The Linguistic Review 2016; 33(3): 365–396 Peter Slomanson* Pragmatic accommodation as a catalyst for the development of (non-)finiteness DOI 10.1515/tlr-2016-0002 Abstract: A morphological asymmetry is shared by certain Dravidian (and Finnic) languages. The phonological shape of a negation element is dependent on the finiteness of the verb it negates. Pragmatic factors are identified that could motivate the development of this shared asymmetry, using evidence from the grammar of a Dravidian-influenced contact language. I will show that contrastive finiteness marking (finite and non-finite morphology) can facilitate the development of pragmatically-motivated linear reordering of affirmative clauses and negated clauses in order to accommodate new information structure conventions, extending the contrast to negated verbs by expanding the functional range of a negative imperative marker. Radical contact languages resulting from collective adult second language acquisition in naturalistic social contexts are typically presumed to feature reduced functional morphology, in which only highly salient contrasts, such as temporal contrasts, are formally instantiated. If a formal finiteness contrast and other relatively marked properties (“complexity”) could develop in a highly analytic contact language that did not previously have them, this suggests that such a sequence of changes is in fact as plausible among genetically-unrelated languages in a sprachbund as it is over longer periods of time in genetically-related languages. We can observe this by examining grammatical change in a language that previously lacked both a finiteness contrast and a corresponding negation asymmetry, but which developed both the contrast and the asymmetry as a result of contact with a genetically-unrelated language that has analogous properties. Sri Lankan Malay (SLM) has undergone grammatical change due to contact with Dravidian (primarily Sri Lankan Muslim Tamil). Several of these changes involve verb morphology and syntax, and are plausibly motivated by discoursepragmatic triggers. Consideration of tense and (non-)finiteness phenomena, as well as their reflexes in SLM negation, suggests a discourse-pragmatic motivation for these changes. Two discourse processes could conspire to motivate the development of the new morphology in SLM. The first is a clausal asymmetry, in which the predicate representing the most recent event is ordinarily in focus, indicated by tense morphology and position of the clause relative to clauses referring to *Corresponding author: Peter Slomanson, University of Tampere, Tampere, Finland, E-mail: slomanson@gmail.com 366 Peter Slomanson subsequent events. The second is the communicative need to reassign focus in certain contexts to a temporally non-primary clause, one referring to an event that did not take place first. In spite of a constraint in Dravidian languages blocking the marking of functional contrasts under negation (so that only a negation morpheme can be prefixed to the verb), negation morphology encodes an obligatory finiteness contrast, optimally supporting these information-structuring processes. The clause describing the most recent event in a sentence remains visibly finite under negation, when a temporally secondary clause is focused. Keywords: pragmatics, information structure, contact languages, morphosyntax, bilingualism, language change, typology 1 Introduction The goal of this paper is to demonstrate the importance of discourse-pragmatic factors and second language acquisition in cases of extensive accelerated morphosyntactic change. This will be based primarily on evidence from contact between two typologically distant languages, a highly analytic Austronesian SVO language and an agglutinating Dravidian SOV language.1 The Austronesian language, vernacular Malay, has become a new agglutinating SOV language Sri Lankan Malay (hereafter SLM), using Malay morphemes for new functional contrasts. The changes that took place can be motivated by appealing to preliminary discourse-pragmatic changes that eventually necessitated morphosyntactic change. I claim, following Prince (1998, 2001) as well as work in creolistics generally, that (adult) second language acquisition facilitates the types of profound structural change that generally occur very slowly over time, if at all, even in highly bilingual and multilingual speech communities in which we find fairly extensive interaction and cultural diffusion between the groups respectively associated with one of the languages.2 Given evidence for extensive L2 acquisition of 1 While I will discuss typological parallels in the expression of non-finiteness under negation, in order to demonstrate that this is an available strategy that can be found in a range of unrelated languages, I do not discuss the extent to which all the grammatical phenomena described in this paper can be found outside Sri Lanka. I demonstrate how discourse-pragmatic change can serve as input to morphosyntactic change, and that is what is expected to generalize to other language pairs. The frequency with which a specific discourse-pragmatic configuration appears cross-linguistically and the number of times this has led to a specific type of grammatical change in contact languages is not central to what I have chosen to show. 2 In Prince’s example, Eastern Yiddish never became a structurally Slavic language, in spite of the length and intensity of contact, although pragmatic borrowing from Slavic is extensive. (See Section 5 of this paper for more discussion.) Catalyst for the development of (non-)finiteness 367 Malay by adult Sri Lankans, the facilitating conditions proposed by Prince apply in the development of SLM. The possibility of earlier shifts in discourse-pragmatic conventions, particularly shifts that take place after a language is acquired by a proportionally large number of adult learners from outside the speech community, should be addressed in discussions and investigations of linguistic change. This possibility should also be addressed in discussions of the apparent origins of areal features in sets of languages. Such shifts may be plausible and even necessary precursors to the greatly accelerated morphosyntactic change that we find in certain bilingual or historically bilingual speech communities. The literature contains many references to syntactic reanalysis of surface configurations. Those reanalyses were based on the increased frequency of pragmatically-marked, but syntactically ordinary constructions that were already available, although optional. In a language that has been subject to contact effects due to changes in external context (such as migration), pragmatic reorganization can also trigger the development of construction types that were syntactically and morphosyntactically impossible in the original grammar. In order to illustrate this process concretely, I will be drawing in the first place on data from SLM, and for purposes of comparison, I will also draw on other languages, particularly Sri Lankan Muslim Tamil, often known by its speakers as Shonam, a language that happens to have in common with the major Finnic languages that the negation of non-finite verbs contrasts morphologically with the negation of finite ones, and that this is accomplished by means of an abessive construction.3 I take Shonam (rather than the language of the Sinhalese majority in the country) to be the model language for most of the changes that have taken place in SLM, based on external and morphosyntactic evidence (Hussainmiya 1987, 1990; Slomanson 2011, 2013a). Investigating triggers for typological shift is of particular relevance to linguists interested in the effects of collective second 3 All SLM examples are either taken from transcriptions of my own field recordings in the Malay village of Kirinda, in southeastern Sri Lanka, or else are constructed sentences whose grammaticality I checked by triangulating judgments with native speaker consultants with proven skills in this area. One consultant, Mohamed Jaffar, has recently acquired extensive formal training in linguistics, and has been presenting his own research on SLM at specialist conferences. His judgments of the relevant sentences in this paper do not differ from those of the other native speakers consulted in the field, although his grammar and lexical inventory may differ in other ways from that of my consultants in Kirinda and elsewhere in Hambantota province. All Shonam examples are from consultants who acquired spoken Shonam in childhood. Examples from other languages are based on my own knowledge (Irish and Sinhala) or were elicited from native speakers (Estonian, Finnish, and Hungarian). For the Estonian judgments, I would like to thank Liina Lindström. The grammaticality of the Sinhala examples was confirmed by native speakers. 