Undeletions in Black South African English

In this paper, I propose some syntactic tendencies by which New Englishes can be better characterised from an internal and a comparative perspective, using Black South African English (BSAE) as a case study and focal point. Many descriptions of New Englishes have simply taken the form of lists of features. "Feature" in New English Studies (as in traditional dialectology) refers to a linguistic item that is characteristic of a particular second language (L2) variety, but not of the relevant superstrate, which is usually standard first language (L1) English of Britain or the USA. Sometimes a feature is characterised not in terms of pure absence versus presence, but in terms of relative frequency, markedness, or a change of function. Such inventories of features are a useful, and perhaps necessary, first step in dialect description. However, they are a long way off from being descriptively adequate, if one may apply Chomsky's (1986) term in an L2 context. For that, one would need to study an individual feature as exhaustively as possible in terms of its function and sociolinguistic distribution in the dialect concerned.


Introduction
In this paper, I propose some syntactic tendencies by which New Englishes can be better characterised from an internal and a comparative perspective, using Black South African English (BSAE) as a case study and focal point. Many descriptions of New Englishes have simply taken the form of lists of features. "Feature" in New English Studies (as in traditional dialectology) refers to a linguistic item that is characteristic of a particular second language (L2) variety, but not of the relevant superstrate, which is usually standard first language (L1) English of Britain or the USA. Sometimes a feature is characterised not in terms of pure absence versus presence, but in terms of relative frequency, markedness, or a change of function. Such inventories of features are a useful, and perhaps necessary, first step in dialect description. However, they are a long way off from being descriptively adequate, if one may apply Chomsky's (1986) term in an L2 context. For that, one would need to study an individual feature as exhaustively as possible in terms of its function and sociolinguistic distribution in the dialect concerned.
Since the first democratic elections in South Africa in 1994, BSAE has emerged from the shadows into the public arena, via the public broadcast media and high profile job opportunities in government and the private sectors. The period immediately following the transition can be characterised as one that engendered a "moral panic" (see Cameron 1995:82-97), evident in the claims about the desecration of English and loss of standards that found its doi: 10.5842/34-0-30 Rajend Mesthrie 76 expression in Letters to the Editor of many newspapers. Nowadays, opposition to the use of Black English in the media is more muted. L2 English accents are becoming more acceptable if one is to judge from the range of callers attracted to phone-in programmes, for instance. But a "complaint tradition" is still in existence in, for example, letters from L1 English speakers to the national radio service, SAFM. 2 Black English is becoming, in Le Page and Tabouret-Keller's (1985) terms, more "diffuse" than in the apartheid era, with the emergence of several new strands, reflecting a diversity of linguistic experiences amongst its speakers. Among these strands are the following: (i) the English of people returning from exile in different parts of the world; (ii) the rise of a new elite whose children share the L1 English norms of their peers in the now non-racial but English-dominant private schools; and (iii) the general increase of Black children at schools outside the townships.
All of these factors suggest that BSAE is likely to develop the range of accent and grammatical types that scholars have characterised as cultivated -general -broad (see Mitchell and Delbridge (1965) for Australia; Lass (2002) for South Africa). The rise of BSAE in the public sphere has been matched by an increase in academic studies since the 1990s. Its socio-educational context is considered by Buthelezi (1995), Wright (1996), and De Klerk and Gough (2002); its phonetics by Van Rooy and Van Huyssteen (2000), and Wissing (2002); and its syntax by , Mesthrie (1997), De Klerk (2003), etc. Gough (1996) provided a list of 23 of the main features of BSAE, which is useful as a quick overview of the variety. The list is repeated in De Klerk and Gough (2002:362-63).
In parts of sub-Saharan Africa, e.g., the use of can be able for can, and a predilection for left dislocation -see Bokamba (1982:83) for sub-Saharan Africa, Schmied (1991:72-73) for East Africa, and Mesthrie (1997) for South Africa. Such widespread similarities raise interesting questions regarding the role of a broad "living" Bantu substrate, complementary to, and integrated with, other language processes involved in L2 learning and stabilisation. The study that follows is therefore of considerable relevance to the study of English in sub-Saharan Africa in general.
Data for this paper comes from an ongoing project on BSAE initiated in 1992. The data base currently comprises 60 interviews undertaken under "Labovian" conditions, adapted as necessary for an L2 context. The interviews give evidence of a range of lects, from speakers fossilised at an early interlanguage stage to the highly proficient, virtually L1 variety of some university students. For the purposes of this paper, it is important to focus on a specific subgroup of 12 speakers who may be characterised as (mid-)mesolectal. These 12 speakers have the following characteristics: (i) they are all fluent in English; (ii) they function in English as their main academic language; (iii) they did not learn English as an L1; (iv) they did not use English in the home as children; (v) they are highly multilingual, often speaking about five other languages; (vi) they have a Southern Bantu language as home language: either an Nguni language like Xhosa (3) or Zulu (1), or a Sotho language like Tswana (4), South Sotho (1), or Pedi (3); (vii) they were interviewed while studying at an English-medium university; and (viii) they use English as a means of interaction with Black peers, but not exclusively. 3 However, these students can be differentiated from acrolectal BSAE speakers, in terms of accent and grammar. Typically, acrolectal speakers have been to multiracial or private schools from an early age, resulting in the acquisition of L2 norms that are quite similar to that of "cultivated" White South African English. 4 Studying the mesolect gives us the best clues about BSAE as a system, as it is not characterised by the lack of fluency one often finds with doi: 10.5842/34-0-30 basilang speakers; it is not a direct "inheritance" of the target language (TL); nor is it yet swamped by "interference" from the standard. See Rickford (1974) for the insights provided by the mesolect for Creole studies, and for the Guyanese Creole continuum in particular.

