues that this type of clause should be viewed as a circumstantial adverbial clause, with the translation of 'There is a person in the circumstance of [verb]ing', which does indeed entail a progressive interpretation of the verb. The quantitative comparisons are unconvincing in arguing for contact-induced developments. Although the progressive marker lagi is not attested in the heritage corpus, it only occurs four times in the baseline corpus. And while the existential construction is used in heritage contexts where baseline speakers use lagi, the overall relative frequency of existential constructions is not higher in the heritage corpus. However, it does not seem coincidental that imperfective e (which also indicates progressive, among other aspectual meanings) in the Surinamese creoles has grammaticalized from a copula which is also used in existential constructions (Winford and Migge 2007: 89).
As in the baseline language, perfect aspect is expressed with the auxiliary wis. To emphasize this focus on the relevance of the event in the present, this marker is often combined with saiki 'now' in the same sentence: (104) saiki-né bocahé wis temu, temu kanca-né wis ungah now-def child-def prf find find friend-def prf happy menèh again 'Now the child had found his friend, he was happy again.' The meaning of wis remains unchanged in the heritage corpus, though it is used more frequently in the baseline language than in the heritage language, in fact even more than twice as often. This may be caused by still lower frequencies of Sranan's completive marker kaba.

Conclusions
To sketch the picture of the TMA-system in Surinamese Javanese in a unified way, we see that there are different things going on in the different categories. The use of arep is more frequent among the heritage speakers, mirroring patterns in Sranan and Dutch. The category of modality seems to have undergone the most changes, although they are still not radical: the forms for the modal verbs 'try' and 'be able to' have been replaced by new forms in Surinamese Javanese, while the syntactic construction had remained the same. The verb 'try', is most clearly the result of language contact, since proberi is a Sranan verb. Within the category of aspect, the original Javanese marker lagi for progressive aspect seems to be less used, in favor of the construction with the existential verb, also perhaps modeled on similar developments in Sranantongo. The perfect marker wis also seems to be less used in the heritage corpus which is more consistent with perfective marking in Sranantongo.

Discussion: Stability and borrowability in Surinamese TMA systems
In this chapter, we have detailed a number of developments in the TMA systems of the Surinamese creole languages, Surinamese Dutch, Sarnami, and Surinamese Javanese. These are summarized in Tab. 12 along with processes relevant to their transformation. Our exploration of TMA systems in these languages clearly demonstrates the central position of Sranantongo and Dutch in contact induced language changes in Suriname. These languages not only exchange linguistic forms and patterns, but also provide them in unidirectional transfer to the other languages in our sample. This speaks to the role of factors that are external to the linguistic system in language contact. Had such factors not played a role, we  would expect a more symmetrical flow of linguistic features across languages. However, the central position of Sranantongo and Dutch in Surinamese society, and that they are the two languages that have traditionally functioned as out group languages, means that they both tend to be contributors of linguistic material rather than recipients. 7 Clearly linguistic structures also play a role, albeit not a very clear one, in the types of changes attested in our corpus. Consider that the adoption of Sranantongo forms in the expression of conative modality in Sarnami and Javanese parallel native forms and are therefore relatively easily incorporated into native structures. Similarly, recent developments in formal marking of urban Ndyuka potential modality do not constitute structural alterations since Ndyuka and Sranantongo index the semantic distinctions within this realm of modality. On the other hand, the core TMA system of Sarnami remains relatively unaffected, which we suspect is the result of its integration into the language's rich morphological system and/ or typological distance from Sranantongo and Dutch. Still, this line of thinking does not account for everything we have described. Consider the af-constructions in Sarnami in which neither a foreign form is introduced to an existing parallel native structure, nor is the construction introduced to fill a gap in the native system.
Returning to the questions of borrowability and stability in TMA systems -Matras (2007: 45-46, see example (3a) above) suggests that modality is most borrowable, followed by aspect, future tense, then other tenses. Taking borrowability as converse of stability, Matras' general hierarchy holds for most of our data. Most developments can be seen in the modal systems of our language sample. The exception to this is Dutch, but since a large portion of Sranantongo's modal system is modeled on Dutch, there's no suitable source by which Dutch could be influenced. Alternatively, our methods of data collection may have simply not elicited the right kind of language use to make Surinamese Dutch modality visible. The aspectual systems in the Surinamese creoles appear to have stabilized rather early on in their development, but the other languages in our sample all show developments in their aspectual systems. Future tense displays the most developments in Dutch and Javanese. There appear to be no direct developments in the realm of past tense, though the reorganization of SD certainly effects when past tense is marked. In fact 7 This is not to say that in the individual Dutch or Sranantongo ideolects of speakers of other in group languages in Suriname, no trace of their L1 is detectable, or even that ethnolectal varieties of Surinamese Dutch and Sranantongo do not exist in Suriname (see Lie 1983). It is the overwhelming tendency in this and other studies that Surinamese languages other than Sranantongo and Dutch do not affect Sranantongo and Dutch in Suriname.
Bereitgestellt von | De Gruyter / TCS Angemeldet Heruntergeladen am | 16.10.19 13:36 past tense seems to be so prevalent that it gets marked twice in the case of some strong verbs in SD.
Even while adhering to the general trend for TMA borrowability proposed by Matras (2007), the relatively limited changes in Surinamese Dutch modality (we only found the tendency for a distinction between two kinds of future, modeled on the Sranan distinction between sa and o) preceding changes in aspect and tense, along with inconsistencies in the hierarchies in those studies detailed by Dediu and Cysouw (2013), (examples (4)-(9) above) suggest that language change, at least within TMA, is (a) conditioned by factors external to the language system, and / or that (b) there are no universally stable TMA structures (despite statistical tendencies).
Our investigation has been largely exploratory, and although we have been able to pinpoint a fair number of changes in the TMA systems of Suriname, there is ample room for continued research on the topic. Firstly, developments among urban dwelling Maroons need to be more thoroughly and systematically investigated in terms of leveling of Maroon varieties in the city. Are we witnessing a 'real' change toward Sranan -remember that the forms are Sranantongo but there is no categorical shift and the formal distinction between positive and negative modality appears to be maintained -or something else? Another issue that merits further research is the historical development of SD. Overall, the influence of ED has become stronger when the migrations to the Netherlands took on a massive form in the 1970's. We can expect earlier forms of SD to show more semantic features that distinguish it from ED. An analysis of the core TMA system of Sarnami would be welcome. And finally, the Javanese section in this paper is perhaps the most exploratory in nature and thus provides a number of avenues to be investigated in more detail: forms used, frequencies, and the particulars of their semantics.