368 Peter Slomanson language acquisition and the ways in which this process can yield radical linguistic change under specific conditions. We can infer from the picture provided by SLM that similar trajectories of change are likely to have arisen in and between other language families in the more distant past, giving rise to areal features and/ or apparent genetic relationships. In Section 2, I introduce the basic questions that have motivated research on SLM. In Section 3, I discuss the view that a new finiteness contrast is somehow anomalous or unexpected. In Section 4, we will see how information structure can motivate morphosyntactic change. In Section 5, I discuss the idea that discoursepragmatic change without collective second language acquisition does not contribute, at least in the short run, to the kind of structural change that we associate with a shift in typological class. This is based on Prince (1998, 2001) that deals with Yiddish, yet the observations were intended to apply to language contact situations involving other languages as well. In Section 6, I present the specific functional and communicative benefit to SLM speakers of instantiating a finiteness contrast where one did not previously exist. In Section 7, we will see that the abessive expression of negation can be shown to encode the non-finite status of a clause in a range of unrelated languages, including Shonam, although this particular strategy was not replicated in SLM. The cross-linguistic suppression of functional contrasts under negation will be discussed. We will see how the finiteness contrast is expressed under negation, in spite of the Dravidian constraint that limits the appearance of functional morphology in negated clauses. In Section 8, I discuss the status of the finiteness contrast under negation as a creative generalization. In Section 9, I present the contrast with the expression of negation in Sinhala. In Section 10, we see how infinitival morphology in affirmative contexts was grammaticalized in SLM. In Section 11, we see how another functional contrast, simultaneity, also follows from the asymmetrical organization of finite main clauses and adjunct clauses in Shonam. 2 Background It is important to understand potential initial triggering mechanisms for the sequence of changes that yield a shift from one typological class to another, particularly in cases in which it is completely unambiguous that the grammars at opposite ends of the process are typologically distinct, based on factors such as constituent order, branching direction, the number and type of grammaticalized functional contrast, and the complexity of phonological words. It is true that shifts from analytic to synthetic status do not necessarily involve Catalyst for the development of (non-)finiteness 369 more than the phonological weakening of free-standing functional elements and constrained distribution, however shifts involving the accretion of new functional contrasts are perhaps more legitimately treated as complexification. Both increases in inflectional morphology and the accretion of new functional contrasts have reflexively been treated as unlikely, or at least as infrequent, under language contact conditions. This applies to language contact cases in which there has been extensive adult second language acquisition, rather than to cases involving an apparently uninterrupted history of stable intergenerational bilingualism. If cases of so-called complexification are indeed infrequent, then it is important for us to understand which conditions yield the exceptional outcomes, exemplified here by the grammar of SLM. The empirical data for cases in which it is absolutely clear that such shifts took place currently involves a small number of languages. The relative scarcity of languages providing us with evidence of complexification in the present day could dissuade us from investigating a potential profusion of such shifts in a past for which we have limited direct evidence. It is the markedness and consequent anomalousness of the grammar that developed in SLM, particularly its inflectional complexity and the precursors of this in clitic phenomena, that has attracted recent research interest (inter alia Ansaldo 2008; Nordhoff 2009; Slomanson 2006, 2008b, 2009; Smith et al. 2004; Smith and Paauw 2006). Contact linguists are accustomed to thinking of precisely those languages as contact languages in which typological change is historically recent, and in which change is therefore visible. This tendency may cause us to inadvertently overlook much older cases in which the grammars of genetically unrelated languages converge as a result of communicatively-motivated factors, such as newly shared discourse conventions. What makes it easy for us to overlook those older cases is the fact that the relevant diachronic stages may be poorly attested and the external facts poorly documented, however this is a challenge to the linguist to engage in plausible reconstruction of earlier stages in a language’s history. Is it possible for a language to preserve its entire lexical inventory, both open and closed class, while at the same time becoming a member of another typological class? This last scenario is exactly what has happened in historical memory to Malay in Sri Lanka, hence the value of the example. The external background to the creation of a Dravidianized variety of Malay is the fact that Malay-speaking Muslims were brought to urban Sri Lanka in the mid-seventeenth century. At that time, Sri Lanka already had a substantial and wellestablished urban Muslim population with a distinctive Tamil-based language and culture. There is extensive historical evidence for widespread interaction between the two Muslim groups throughout the subsequent period, including 370 Peter Slomanson popular nineteenth century religious materials for a Malay audience written in Shonam in modified Arabic script, as well as bilingual religious and secular nineteenth-century texts, and other forms of documentation. There is also considerable evidence for ethnolinguistic intermarriage in the Muslim communities (Slomanson 2013a).4 The linguistic outcome of contact between Malay and Shonam, its model language, is that SLM is now for all intents and purposes a Dravidian SOV language, albeit one of largely Malay lexical base. The Malay lexical source varieties are highly analytic SVO languages, whereas Shonam is highly synthetic, and SLM has moved quite a distance in that direction as well, in both verbal and nominal morphology. Syntactic and morphosyntactic borrowing, including inter alia cross-categorial reversal of headedness/branching direction, are much more difficult, much more incremental, and therefore much slower than lexical borrowing (see Lehmann 1990 for an overview).5 If that is so, then a case of minimal lexical change, at least with respect to phonological shapes, accompanied by maximal 4 Given that the external question of intermarriage between Malay speakers and Shonam speakers has proven to be controversial, given its relevance to the claim that pre-twentieth century linguistic influence came largely from Shonam, and to a significantly lesser extent from Sinhala and its native speakers, Bakker (2013) cites genetic literature (Papiha et al. 1996) as evidence against extensive intermarriage. Based on the results of that study, Malays are the most endogamous Sri Lankan ethnic group. As intriguing as these findings may be as a new variety of evidence in the controversy, the study cited, not surprisingly for a research paper in population genetics, does not take socio-historical factors or regional variation into account. The Malay samplewas drawn from 103 participants in the Colombo area, whereas SLM speakers are found in sociologically and culturally divergent communities all over the island. The communities in the Colombo area are markedly stratified by social class, with attitudes toward marriage with non-Malay Muslims not consistent across communities or social groups. The sample in the 1996 genetic study may not be representative of the entire population of approximately forty thousand people, among whom there is considerable anecdotal evidence of a long history of frequent intermarriage with Shonam speakers. Consequently the existing historical record, including extensive marriage records that are explicit about ethnolinguistic identities, cultural evidence from texts, including bilingual religious and secular nineteenth century documents in modified Arabic script, written by and for ethnic Malays cannot be disregarded. 5 An alternative view, strongly associated with Weinreich (1957), holds that syntax is as easily acquired as lexical material, and that free-standing elements are necessarily much more easy to borrow than bound morphology. I see no reason to concur with the first generalization, while the second is a tendency that seems to be dependent in practice both on the respective grammars of the languages in contact and on the acquisition context itself. Catalyst for the development of (non-)finiteness 371 morphological and syntactic change, bears explaining. Neither collective language shift nor a break in intergenerational transmission are contributing factors in this case, adding to the explanatory challenge. 3 The development of (modeled) non-finite structures Contact linguists have tended to regard the development of morphologically non-finite forms such as participles and infinitives as unlikely to occur in the majority of cases in which morphological marking of non-finite verbs had been leveled at an earlier stage in the history of contact. This is based on examination of a large number of canonical creoles that developed from pidgins (cf. Bickerton 1984; Mufwene 1989). At the same time, it would be misleading to assume that non-finite structures have insufficient functional and pragmatic impact cross-linguistically in the languages in which they already appear when discourse-pragmatic conventions are not taken into account, whether or not they are reinstantiated in a particular contact language, as attested by adverbial adjunct clauses in which the main verb is a participle, and by the frequent purposive and irrealis interpretations assigned to infinitives. A cross-linguistically generalizable correspondence between the potential pragmatic utility of a particular functional contrast and the likelihood of its instantiation in contact languages generally is likely to prove elusive. Obviously the acquisition as a second language of a model language that makes extensive use of participles and infinitives as discourse devices will increase the chance that the same structures will somehow develop be in the contact language, given sufficient collective bilingualism in the contact language and the model language. 