The notion of an 'antideletion'
I claim that -far from being a problematic and error-ridden dialect with a miscellany of nonstandardisms -from the upper mesolect and beyond, BSAE can be seen as a coherent system, whose differences from what have come to be the standard systems of English can be characterised from the vantage point of deletion processes that are commonly assumed in generative analyses of English, or are reported in the dialectological literature. For want of a better word, I coin the term "antideletion" for this phenomenon, which can be refined into the following three types: (i) Type A (Undeletion Proper): This type restores an element that is often assumed to be deleted or to have an empty node in generative analyses of English.
(ii) Type B (Non-deletion): This type shows the presence of a feature of standard English that is deleted in some (L1) dialects of English.
(iii) Type C (Insertion, the opposite of deletion): This type inserts elements that are not found in the underlying structure of standard English.
In section 4, I present seven proper undeletions of type A, whilst types B and C are discussed in Mesthrie (2006). The syntactic framework I use is an eclectic "descriptive syntax" one, which draws on insights from Chomskyan and other branches of linguistics (e.g., typology), without committing itself to a particular version of a particular theory. This approach is best exemplified by the landmark 1842-page Cambridge Grammar of the English Language (Huddleston and Pullum 2002), which presents findings based on over 50 years of syntactic theory relating to English in ways that are accessible outside the generative paradigm (e.g., to applied linguists). Like Huddleston and Pullum, I find the notion of 'underlying structure' to be a useful one for descriptive purposes.

An introductory comparison with other varieties of South African English
Ultimately, this paper is concerned with characterising BSAE in its own terms.

81
Colloquial WSAE, on the other hand, is also a source of input in the workplace and socially.
However, social contacts between Black and White South Africans were still limited in the early to mid-1990s when the data was collected. Descriptions of colloquial WSAE syntax can be found in Branford (1991), Lass (2002), and Bowerman (2004). Other varieties of South African English which themselves started out as L2 varieties are also relevant, e.g., Afrikaans English (Watermeyer 1993), Cape Flats English (Malan 1996), and South African Indian English (Mesthrie 1992). No corpus of South African English is yet available that will allow the use of these varieties as control groups to verify that the properties of BSAE described in this paper are indeed found to a statistically significant degree in this variety alone. However, the descriptions that are available in the sources cited are sufficient insofar as they are either interview-based descriptions of individual varieties (Mesthrie 1992;Watermeyer 1993) or careful summaries of a range of studies (Malan 1996;Bowerman 2004). None of these studies point to a tendency to deletion in the varieties described.
For example, Branford (1991:223-224) devotes the equivalent of a whole page of her Dictionary of South African English to an entry called "Omissions". Apart from an entry on "articles", it is the only such general grammatical entry in the work. She lists ten such omissions covering a variety of categories, as shown in table 2.

Undeletions proper
In this section, I present seven instances of undeletion in BSAE, namely that of the complementiser that, infinitival marker to, pronoun (including resumptive pronouns, left dislocation, and dummy it), to be in small clauses, and the occasional undeletions with idiomatic whconstructions.

Complementiser that
The constraints on the use, variability, and deletion of that in std Eng are reasonably clear. It is variable in unmarked complements with verbs like say as in (1) and (2); mandatory in extraposition as in (3); and categorically deleted in direct speech as in (4) and (5): (1) She said that she'd go for a walk.
(2) She said she'd go for a walk.
(3) That she'd go for a walk was clear to us all. (*Φ for that) (4) She said, "I'll go for a walk". (*that for Φ) (5) She thought, "He's not a bad chap". (*that for Φ) In std Eng, the difference between direct and indirect speech is signalled by deictic changes to pronouns and tense markers within the quoted clause, as can be seen by comparing examples (2) and (4) (6) to (8): (6) So she was warning us that, "You'd better learn this language because, like, you're going to Cape Town".
(7) They'll just tell you that, "We have been using Fanakalo".
(8) They announced that, "We are going to do all courses in Afrikaans".
The quotative nature of the subordinate clauses is also signalled by falling intonation accompanying that. The use of that in sentences like (6) to (8) is not mandatory, and cooccurs with the standard use of zero with quotative verbs, as illustrated in (9) -(10).
(9) A lot of people will usually say, "What does it mean?" (11) I just said that, "No, I cannot do what everyone is doing".
(12) You can hear that, "No, they are using a different language now".
(13) They want to show us that, no, they can speak English.
(14) We just saw that "Oh, we are at home now".
In (11) and (12), no is a discourse marker, equivalent to something like in fact. Sentence (13), whose subordinate clause is not expressed in direct speech, shows this discourse marking function clearly. This counts as "semi-indirect" style. These instances of no and oh do not alter the analysis of direct speech; they are noted here simply because they appear fairly frequently. Anthea Fraser Gupta (personal communication 2005) points out that these are also not uncommon in L1 varieties of English. Table 3 gives the statistics of the use of that in direct speech amongst the 12 mesolectal speakers studied. "Deleted that" refers to the absence doi: 10.5842/34-0-30
I now turn to other instances of undeleted that. In (15) below, that makes an appearance in clefted whconstructions, contrary to TL norms. The equivalent in std Eng to (15) has Φ in place of that as in (16). Note that (17), the unclefted equivalent of (16) in the TL, allows that or zero. Likewise, (18) from BSAE shows wh + that, this time in a combination of indirect marker when + complementiser that. However, the std Eng equivalent, given as (19), disallows that after when.  (20). In this sentence, that is not an ad hoc interpolation determined by maybe; it is a regular pattern that turns up in formal styles.
Sentence (21) is taken from a postgraduate student's essay, while (22) is from a radio interview. The structure of the standard equivalent of (22) can be represented as in (23).
(20) As you might have heard, maybe, that women were quite restricted.
(21) As it is mentioned above that one of the speakers said that she sometimes gets tired of speaking in English … (22) As you know that they are from the Ciskei. (that for Φ) Siegel 2001). At the next stage of selection, such competing forms are narrowed down according to certain typological principles, if the TL is not readily available. Mufwene (2001) argues that such influences are driven by the language ecology of the variety being established. If the pull of the TL is strong (e.g., it is readily available in the classroom and outside), then selection turns out to be largely a matter of "weeding out" the non-TL forms from the pool of variants. I propose that mesolectal BSAE shows the interplay of several phenomena, namely: (i) substrate influence from the Nguni and Sotho languages, which I explore more fully in work in progress; (ii) discourse effects in common with colloquial English; (iii) typological regularisation of the pool of variants; and (iv) the pull of TL grammar as the influence of education becomes more forceful in the experience of present-day mesolectal speakers. 6