4 A discourse-pragmatic function for (new) non-finite verb forms In the language development scenario presented here [for SLM], in which typologically-discordant grammars were in contact, there was a communicative benefit to replicating a cross-clausal finiteness asymmetry in the morphosyntactically-restructured grammar, or in what would become a restructured grammar. The finiteness contrast functions as an information-structuring device. The 372 Peter Slomanson information in question is the hierarchical status and temporal ordering of predicates (technically of clauses) in an utterance in which a sequence of related events are listed. The discourse-pragmatic function of the finiteness contrast in this context is to formally distinguish the most recent event from subsequent ones. The most recent event can take place prior to the actual utterance or not, but it must be tense-marked. The relationship between the clauses is encoded by expressing the temporally primary event (the last event to take place) last, as a tense-marked verb, and the subsequent events first, and in the order of their occurrence, as participles (1). This could be achieved through linear ordering alone, however the morphological contrast permits the reordering of event clauses, in order to focus a temporally secondary event clause, now to be overtly marked as non-finite. This enables an override of the inherent clausal focus on the most recent event, marked through the use of tense morphology, the only available reflex of finite status. In (1), the clause with the enlarged bold font is the temporally-primary matrix clause that is explicitly tense-marked. The preceding clauses are participial adjunct clauses representing temporally secondary events, marked as non-finite. The aspectual/participial and event prefixes are underlined in the example. (1) SLM iskuul na a(bbi)s-pi, mulbar a(bbi)s-blajar, Miflal school P ASP.NFN-go Tamil ASP.NFN-learn Miflal attu=nyanyi su-tulis. IND=song PST-write ‛Having gone to school, (and then) having learned Tamil, Miflal wrote a song (in it).’ This is a sequence of clauses, each within its own intonation phrase, but clearly falling within a single sentence, as the initial clauses are non-finite adjuncts, and obviously dependent. This type of sentence is not required in an analogous context in English, of course, and the English translation provided sounds rather odd or stilted, but it is nonetheless grammatical, and adequately illustrates what is required in languages other than English that are spoken in Sri Lanka. Note that agreement of any type as an additional reflex of finiteness has never developed in SLM, and this is unsurprising, given the fact that although subject-verb agreement is characteristic of more widely-spoken Tamil varieties, it is not characteristic of older and conservative Shonam varieties. As we will see, the pragmatic motivation for the finiteness contrast in affirmative contexts in SLM carries over to negative contexts, in spite of the fact that tense information is obscured in negative contexts, yielding finite negation markers (tara, tuma) and a single non-finite negation Catalyst for the development of (non-)finiteness 373 marker (jang). Each of the clauses in (1) can be negated using one of these markers, with tara replacing the past tense marker su (si), and jang replacing the participial a(bb)is. This yields the sentence “Not having gone to school, not having learned Tamil, Miflal did not write a song in it.”6 What alternatives are there for a grammar to select, in order to express the sequence of events in (1)? Possible choices include inter alia the earlier Malay pattern, requiring a chain of clauses of equal status, with the focusing of a temporally non-primary clause signalled prosodically. This was disfavored however, with the adoption of the Sri Lankan areal pattern in which peripheral focus is strongly preferred over exclusive dependence on prosodic focus-marking. The reordering of event clauses is a discourse context in which grammaticalization to mark the finiteness contrast is motivated, in order to preserve the formal indication of the relative temporal status of the events. It is not motivated by an indiscriminate process of morphological convergence, in which a language necessarily adopts even the most marked features of its model language. That is a scenario that the term metatypy (grammatical change involving convergence on the grammar of an unrelated ambient language in which the speakers of the changing language are also fluent), as invoked in Ansaldo (2011) may label, however the use of such a label in this context does not provide any communicative or grammatical motivation, hence it has no explanatory value with respect to the actual process of linguistic change (Slomanson 2012). A discourse culture-specific pragmatic motive proposed here, in which discourse conventions first shift, and this subsequently triggers morphosyntactic change, is more plausible than random unmotivated selection from a feature pool, a scenario in which even markedness and economy considerations appear to play no role. How would speakers of SLM reassign focus to a temporally non-primary clause, given the discourse culture in which they had become participants? 6 The negative non‐finite construction was described in Slomanson (2006, 2008b, 2009, 2013b). It was also described for Upcountry SLM in Nordhoff (2009), with the local pronunciation jamà. Nordhoff translates jamà as ‘without’, which is idiomatically expedient, however somewhat misleading, since the form occurs with verbs and adjectival predicates only. This is therefore not a functional extension of the preposition ‘without’ as with aamee in Shonam, gan in Irish, and so on (see Section 7). In his examples, he shows that jamà can be used in an adverbial context without necessarily marking completive aspect, so without the temporal sequencing I argue here to be the origin of the finiteness contrast in SLM. Nordhoff’s observation is cross‐ dialectally accurate. Jamà/jang/je can and does now negate any non‐finite verb or non-finite converted adjective in any type of adjunct or complement clause. 374 Peter Slomanson The answer to this in (1), repeated here below, in which the interclausal finiteness asymmetry is exemplifed, renders the clause sequence in (2) possible. Here the participial status of the predicate in the dislocated adjunct clause preserves the relative temporal status of the event, while simultaneously marking the event as the focus of the sentence as a whole. (1) SLM iskuul na a(bbi)s-pi, mulbar a(bbi)s-blajar, Miflal school P ASP.NFN-go Tamil ASP.NFN-learn Miflal attu=nyanyi su-tulis. IND=song PST-write ‛Having gone to school, (and then) having learned Tamil, Miflal wrote a song (in it).’ (2) SLM iskuul na a(bbi)s-pi, Miflal attu=nyanyi school P ASP.NFN-go Miflal IND=song su-tulis, mulbar abbis-blajar. PST-write Tamil ASP.NFN-learn ‛Having gone to school, Miflal wrote a song, having learned Tamil.’ The interclausally asymmetrical negation morphology follows from this, since for independent reasons, independent morphological reflexes of finiteness and of its absence are suppressed under negation in the Dravidian sprachbund, as they are in somewhat contrasting ways in Finnish and Estonian. If in a contact language, in this case SLM, the finiteness contrast becomes highly salient as a result of a shift in discourse organization, then it becomes increasingly necessary for non-finite clauses to signal non-finiteness overtly. This applies equally to negated non-finite clauses. In Section 7, we will see examples of negation asymmetries in other languages, however the motivation for these asymmetries is not necessarily the same as the motivation in SLM. In SLM, reflexes of finiteness are associated with the pre-verbal position, however due to the Functional Stacking Constraint in that language blocking the stacking of bound morphology in that position (Slomanson 2006, 2007, 2009, 2011), no marker of non-finiteness (infinitival or participial) can co-occur with a marker of negation. The development of a specifically non-finite negation marker, through functional extension, increased the syntactic and pragmatic transparency of negated clauses. In spite of the apparent creativity in this development, non-finite clauses could reasonably be expected to develop some overt Catalyst for the development of (non-)finiteness 375 morphological indication of the cross-clausal status asymmetry that is clearly marked in affirmative clauses.7 5 Pragmatic change versus syntactic change Ellen Prince (1998, 2001), writing about Yiddish, describes a kind of feeding relationship between pragmatic change and syntactic change. Her claim, expressed most explicitly in Prince (2001), is that grammars in general can accommodate new information-structuring requirements to the extent that their existing syntactic systems permit, but that actual systemic syntactic change requires more dramatic events. In her (1998) article, Prince cites one case that is actually similar in a sense to the SLM case discussed here, since it also has to do with the temporal marking of verbs relative to each other in sequences of clauses. In brief, this involves a change from the Germanic system in which temporally primary clauses referring to events preceding the time of utterance are expressed as past or perfect, whereas non-primary events are pluperfect. As a result of contact with Slavic, all events preceding the time of utterance are simply expressed as perfect. The decline of the Germanic system eventually led to a dramatic loss of triggering evidence for the original Germanic pluperfect, and this precipitated changes in the grammar. Note that while this is a grammatical change triggered by pragmatic change itself, triggered by language contact, it involves the loss of an existing construction, rather than the addition of an entirely new morphosyntactic contrast such as finiteness. For such an accretion, dramatic events are likely to involve acquisition context and demography, and the acquisition is likely to involve adult learners. If the diachronic changes in the grammar of Yiddish discussed by Prince involved a kind of loss, as a result of syntactic reanalysis, the development of the finiteness contrast in SLM involved a striking accretion. Under the circumstances, if Yiddish had no finiteness contrast, would it have been likely for it to grammaticalize the necessary morphology in a similar way? The introduction of a completely new functional contrast, and the development of corresponding 7 As we can see in Section 8, although there is no explicit future marking under negation, the contrast between an irrealis interpretation (tuma) and a default past interpretation (tara), may be a response to the suppression of independent tense morphology under negation. Such solutions have been documented for historical Dravidian languages in which, for example, a negated infinitive may represent past tense, and a negated participle non-past, circumventing the suppression of explicit tense morphology. 