Infinitival marker to
A fairly similar type of analysis holds for infinitival to, which is mandatory after most verbs that select infinitive clauses as complements, as in (24). On the other hand, there is a small set of "bare infinitivals" (let, make, have) which behave in the opposite fashion, as shown in (25) to (27) However, Φ rather than to is mandatory in BSAE mesolect in the first person forms let me go and let's go. There is a third set exemplified by help, which allows to or Φ in std Eng, as shown in (32). This variability is also found in my data base, as examplified by sentences (33) and (34).
(32) She helped me (to) find my rabbit.
(33) I think it would help me Φ write better.
(34) I had to leave it with somebody to help me to come to UCT.
Sentence (35) shows that negative sentences also permit undeletion.
(35) So that's what makes one not to know which language to speak.
As far as std Eng is concerned, such use of to in negative constructions would, in my judgement, be less marked than in positive sentences. Once again, standard zero forms instead of undeleted to also occur in BSAE, as in (36) and (37).
(36) You must let your child Φ speak English.
(37) My dad used to make me Φ read the newspapers.
The undeletion of to is noticeable with the form to be, as shown in (38) and (39).
(38) Treat that person as a person and maybe pointing out things that can make that person to be the character that he is … (39) … and it challenges me or it makes me to be challenged … 9 Std Eng here allows either Φ or be but not *to be. leaving, and I felt it crawling is to be followed up in future work. There is at least one instance of the substitution of infinitival V + -ing of the TL by to + V with an admittedly nonsensory verb in my data base: We didn't even mind to watch at night. Table 4 gives the statistics for the occurrence of to versus zero. Excluded from the count are first person subjunctive expressions let me/us V, which occurs invariantly in the mesolect.
Thus, for the purposes of this analysis, phrases like let me go, let's say, and let me say can be considered idiomatic invariant forms, rather than ones involved in syntactic/parametric variation.  Rather similar to the behaviour of that and to discussed above are undeletions involving pronouns. The most obvious of these is the use of resumptive pronouns, though other retentions are also prominent.

Resumptive pronouns
This feature has been identified as a prominent feature of BSAE; Gough (1996:61) (42) show the retention of a resumptive (or shadow) pronoun in the relative clause. However, such examples are infrequent and probably subject to a strong grammatical constraint. That is, resumptive pronouns do not occur as subjects within the relative clause; the relevant statistic in my data is 67 zero subject relatives versus 0 resumptive subject pronouns. This is also true of the data provided by Bokamba (1982:83) and Gough (1996:61).
The absence of resumptive pronouns in subject relatives may well be a sub-Saharan L2 English "universal". In oblique position (usually direct object and genitive), resumptive pronouns are an option, as table 5 shows. rrences of zero in oblique position .4 The constraint against resumptive pronouns in subject relative position requires explanation, as it goes against the general undeletion habit of BSAE. It is probably related to the frequency of subject copy pronouns in main clauses, to which I now turn.

Left dislocation
Resumptive pronouns are closely related to other forms like appositional pronouns occurring after a relative clause and pronouns in left dislocation constructions. Some writers on BSAE have erroneously, though perhaps understandably, referred to these appositional pronouns as "resumptive pronouns". Jespersen (1927:72) had noted that a relative clause could be "resumed later" by a personal or demonstrative pronoun, in long sentences in which the speaker has failed to keep track of the initial NP.
The following examples are from my data base.
(43) The last time I was in Natal, it was in 1981.
(44) The people who are essentially born in Soweto, they can speak Tsotsi.
(45) Yes, most of them, I call them confused scholars.
These are left dislocations with complex NPs, including those with relative clauses as in (43) and (44). Left dislocation involves the use of an appositional (or copy) pronoun, which, unlike its resumptive counterpart, occurs in the main clause. 11 And, unlike their resumptive counterparts in BSAE, appositional pronouns may occupy subject position; in fact, they frequently do. Mesthrie (1997)  According to Prince (1981), left dislocation, as in (46), involves a fronted NP with a copy pronoun in the main clause. Left dislocation is used to reintroduce information that has not been talked about for some time, or for contrastive purposes, when speakers go through lists and make comments about each individual element in the list. Fronting, on the other hand, puts old information first (in topic position). This topic must be already evoked in the doi: 10.5842/34-0-30 discourse or stand in a salient set relation to something already in the discourse (Prince 1981).
Thus, Zulu in (47) has already been evoked in the question and is therefore fronted in the interviewee's response. In the focus movement construction in (48)