376 Peter Slomanson morphology through grammaticalization seems to be a function of adult second language acquisition, when suitably motivated by discourse culture pressures and by acquisition context. This can lead not just to simplification through deflection, which is not itself a response to pragmatic factors, but second language acquisition can also lead to morphological complexification, motivated by pragmatics, as we have seen. When the original grammar is destabilized, interlanguage development permits a degree of restructuring that can include the accretion of new functional contrasts and the grammaticalization of supporting morphology. These outcomes, in which we do not find the same degree of restructuring in Yiddish as in SLM, stems from the fact that whereas Yiddish speakers became bilingual speakers of Slavic languages, and adopted a range of Slavic discourse conventions, as constrained by their existing Germanic grammar, acquisition of Yiddish by Slavic speakers was negligible. In the case of SLM, in contrast with the Yiddish case described by Prince, large numbers of native Shonam speakers acquired Malay, with “large” to be understood as relative to the L1 community, although there is no historical evidence that the L1 community was ever actually outnumbered by learners. I take the grammar of SLM to have developed ultimately through the convergence of a second language grammar used by Shonam speakers and a first language grammar used by Malay speakers, largely in bilingual households. The changes began with pragmatic accommodation. This does not imply that all households were bilingual, but only that the L2-influenced lect acquired the status of a model for ordinary vernacular functions within the community. Note that the persistence of SLM in bilingual homes continues to this day. I have interviewed non-Malays who have acquired fluent SLM only subsequent to marriage to a native speaker. The likelihood of this scenario may be a function of the density and multiplexity of local Malay networks, and is a matter that is badly in need of sociolinguistic and anthropological research. There have been references in the literature (Ansaldo 2008; Nordhoff 2009) to the relative potential grammatical influence of Shonam versus Sinhala in which the authors appeal to absolute speaker numbers in the country (Ansaldo 2008) and to speaker numbers in the regions in which most SLM speakers have lived (Nordhoff 2009). These arguments fail to take into account the organization of small ethnic and religious communities in many south and southeast Asian communities until recently, and certainly the absolute priority assigned historically to religious identification over any other social category in Muslim communities in Sri Lanka. This is not to say that SLM speakers and other Muslims had no contact with native Sinhala speakers (most of whom are Buddhist and a minority of whom are Roman Catholic), however historical Catalyst for the development of (non-)finiteness 377 marriage patterns, residential patterns, and evidence from nineteenth century texts, including bilingual Malay-Shonam didactic materials, demonstrate much greater network density and multiplexity across Muslim linguistic communities than between Muslims and Sinhalese.8 They also illustrate the high value 8 In this as in other diachronic research on minimally attested (stages in the development of) contact languages, the external history is critical to our understanding of internal changes, and this must often be pieced together based on evidence from archival investigation. Where there is no discussion of external evidence in grammar-oriented papers, this leaves the diachronic contact linguist open to criticism in later work, if the historical basis for linguistic hypotheses is in any way controversial. Ansaldo (2008) dismisses the sociohistorical claim in Hussainmiya (1987, 1990) that Malays in Sri Lanka have consistently tended to marry Shonam (“Tamil”) speakers with some frequency (and to marry Buddhists only very rarely). Hussainmiya’s claim is based in part on evidence from head thombos (colonial population registries) without provision of actual evidence. In support of his rejection of that claim, Ansaldo claims to have personally examined thombos (specifically) associated with the period of 1678–1919. Unfortunately, the period in Ansaldo’s claim extends well over a century beyond the point at which colonial administrations ceased to maintain these registries, or to record Muslim marriages at all. (The Dutch administration did so, but the British administration did not.) Ansaldo’s position is therefore considerably weakened by this claim, which is central to his dismissal of Hussainmiya’s historical research and the subsequent linguistic research in which Tamil/ Shonam has to some extent been treated as more significant than Sinhala, but not without providing linguistic argumentation and evidence that Ansaldo himself never challenges. Ansaldo objects to what he refers to as a “Tamil bias” in the work of other linguists who have published papers on SLM, including Bakker (2000), and Smith (2004) in particular. In Ansaldo (2011), the author rejects my criticism of his Tamil bias claim in Slomanson (2011), by questioning the existence of Shonam and “Mloaish”, a red herring. Shonam is the conventional glottonym in and for Sri Lankan Muslim Tamil, and “Mloaish”, a regrettable typesetter’s error in the spelling of “Moorish”, a frequent ethnonym for Shonam-speaking Muslims. Ansaldo states that Shonam “had not been uncovered at the time (of Ansaldo 2008) and could not have been referred to in any possible way”, and that “Shonam and Mloaish come across as somewhat ad hoc lingistic constructs and are thus impossible to evaluate”. A glottonym is not a linguistic construct. He further states that I (1) “criticized archival data, in support of a Tamil bias”, and (2) “failed to provide archival evidence” in favor of my claim of frequent Malay-Moorish intermarriage. What is suggested by Ansaldo’s locution in (1) is that I questioned the existence of archival data that do not show frequent intermarriage, essentially negative evidence, (2) without providing new historical evidence. (Ansaldo has only offered negative evidence to support his own historical position.) Ansaldo’s objection in (2) is legitimate and this omission was corrected in Slomanson (2013a), a paper largely dedicated to resolving this disagreement in the literature on the historical facts, rather than the internal linguistic ones. My investigation of Muslim marriage records in the nineteenth and early twentieth centuries shows that another type of record (i. e. not the colonial head thombos), called a kadutam, was kept in mosques, in accordance with British colonial policy, in which certain types of record keeping were communalized. The evidence there for highly frequent Malay-Moor (i. e. Shonam-speaking) intermarriage is clear. I also provided unambiguous photographs of earlier marriage records from the head thombos, evidence of the type that Ansaldo claimed does not exist. In a published review 378 Peter Slomanson historically assigned to Shonam in Sri Lankan Malay communities, as a language of religious education and process. That value persists in the most thoroughly SLM-speaking community, the village of Kirinda in southeastern Sri Lanka, in which all education from pre-school onwards is Tamil/Shonammedium and religious knowledge in Shonam is disseminated by loudspeaker from the largest mosque. Criticism by Ansaldo and Nordhoff (for example Nordhoff 2012; Ansaldo 2014) have focused on their objection to the relevance of second language acquisition and bilingualism within households, without taking the matter of network density and multiplexity and historical diglossia (with Shonam as a communal high language) into account. of the anthology in which Slomanson (2013a) appears, Ansaldo neglects to evaluate the evidence or argumentation in my paper, but instead returns, as in Ansaldo (2011) to the glottonym, a red herring, with the following comments: “The main claim Slomanson wishes to make in this chapter is that, unlike what is argued under the “Tamil bias” label, namely the idea that the intense Malay-Tamil bilingualism mentioned above is, basically, historical fiction, Tamil-Malay marriage did indeed happen on a significant scale. This may be confirmed if the new or reinterpreted historical material presented is accurate. The implication of this would be that Tamil is indeed a fundamental element in the formation of SLM. This can be additionally corroborated by the role of a variety of Tamil known as Shonam allegedly spoken in SLM communities. Unfortunately Shonam, as a variety, remains as elusive as ever in this volume under review. No other author apart from Jaffar makes reference to it, and he does so in passing. Slomanson does not give us any clue as to its linguistic relevance in the