Dummy it
Dummy or pleonastic it occurs in std Eng sentences like (49) and (50). In adjunct comparative clauses like (51), however, dummy it does not surface in std Eng.
(50) It can be said that children are highly adaptable.
(51) As Φ can be seen from recent statistics, the birth rate is declining.
BSAE mesolect notably favours the retention of it in constructions exemplified by (51). This construction belongs to a formal "expository" style and is not common in casual speech.
Dummy it does surface regularly in BSAE university students' essays, as exemplified by (21) above and (52) below, which pertains to the main clause in a for … to construction. However, examples can be found in speech too. Sentences (53) and (54) were taken from national radio, whilst (55) is from my data base.
(52) For her to use the word 'shame', it doesn't mean that there is no other word in Zulu.
(53) As I made it clear before, I am going to talk about solutions, not problems.
(54) As it is the case elsewhere in Africa, much can still be done for children. (std Eng: As is the case ...) (55) Take it for example a person who's North Sotho … Potentially, BSAE restores dummy it in the whole range of expressions like as is widely known, as happens frequently, as will be obvious, etc.

To be in small clauses
In the syntax literature, the term "small clause" is used for structures exemplified by (56) (Radford 1988:324 (58) and (59)  Given the properties discussed in this section, the following principle can be enunciated:

Principle 1
If a grammatical feature can be deleted in std Eng, it can be undeleted in BSAE mesolect.
However, since all the tables indicate that even in BSAE such undeletions are not mandatory, the following corollary is necessary: 13 doi: 10.5842/34-0-30

Corollary 1
If a grammatical feature can be deleted in std Eng, it can also be (variably) deleted in BSAE mesolect, at a lower rate of frequency. 14

Related clusters of features in BSAE
In Mesthrie (2006), I examine two related clusters of features in BSAE mesolect. The first concerns a set of linguistic items which are not thought of as features, since they accord with the grammar of std Eng. However, they are well-known in the syntax literature, since some non-standard varieties do indeed delete them variably. I argue, therefore, that it is typologically significant that BSAE does not, for example, generally delete the copula, whereas varieties like African American Vernacular English (Labov 1972) and Singapore English (Ho and Platt 1993) often do. The second set discussed in Mesthrie (2006) concerns additions rather than "undeletions". These are features which frequently insert an additional element into the syntax, contrary to the patterns of std Eng. An example would be the use of cross-clausal conjunctions as in (62).
(62) Although I'm not that shy, but it's hard for me to make friends.

Conclusion
I have suggested that, far from being an unstable L2 variety (as error analyses have implied), mesolectal BSAE is quite regular. It incorporates a subystem which is typologically consistent as an undeleting version of the TL (std Eng). The statistics provided show that undeletions do not occur categorically for any construction; rather there is variability. This variability seems to encompass the principles outlined by sociolinguists (Labov 1972;Chambers 2003) for L1 variation. In this instance, though, the origins of variability lie in the interplay between an undeleting system and a more standard system which speakers have to pay attention to, since they use their L2 more in educational and formal settings than in casual in-group speech.
The concept of 'undeletion' will, I believe, help characterise other New Englishes too, especially other varieties in Africa. 15 Its origins must lie in the nature of the substrata, in which deletion and movement rules are rare. The details are beyond the scope of this paper (see Van  16. Similarly, it would seem to me that the deleting tendencies of Singapore English are due to the majority substrate language, Chinese.