formation of SLM, so we cannot possibly appreciate it. Most authors in this book support a different model, akin to metatypy which invokes more than two languages in genesis: Bakker, Gair, Smith as well as Nordhoff in the final chapter who clearly states that the “creolization” approach, which builds on this idea of Tamil-Malay intermarriage, is not supported by a multilayered approach, while typological convergence, which does not require any role of Tamilinfluenced bilingualism, is likely to be correct for significant stages of the language.” I have been referring explicitly to Muslim Tamil and/or Shonam (two names for the same variety) since Slomanson (2006), and have provided data examples. The relevant variety is therefore not “elusive”. The chapter referred to contains photographic plates with the relevant evidence (marriage records, a pedagogical glossary in SLM, Shonam, and Arabic, and representative eighteenth-century thombo entries). The concept of metatypy does not require more than one model language. Typological convergence is simply a descriptive label for a collection of changes that undeniably took place, and does not preclude a non-primary role for an additional language (i. e. Sinhala as an adstratal language), nor have I. Mohamed Jaffar is a linguist and native speaker of SLM and Shonam, on whose advice I began using the native glottonym in my papers. The most extensive linguistic argumentation for the primacy of Tamil/ Shonam influence is found in Smith (2012a) and (2012b). Catalyst for the development of (non-)finiteness 379 Although the SLM case is an example of typological change resulting from effects of second language acquisition, the accretion of a finiteness contrast is not evidence of imperfect acquisition of a target grammar, but of communicatively-motivated morphosyntactic support for altered information-structure conventions. These conventions would have been far more salient to bilinguals than formal morphological and syntactic contrasts if those contrasts were not motivated by accommodation to local communicative-pragmatic norms. The development in SLM of tense and finiteness as obligatory grammatical contrasts is evidence of a secondary stage in speaker accommodation to the informationstructuring demands of a different discourse culture, a discourse culture in which finite status consistently signals most recent event status, and non-finite status signals temporal subordination. 6 The pragmatic value in developing non-finite morphology As we have seen, a plausible pragmatic motivation for the development of a finiteness/non-finiteness contrast in SLM can be identified in the fact that the discourse culture associated with the Sri Lankan sprachbund, as interpreted by speakers of SLM (some of whom were second language speakers), associates the sentential periphery with constituent focus, not just of nominal constituents, but of clauses. Returning to the example in (1), repeated below for convenience, note that the first two events are not literally in a temporal sequence, although we should understand that going to school preceded Miflal’s learning Tamil. (1) SLM iskuul na a(bbi)s-pi, mulbar a(bbi)s-blajar, Miflal school P ASP.NFN-go Tamil ASP.NFN-learn Miflal attu=nyanyi su-tulis. IND=song PST-write ‛Having gone to school, (and then) having learned Tamil, Miflal wrote a song (in it).’ His going to school was not completed prior to his learning Tamil. The sequence matters with respect to the onset of each activity, but not its completion. It is the non-primary temporal status of the non-finite adjunct clauses that is most salient 380 Peter Slomanson to speaker and listener, more so than their sequence with respect to each other. In the varieties of Malay originally brought to Sri Lanka, all the verbs in this sequence are likely to have been unmarked for aspectual status, and tense-marking, whether free-standing or phonologically-dependent, was unavailable. This means that an L1 Malay speaker in Sri Lanka in the process of accommodating Sri Lankan discourse conventions would be forced to depend on prosody, a disfavored strategy, and on the linear ordering of clauses, which would prevent their reordering for focus. The Sri Lankan languages, including SLM, are left-branching SOV languages, and the periphery is not pragmatically neutral, with focus a highly communicatively salient function in the sprachbund. Non-phonological focus marking, both syntactic and morphological, is preferred over focus marked only by prosody. The clause containing the most recent event normally appears in the right-most position as the (tensed) matrix clause. Clauses not representing the most recent event must be explicitly marked as non-finite, so that the status of the most recent event, which will always be tense-marked, will retain its status when a temporally non-primary event is focused through dislocation to the right-peripheral focus position. Note that this then is not simply a variant order, but is perceived as a pragmatically-marked order.9 7 The suppression of functional morphology under negation and the abessive expression of non-finite negation There is cross-linguistic variation in the ways in which non-finiteness is morphologized. We find this variation not just in the form of the verb itself, 9 The cross-linguistic parallel can be seen from the following Muslim Tamil examples, that correspond with the SLM examples in (1) and (2): (i) iskul-ukku pee-thu, tamil paad-icci, Miflal paath-ondu elludinaan. school-ALL go-PRT Tamil learn-PRT Miflal song-DET learn-PST ‘Having gone to school, (and then) having learned Tamil, Miflal wrote a song (in it).’ (ii) iskul-ukku peethu, Miflal paath-ondu elludinaan, tamul paadicci. school-ALL go-PRT Miflal song-DET learn-PST Tamil learn-PRT ‘Having gone to school, Miflal wrote a song, having learned Tamil.’ Catalyst for the development of (non-)finiteness 381 that is, in the extent to which tense and agreement are visible in the form of the negated verb. We also find it in the form of the negation element. Crosslinguistic contrasts in the phonological shape of negation elements may be a function of the general finiteness asymmetry between clauses in these languages, or they may have additional causes. One particular strategy is found in a number of completely unrelated languages, including Shonam, although not including SLM. Non-finite negation can be realized using a closed class element that conveys the sense of the adposition “without” that we otherwise find in nominal contexts (PPs and case-inflected NPs). In Estonian and Finnish, for example, non-finite verbs can take a nominalizing suffix, followed by abessive case inflection, as in the examples in (3) and (4). (3) ESTONIAN Läk-si-me kooli, enda-l kodutöö tege-ma-ta. go-PST-1PL school self-ADE homework do-NMZ-ABE ‛We went to school, not having done our homework.’ (4) FINNISH Meni-mme koulu-un teke-mä-ttä läksyjä-mme. go-1PL school-ILL do-NMZ-ABE homework-1PL ‛We went to school, not having done our homework.’ The asymmetry is certainly not universal within the larger Finno-Ugrian family however. In Hungarian, for example, formal infinitives are easily negated with nem (as are finite verbs), as in nem dolgozik (lit. “not to work”), so that it is unnecessary to resort to nominalizations with an abessive element, such as (a) nélkül (“without”), to express the same meaning. Whereas in Finnic, however, using an abessive suffix with a verb requires the insertion of a nominalizing affix, in Tamil and Shonam, use of the abessive marker in postpositional phrases requires the insertion of a negative existential marker, rendering the apparent postpositional phrase clause-like. Verbs and clauses may be case-marked in Tamil varieties and in the Sri Lankan sprachbund generally. (5) NOMINAL CONTEXT (SHONAM) Miflal Kirinde-kku poov-ille, [sor-ill-aamee.] Miflal Kirinda-DAT go-NEG rice-NEG.EXS-ABE ‛Miflal did not go to Kirinda without rice.’ 382 (6) Peter Slomanson CLAUSAL CONTEXT (SHONAM) Miflal Kirinde-kku [sor tind-aamee] poov-ille. Miflal Kirinda-DAT rice eat-ABE go-NEG ‛Miflal did not go to Kirinda without having eaten rice.’ The fact that no additional morpheme is required in the clausal context (6) suggests that “withoutness” may be more strongly associated with the verbal domain than are other adpositions. In a language such as Shonam, in which categorial contrasts are not weak generally, this may point to facts about grammatical abessiveness and why the subordination (adjunction) of negative clauses is frequently made possible cross-linguistically by the use of abessive constructions, as opposed to another grammatical strategy. Abessiveness is also found in the negation of non-finite clauses elsewhere, as in the Irish examples in (7) and (8), showing how the counterpart of English “without” is the (only) negative infinitival complementizer in that language. Irish is an Indo-European language with no significant history of contact with the other languages considered in this paper, demonstrating the fact that this is one of the negation strategies available to (not necessarily related) languages with non-finite complementation. (7) NOMINAL CONTEXT (IRISH) [Gan a chuid seacláid-e,] ní raibh an without POS PTV chocolate-GEN NEG be.PST DET páiste sásta. child happy ‛Without his chocolate, the child was not happy.’ (8) CLAUSAL CONTEXT (IRISH) Is deacair le páiste [gan a chuid seacláid-e a ithe.] COP difficult P child without POS PTV chocolate-GEN INF eat ‛It is difficult for a child not to eat his chocolate.’ Although the abessive construction in Shonam could have served as a direct model for non-finite negation in SLM, the way that this functional configuration was grammaticalized in SLM was creative and did not involve structural transfer or calquing. As we have seen, non-finite clauses in certain non-South Asian languages, including the Finnic languages and Irish, are negated using different morphological devices from those used for negation in finite clauses. The contrast between affirmative and negative verbs in finite contexts that we find in the grammar of SLM, based on a constraint in Shonam (and absent entirely from Catalyst for the development of (non-)finiteness 383 Sinhala) is found in other languages. The asymmetry that this yields has been investigated for a large number of languages in Miestamo (2005).10 In Finnish, functional contrasts deleted from a negated verb will reappear on a free-standing negation morpheme (9, 10), however in Estonian, functional information that is associated with finiteness and deleted under negation is not similarly restored (11, 12) elsewhere in the clause. In this respect, Tamil varieties align with Estonian, rather than with Finnish. (9) FINNISH Halua-n maito-a. want-1S milk-PTV ‛I want milk.’ (10) FINNISH E-n halua maito-a. NEG-1S want-ø milk-PTV ‛I do not want milk.’ (11) ESTONIAN Soovi-n piim-a. want-1S milk-PTV ‛I want milk.’ (12) ESTONIAN Ma ei soovi piim-a. 1S NEG-ø want-ø milk-PTV ‛I don’t want milk.’ (13) TAMIL Miflal sor tin-d-aar. Miflal rice eat-PST-3S ‛Miflal ate rice.’ 10 There are consequently two relevant negation asymmetries: the asymmetry in Miestamo (2005) in which negated finite verbs cannot display functional features other than negation, and the finite/non-finite negation asymmetry presented here, instantiated on non-finite verbs as abessive negation in Finnic and Tamil/Shonam, although not in SLM. 384 (14) Peter Slomanson TAMIL Miflal kozumbu-kku poo-v-ille. Miflal Colombo-to go-INF-NEG ‛Miflal did not go to Colombo.’ Although there is no agreement in conservative varieties of Shonam, nor in SLM, it is the morphology of contrastive tense that is similarly neutralized under negation.11 Malay varieties that were brought to Sri Lanka had optional freestanding aspect markers, but no morphological or syntactic instantiation of a finiteness contrast. (15) SLM Miflal nasi su-makan. Miflal rice PST.FIN-eat ‛Miflal ate rice.’ (16) SLM Miflal nasi tara-makan.12 Miflal rice NEG.FIN-eat ‛Miflal does/did not eat rice.’ 11 For many speakers, the pre-verbal finite negator tara always has past tense reference, whereas tuma always has non-past reference. For other speakers, tara can be general present, whereas tuma is habitual present. For predicates in which adjectives function as verbs (a process that is not equally productive in all varieties or for all speakers), tara can have both present and past tense interpretations, depending on the context. What is significant is that there is no dedicated tense morphology under negation, and that no speakers have dedicated tense morphology consisting of the three tenses, past, present, and future. Negation morphology also suppresses pre-verbal modality markers, such as bole/bələ/bər(ə) and mes(ti). It is possible for a negated modal predicate, such as bole, to appear before a lexical verb, but the predicate is not then phonologically bound to the verb. This is why mes(ti), which cannot occur as a predicate, can never be negated in the position immediately to the left of a lexical verb. 12 The vowels in the form displayed in SLM example sentences in this paper as tara- are usually reduced to shwa and whether or not the final vowel is heard is determined by phonotactic factors. (It is never heard prior to a vowel-initial verb stem. There is a post-verbal form tra, which is a contraction of tar(a)-ada, a negated existential marker or semantically empty auxiliary, depending on the context. Nordhoff (2009) uses th rather than t for these forms, in deference to the local convention of spelling dental variants of /t/ in this way when Romanizing words from Sri Lankan languages. I have not adhered to that convention, consequently tra and thra in different texts dealing with SLM refer to the same form. Catalyst for the development of (non-)finiteness 385 The evidence that we have seen suggests a potential implicational relationship between, on the one hand, the existence of a finiteness contrast and the suppression of the morphological reflexes of finiteness on lexical verbs under negation, and, on the other hand the contrastive phonological shape of negation elements in finite and non-finite clauses. This suggestion awaits broad typological comparison to confirm or disconfirm its validity. 8 Expression of the finiteness contrast under negation In Section 7, we saw examples of a finite/non-finite negation asymmetry in Estonian, Finnish, and Shonam. In SLM, the development of an analogous clausal asymmetry strengthens the marking of temporal hierarchy, particularly in sentences in which a temporally non-primary clause is in focus. Negation elements marking finite predicates in SLM are variants of the Malay form tara (17), or of the form tuma (18). Tuma, in etymological terms, is a contraction of tara (negation) and mau (volition). Although their phonological shape signals that they are finite, and tense, in the absence of agreement, is the only morphological exponent of finiteness, these negation elements are essentially unspecified for tense, although tuma is interpreted by some speakers as irrealis, and by others as habitual non-past, in the sense of “I never (will) do this”. There is no explicit past-present-future contrast under negation, as there is in affirmative contexts. Tara and tuma only negate verbs that, in affirmative contexts, can be tense-marked. Negation elements marking nonfinite verbs in SLM, whether these are participial or infinitival, are variants of the Malay form jang (from jangan) (19). This is historically the negative imperative marker in Malay, a function that it also retains in modern SLM. Jang can only precede verbs that cannot be tense-marked in affirmative contexts, including participles, infinitives, and imperatives. (17) SLM Miflal nasi tara-makan. Miflal rice NEG.FIN-eat ‛Miflal did not eat rice.’ (18) SLM Miflal nasi tuma-makan. Miflal rice NEG.FIN-eat ‛Miflal does/will not eat rice.’ 386 (19) Peter Slomanson SLM Nasi jang-makan na, Miflal si-hendat. rice NEG.NFN-eat P Miflal PST-tired13,14 ‛Not having eaten rice, Miflal became tired.’ The contrast is motivated by the fact that SLM, like Shonam and other Tamil varieties, suppresses tense morphology and other markers of finiteness status under negation. This may be regarded as a constraint that must be circumvented in some way, if the relevant information structure contrast, salient as we have seen in the discourse culture, is to be maintained in negation contexts. It is necessary for one’s interlocutor to know which clause is finite and which ones are not, regardless of whether these are affirmative or negative clauses. SLM, in order to accommodate the discourse conventions of Shonam, also affixes non-finite negation morphology to any negated lexical verb, albeit preverbally, rather than post-verbally, using the element jang, diachronically a regrammaticalized Malay imperative marker (Slomanson 2006, 2009). SLM has creatively extended the functional scope of jang to encompass all non-finite negation contexts, including participles, in adjunct clauses and infinitival complement clauses. This approximates the Dravidian model. We can see that replication of non-finite negation is not a simple process of calquing from Shonam, since SLM did not select the literal equivalent of “without”. It did not even select its equivalent in SLM, the complex postposition tra na(ng),15 which is in fact also a creative replication, loosely modelled on Shonam. Tra na(ng) was reserved for nominal constituents that are not 13 Adjective to verb conversion is largely productive in Kirinda. There is a certain degree of regional variation in the semantic classes of adjectives to which this process may be applied. 14 The form si occurs in Kirinda, where it has replaced su. Hendat is also a local SLM word that is unfamiliar to speakers in other areas, although it may have been used elsewhere in earlier periods. 15 The status na(ng) is unusual in that it is the one closed-class item that is borrowed from Javanese (Slomanson 2006), possibly through the Java Malay varieties originally brought to Sri Lanka in the seventeenth century. In Javanese, it is a free-standing allative adposition, whereas in SLM, it performs both the allative function, and the function of a dative clitic without allative meaning. My claim that it is a clitic is based on my observation that the indefinite determiner attu variably appears pre-nominally and post-nominally, and is itself therefore a clitic. When attu is post-nominal, it intervenes between the noun and na(ng), but na(ng) cannot be made to follow a parenthetical remark introduced by an intonational break, as in the English sentence “We’re going to, you know, a mosque”. Catalyst for the development of (non-)finiteness 387 clausal, as seen in the following examples. In (20), “without rice” is expressed with the postposition tra-na(ng), whereas in (21), an infinitival clause, nasi makan tra-na(ng) would be ungrammatical. The function of jang in examples (21) through (23) and in example (25) is to express negation and the absence of finiteness. The function of tara in (22), (23), and (25) is to express negation and finiteness. Based on the Malay lexical source varieties, this is likely to have been the original negation morpheme in all but imperative contexts. As morphological tense prefixation developed, yielding forms such as si-pi, (20) and (21) with past tense reference, a(rə)-pi with present tense reference, and a(n)ti-pi with future tense reference, these contrasts could not be expressed under negation. This is a Dravidian constraint not reflected in the grammar of Sinhala. (20) NOMINAL CONTEXT (SLM) Miflal Kirinde na si-pi, [nasi tra-na] Miflal Kirinda P TNS-go rice NEG-DAT ‛Miflal went to Kirinda without rice.’ (21) (ADJUNCT) CLAUSAL CONTEXT (SLM) Miflal Kirinde na si-pi, [nasi jang-makan na.] Miflal Kirinda P TNS-go rice NEG.NFN-eat P ‛Miflal went to Kirinda, not having eaten rice.’ (22) PARTICIPIAL CLAUSE (SLM) jang-tidur, Miflal nasi tara-makang. NEG.NFN-sleep Miflal rice NEG.FIN-eat ‛Not having slept, Miflal did not eat rice.’ (23) PURPOSIVE INFINITIVAL CLAUSE (SLM) jang-tidur na, Miflal nasi tara-makang. NEG.NFN-sleep P Miflal rice NEG.FIN-eat ‛To not sleep, Miflal did not eat rice.’ (24) IMPERATIVE CLAUSE (SLM) Nasi jang-makang! rice NEG.NFN-eat ‘Don’t eat rice!’ 388 (25) Peter Slomanson SLM iskuul na jang-pi na, mulbar jang-blajar na, Miflal school P NEG.NFN-go P Tamil NEG.NFN-study P Miflal attu=nyanyi tara-tulis. IND=song NEG.FIN-write ‛Not having gone to school, (and then) not having studied Tamil, Miflal did not write a song in it.’ Note that (26) is the negated version of (1), which I will repeat here for convenience. (1) SLM iskuul na a(bbi)s-pi, mulbar a(bbi)s-blajar, Miflal school P ASP.NFN-go Tamil ASP.NFN-learn Miflal attu=nyanyi su-tulis. IND=song PST-write ‛Having gone to school, (and then) having learned Tamil, Miflal wrote a song (in it).’ (26) SLM iskul na jang-pi na, mulbar jang-blajar na, school P NEG.NFN-go P Tamil NEG.NFN-study P Miflal atu=nyanyi tara-tulis. Miflal IND=song NEG.FIN-write ‛Not having gone to school, (and then) not having studied Tamil, Miflal did not write a song in it.’ The SLM negation contrast is partly sprachbund-discordant (Slomanson 2010), since the function and distribution of jang does not map straightforwardly from Shonam or Sinhala negation morphology. It is reasonable to treat this as a generalization of a morphosyntactic feature of Shonam however, that extends to all nonfinite verbal contexts, infinitives, participles, and imperatives, though in Shonam, imperatives are not marked in the same way as other non-finite verb forms. 9 The contrast with Sinhala The view in Ansaldo (2008) and Nordhoff (2009) that there is as much reason to attribute grammatical restructuring in SLM to the influence of Sinhala as to the influence of Shonam is difficult to defend, based on social history and on Catalyst for the development of (non-)finiteness 389 morphosyntactic evidence.16 There are interclausal asymmetries in Sinhala negation morphology, but there is a significant contrast between the way these are instantiated in Sinhala and the other languages in the sprachbund. In Sinhala, negation is ordinarily post-verbal and triggers the instantiation of focus morphology on the verb. The negated verb is necessarily in focus. Whereas the SLM contrast is strictly a finiteness/non-finiteness contrast, the Sinhala contrast strictly reflects matrix/non-matrix status. This is because non-matrix clauses are frequently tense-marked and finite, though they do not and cannot bear conventional negation morphology. Examples (27) and (28) demonstrate this. (27) NEGATION OF INFINITIVAL NON-MATRIX VERB (SINHALA) Farida porondu unaa [mas no-ka-nnə]. Farida promise AUX.PST meat NEG.NMX-eat-INF ‛Farida promised not to eat meat.’ (28) NEGATION OF NON-MATRIX TENSED VERB (SINHALA) [oyaa no-ya-nəwa nam] honda-yi. you NEG.NMX-go-PRS if good-PRD ‛(It’s) good if you don’t go.’ 10 Infinitival contexts The infinitival construction in SLM which takes the allative/dative marker na(ng) may have begun as a purposive adjunct, a function it has retained. The construction has been generalized to clausal complement contexts. A pre-verbal infinitival marker, of the form mə-V, also developed from the phonological weakening and (re)grammaticalization of the volitive/irrealis 16 Nordhoff (2009), based on the author’s investigation of the highland (“upcountry”) varieties, particularly Kandy Malay, argues for significant parallels between SLM and Sinhala phonology. Phonological parallels, however, may have developed more recently than morphosyntactic ones. There is apparent time evidence of recent change in the intonational phonology of SLM. This change happens to coincide with the period in which the traditional role of Shonam in SLM culture (as an important language of Islamic practice and education) began to weaken. More significantly still, that was precisely the period in which both the general importance of Sinhala in the SLM communities as well as the density and multiplexity of Malay-Sinhala contacts increased dramatically, primarily subsequent to Sri Lankan independence, after which these contact continued to increase, since a majority of Malay-speaking children began to be educated through the medium of Sinhala. I believe that this trajectory of recent phonological change would be revealed in an apparent time study involving the oldest fluent native speakers, leaving contact with Sinhala strictly peripheral to the glottogenesis of SLM. 390 Peter Slomanson element mau/mo. I take this construction to have appeared subsequent to the instantiation of non-finite participial adjuncts, of the type we have seen in (1). While the finiteness contrast arose as a way to displace and focus temporally non-primary clauses, the development of infinitival morphology and its corresponding negation with jang follows in part from the development of contrastive tense morphology. This enabled tensed verbs in SLM to take bare VP complements (i. e. INF-V TNS-try, meaning “try to V”). (29) INFINITIVAL BARE VP COMPLEMENT (SLM) Musba waghanam-yang mə-dapat na si-liyat. Musba vehicle-ACC INF-get P TNS-see ‛Musba tried to get the vehicle.’ Note that in Shonam, there are actually different types of infinitival adjunct, with dative-marked verbs in adjunct clauses most likely to be purposive, whereas other types of infinitival adjunct, as well as infinitival complements have specifically (i. e. non-dative) infinitival morphology, as in (30). (30) EMBEDDED INFINITIVAL COMPLEMENT (SHONAM) Miflal-ukku [viTT-ukku poo-k-] oonum. Miflal-DAT house-to go-INF want ‘Miflal wants to go home.’ In SLM, the construction with na (a variant of nang) in (29) variably with and without the prefix mə, is the only infinitival complementation strategy in SLM. The trajectory from purposive to infinitive has been treated as a universal tendency, for example in Haspelmath (1989). The trajectory stands apart from, but ultimately complements the scenario in which temporally subordinate adjunct clauses with the meaning “having done x” were responsible for the introduction of the finiteness contrast into the grammar of SLM (so that those same clauses could be dislocated and focused). 11 Simultaneity SLM has a temporal conjunction of simultaneity, ambe(l).17 A clause containing this conjunction, can be negated using the non-finite negation marker jang, as in 17 I am including the word-final segment /l/ in the discussion for etymological reasons and because it is still present in the most widely-spoken regional varieties, however it will be absent from my examples from Kirinda, where the conventional form is now ambe, without the final /l/. Catalyst for the development of (non-)finiteness 391 the clause final sequence jang-makan-ambe, meaning “while not eating.” In contrast with sequential event clause sequences, the presence of simultaneous event clauses cannot motivate the development of new grammaticalized morphology, however the status of jang as a non-finite negation marker has been generalized to non-sequential contexts. In the scenario that yielded the sequential sentence type, and ultimately the (non-)finiteness contrast, the discourse organization favored by the Sri Lankan sprachbund to express a series of related but sequential events requires a series of adjunct clauses, followed by a primary clause adjoined to an auxiliary. As we have seen, the tensed matrix clause marks the most recent event as the primary event, while participial adjuncts refer to subsequent events (31). In the sentence type involving simultaneity in which ambe(l) appears (32), ambe(l) is a temporal complementizer. It marks progressive aspect, in the manner of English “while”, and cannot precede the lexical verb. The auxiliary ada occurs in its own clause, the highest clause in periphrastic constructions. The evidence for this is the fact that ada, a semantically empty auxiliary, can be tense-marked and negated independent of a lexical verb associated with it (Slomanson 2008b, 2009), and this is also the case with the construction involving a lexical verb and ambe(l), as in the construction taksir ambe (si-)ada in (32).18 It is possible to reverse the order of the ambe(l) clauses, so that the right-most predicate is focused, however the tense interpretation will remain the same. The tense that is optionally marked on the auxiliary has scope over the embedded clauses. (31) THE PERFECT CONSTRUCTION (SLM) abbis baung, tee abbis minung, Rikas iskuul ASP.NFN awaken tea ASP.NFN drink Rikas school na abbis pi ada. P ASP.NFN go AUX ‛Having awoken, having drunk tea, Rikas has gone to school.’ (32) THE SIMULTANEOUS/PROGRESSIVE CONSTRUCTION (SLM) tee minung ambe, Rikas taksir ambe ada. tea drink PRG.CMP Rikas think PRG.CMP AUX ‛While drinking tea, Rikas is thinking.’ 18 The constructions in (32) and (33) are most closely associated with the southeastern variety. The prefix e- is a regional variant of abbis-, the aspectual prefix in (31). 392 (33) Peter Slomanson THE PROGRESSIVE PERFECT CONSTRUCTION (SLM) Rikas ini ari-pada ka baru ruma attu e-rikat Rikas DET day-PLU in new house IND ASP.NFN-build ambe (a-duudung). PRG.CMP (PRS-AUX) ‛Rikas has been building a new house these days.’ We would expect an SLM aspect marker to appear in pre-verbal position in adjunct clauses, however ambe(l) does not, suggesting that ambe(l) is a temporal conjunction or complementizer. The deletion of /s/ is not surprising, given that s-aphaeresis is historically a feature of the Malay vernaculars brought to Sri Lanka from Indonesia, as well as of SLM itself. As an indication of the etymology of ambe(l), the Malay form sambil sometimes (immediately), without aphaeresis, preceded the verb in the Malay of the generation raised in the period preceding and immediately following the turn of the previous century (Mohamed Jaffar, personal communication), although this usage has since vanished. The pre-verbal position in finite clauses is now associated exclusively with markers that are interpretable as tense-bearing, but not with aspect markers, since tense and aspect-marking never merge. Consequently, the position can therefore still be occupied by kapan, a temporal conjunction,19 to mark the relationship between two predicates at some (unspecified) place 19 The temporal conjunction clearly follows the Functional Stacking Constraint, whereas the homophonous interrogative element does not. The evidence for this is the fact that overt tensemarking is ungrammatical after kapan in an embedded clause. (iii) (iv) Rikas iskuul attu kapan si-kuttumun? Rikas school IND when PST-see ‛When did Rikas see a school?’ Rikas iskuul attu kapan *si-kuttumun a-suuka. Rikas school IND when PST-see TNS-like ‛Rikas liked it when he saw a school.’ Ansaldo (2009: 136) contains a surprising and unlikely token of pre-verbal ambe, that violates the Functional Stacking Constraint, by occurring adjacent to a temporal marker, and in which it is inaccurately glossed as “take.” He then cites me as claiming in Slomanson (2006) that “Indonesian Malay” has an identical pre-verbal particle marking progressive aspect, whereas I could only have stated that Indonesian sambil is a left branching complementizer, analogous with English “while,” which does not appear immediately adjacent to any verb. In fact, I did not discuss the construction in that publication. I presented the construction and my views on its development in Slomanson (2008a), however not as a “historical adposition”, as Nordhoff (2009: 176) states that I did. Catalyst for the development of (non-)finiteness 393 in time (34), when the predicate with kapan refers to a state existing at that time (i. e. “when he knew”) or to an action with no overt reference to its aspectual status (i. e. “when he came”, but not “when he was in the process of coming”). Kapan therefore marks past simultaneity without the progressive meaning marked by ambe(l) (Slomanson 2011). (34) SLM Rikas iskuul attu kapan-kuttumun a-suuka. Rikas school IND when-see TNS-like ‛Rikas liked it when he saw a school.’ 12 Conclusion We have seen that SLM features a discourse configuration, the temporal stacking of a sequence of clauses, that could serve as a catalyst for morphosyntactic changes, including the introduction of contrastive finiteness morphology. The cross-clausal finiteness contrast in negation marking was replicated in turn, in order to circumvent the obscuring of this contrast under negation. This replication was accomplished without borrowing or calquing the analogous Shonam non-finite negation morpheme. Instead, the functional scope of a near analogue in Malay to the Shonam element, a negative imperative marker, was expanded. The non-finite negation morpheme in SLM, jang, nevertheless conveys the functional meaning associated with the Shonam abessive suffix aamee. The element jang constitutes a reflex in negation contexts of the finiteness contrast in SLM that has high communicative salience in the discourse culture in which Sri Lankan Malays have participated since their ancestors were brought to Sri Lanka by the Dutch. The contact grammar function of jang can therefore be attributed to the need to continue to mark the temporal status of a late-occurring event as secondary, even after it is focused through dislocation to the right (focal) periphery of the sentence. Abbreviations 1S 3S ABE ACC ADE ASP first person singular third person singular abessive accusative adessive aspect 394 AUX CMP COP DAT DET EXS FIN GEN ILL IND INF NEG NFN NMX NMZ P PL POS PRD PRG PRT PST PTV TNS Peter Slomanson auxiliary complementizer copula dative determiner existential finite genitive illative indefinite infinitival negative non-finite non-matrix nominalizer preposition/postposition plural possessive predicate progressive participle past tense partitive tens References Ansaldo, Umberto. 2008. Sri Lanka Malay revisited: Genesis and classification. In Arienne Dwyer, K. David Harrison & David S. Rood (eds.), A world of many voices: Lessons from documented endangered languages (Typological Studies in Language), 13–42. Amsterdam: John Benjamins. Ansaldo, Umberto. 2009. Contact languages: Ecology and evolution in Asia. Cambridge: Cambridge University Press. Ansaldo, Umberto. 2011. Metatypy in Sri Lanka Malay. In Rajendra Singh & Ghanshyam Sharma (eds.), Annual review of South Asian Languages and linguistics, 3–15. Berlin: De Gruyter Mouton. Ansaldo, Umberto. 2014. SLM is dead, Long Live Sri Lanka Malay: A review article of Nordhoff 2013. Journal of Language Contact 7(2), 381–391. Bakker, Peter. 2000. Rapid language change: Creolization, intertwining, convergence. In Colin Renfrew, April M. McMahon & Larry Trask (eds.), Time depth in historical linguistics, 585–620. Cambridge: McDonald Institute for Archeological Research. Bakker, Peter. 2013. Sri Lanka Malay: New findings on contact. In Sebastian Nordhoff (ed.), Extreme language contact – The case of Sri Lanka Malay, 53–83. Amsterdam: Brill. Bickerton, Derek. 1984. The language bioprogram hypothesis. Behavioral and Brain Sciences 7. 173–221. Catalyst for the development of (non-)finiteness 395 Haspelmath, Martin. 1989. From purposive to infinitive – A universal path of grammaticization. Folia Linguistica Historica X(1–2), 287–310. Hussainmiya, Bachamiya Abdul. 1987. Lost Cousins: The Malays of Sri Lanka. In Dunia Melayu Occasional Paper 2. Bangi: Universiti Kebangsaan Malaysia. Institut Bahasa, Kesusasteraan dan Kebudayaan Melayu. Hussainmiya, Bachamiya Abdul. 2008 [1990]. Orang Rejimen: The Malays of the Ceylon rifle regiment. Bangi: Universiti Kebangsaan Malaysia. Institut Bahasa, Kesusasteraan dan Kebudayaan Melayu. Lehmann, Winfred P. 1990. Syntactic change. In Edgar C. Polomé (ed.), Research guide on language change, 365–388. Berlin: Mouton De Gruyter. Miestamo, Matti. 2005. Standard negation: The negation of declarative verbal main clauses in a typological perspective. Berlin: Mouton de Gruyter. Mufwene, Salikoko. 1989. On the so-called ‘infinitive’ in Atlantic creoles. Lingua 77. 297–330. Nordhoff, Sebastian. 2009. A grammar of upcountry Sri Lanka Malay. Amsterdam: University of Amsterdam Ph.D. dissertation. Nordhoff, Sebastian. 2012. Commentary on Smith’s papers. Journal of Language Contact 5. 73–79. Papiha, S.S., S.S. Mastana & R. Jayasekara. 1996. Genetic variation in Sri Lanka. Human Biology, 68, 707–737. Prince, Ellen F. 1998. The borrowing of meaning as a cause of internal syntactic change. In Monika S. Schmid, Jennifer R. Austin & Dieter Setin (eds.), Historical Linguistics 1997 (Current issues in linguistic theory), 339–362. Amsterdam: John Benjamins. Prince, Ellen F. 2001. Yiddish as a contact language. In Norval Smith & Tonjes Veenstra (eds.), Creolization and contact, 263–289. Amsterdam: John Benjamins. Slomanson, Peter. 2006. Sri Lankan Malay Grammars: Lankan or Malay? In Ana Deumert & Stephanie Durrleman (eds.), Structure and variation in contact languages, 135–158. Amsterdam: John Benjamins. Slomanson, Peter. 2008a. Progressive Aspect in Sri Lankan Malay and the Syntactic Status of ambe(l). Paper presented at the Twelfth International Symposium on Malay/Indonesian Linguistics, at the University of Leiden, The Netherlands. Slomanson, Peter. 2008b. The perfect construction and complexity drift in Sri Lankan Malay. In Peter Cole & Gabriella Hermon (eds.), Lingua vol. 118, 10, 1640–1655. Amsterdam: Elsevier. Slomanson, Peter. 2009. Morphological finiteness as increased complexity in negation systems. In Enoch Aboh & Norval Smith (eds.), Complex processes in new languages. Amsterdam: John Benjamins. Slomanson, Peter. 2010. Pragmatic accommodation as a motivation for typological shift. In Congressus XI Internationalis Fenno Ugristarums (FU11) Piliscsaba 2010, Pars V, 293–300. Budapest: Reguly Társaság. Slomanson, Peter. 2011. Dravidian features in the Sri Lankan Malay verb. In Claire Lefebvre (ed.), Creoles, their substrates, and language typology, 383–409. Amsterdam: John Benjamins. Slomanson, Peter. 2012. The Significance of Sri Lankan Malay and the Importance of Explicit Process Models. Paper presented at the 16th International Symposium on Malay and Indonesian Linguistics, University of Kelaniya, Sri Lanka. Slomanson, Peter. 2013a. Known, inferable, and discoverable in Sri Lankan Malay research. In Sebastian Nordhoff (ed.), Extreme language contact – The case of Sri Lanka Malay, 85–119. Amsterdam: Brill. 396 Peter Slomanson Slomanson, Peter. 2013b. Sri Lankan Malay. In Susanne Maria Michaelis, Philippe Maurer, Martin Haspelmath & Magnus Huber (eds.), The survey of pidgin & creole languages. Volume 3, contact languages based on languages from Africa, Asia, Australia, and the Americas, 77–85. Oxford: Oxford University Press. Smith, Ian, Scott Paauw & Bachamiya Abdul Hussainmiya. 2004. Sri Lanka Malay: The state of the art. In Rajendra Singh (ed.), Yearbook of South Asian Languages 2004, 197–216. Berlin/New York: Mouton de Gruyter. Smith, Ian & Scott Paauw. 2006. Sri Lanka Malay: Creole or Convert? In Ana Deumert & Stephanie Durrleman (eds.), Structure and variation in contact languages, 159–181. Amsterdam: John Benjamins. Smith, Ian. 2012a. Adstrate influence in Sri Lanka Malay: Definiteness, animacy and number in accusative case marking. Journal of Language Contact 5. 5–22. Smith, Ian. 2012b. Comments on Nordhoff’s “Establishing and Dating Sinhala Influence in Sri Lanka Malay”. Journal of Language Contact 5. 58–72. Weinreich, Uriel. 1963 [1957]. Languages in contact: Findings and problems. The Hague: Mouton. Copyright of Linguistic Review is the property of De Gruyter and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.