1. Introduction

1.1. English in Kenya and the International Corpus of English

English is not only the native language of millions of speakers in Great Britain, the United States of America, Canada, Australia and New Zealand, but also a second or additional language used both intranationally and internationally, for example, in India, the majority of countries in Africa, and Hong Kong. Platt/Weber/Ho (1984) classify the various functions in three broad categories: English as a native language (ENL), English as a second language (ESL) and English as a foreign language (EFL). Since then English as an international language (EIL) has been added as part of the latter category (Schmied 1991a: 33 ff.). These terms are used to refer to an individual's command of English but also to the position of the language in various countries. EIL is used to refer to the variety used by non-native speakers to communicate. In Kenya English is a second or additional language. It is prestigious, being the language of secondary and tertiary education, the High Court, Parliament and government institutions. The language is, of course, a remnant of Kenya's colonial past but English still remains of utmost importance in Kenya because of the status it bestows on its speakers and the fact that it fulfils the function of expressing part of a Kenyan's identity. It serves as a lingua franca between people of various ethnic groups, particularly in the bigger towns and the capital, Nairobi.
Kenyans first learn an indigenous or ethnic language, their mother tongue, in their homes and at primary school. In addition, Kiswahili may be learnt at school and a little later, English. Formal teaching in the classroom is, however, only part of the way in which English is acquired in ESL countries. It may also be spoken by some family members and among the peer group. In secondary education English is the medium used and the language acquisition process is continued outside the classroom when it is used to communicate. It is also passively acquired through reading the newspapers and watching television.
Whereas Kenyan school syllabuses base their principles of English language education on the British Standard, day-to-day confrontation with the language as it is spoken by fellow Kenyans leaves a considerable impression on the speakers and deviations¹ from the set norm are reinforced. Differences from British Standard English are most apparent in pronunciation (cf. Schmied 1991a: 57 f. for examples and a discussion of pronunciation features of English in Africa) but vocabulary, grammar and idiomaticity also show variation. Zuengler (1982: 115) determines "evidence of lexical, semantic, and syntactic nativization"². Lexemes borrowed from African languages are used to express concepts, particularly those referring to traditional cultural spheres, for which no equivalent can be satisfactorily found in British Standard English because English counterparts would lack the necessary culture-specific connotations. These "Africanisms"³ are integrated into the sentence structure as if they belonged to the language system (cf. Schmied 1991a: 76 ff. for more detail). Certain grammatical features deviate from the British norm, such as the use of the definite and indefinite articles, the expansion of the progressive form to stative verbs, non-count nouns in plural form, preposition usage. Kachru (1982: 46) also lists similar characteristics of grammatical deviation in "African varieties of English" and also includes in his description "a tendency to use complex sentences".
A detailed analysis of such linguistic features in Kenyan English novels of the late 1980s, followed by an attempt to find sustainable rules to explain specific deviation, was undertaken in an unpublished M.A. thesis (Hudson-Ettle 1990). Whereas innovation in lexical or idiomatic usage may signify a conscious act on the part of the author⁴ (and the vivid imagery certainly enriches the English language), grammatical deviation is more likely to be unconscious and less acceptable, constituting "errors" which derive from "inadequate proficiency, insufficient knowledge of subrules or from overgeneralisation or false analogy" (cf. Hudson-Ettle, Diana/ Josef Schmied 1992: 19). The analysis undertaken by the present author in 1990 examined, however, the language of one novel each of only six authors, and a wider representative corpus would have to be analysed systematically to discover whether features of usage are idiosyncratic or whether some occur frequently in other contexts. Corpus linguistic analysis could, for example, determine the extent to which progressive forms have been expanded to static verbs, find the verbs affected and compare the frequencies of usage.
Deviations are perpetuated in everyday usage and Kenyan English language teachers may not always be aware that what they are saying or writing is not British Standard (the common pick for the British pick up was not always recognised as deviant). Well established deviations are passed on from teacher to learner and in daily usage. The futility of such a situation underscores the need for clear guidelines, for the establishment of a norm for English in Kenya, for reference works such as grammars and dictionaries, which would include grammatical features and lexemes which have become part and parcel of the Kenyan use of English and would be specified as such, thus allowing the Kenyan student and teacher to check his/her usage. Where the language is intended for an international audience or readership, adaptations could then, if applicable, be made.
Interest in differing forms or varieties of English distributed all over the world has shown that there is a need for a large base of concrete data so that a clear picture of the state of the language can be obtained. Teams of linguists under the auspices of ICAME (International Computer Archive of Modern English) undertook the task of collecting spoken and written material from various English-speaking regions. Ultimately the ICE (International Corpus of English) will enable linguists to compare the language of texts from different categories with that of the accepted standards (British and American) as well as with the language of other varieties. The corpus will provide data for lexical, grammatical, syntactic and discourse studies and analyses and allow a detailed investigation of the features of the "New Englishes". It will be possible to locate those features which may vary from the Standards but occur in other Englishes and may therefore be more acceptable to, for example, the African teacher of English (Schmied 1990: 258).
This study is based on part of the Corpus of English in East Africa, which consists of data from Kenya and Tanzania. As collection and computerisation of the corpus material for Kenya had to be accomplished parallel to the present study, analysis of a ready-made and grammatically tagged databank by computer application alone was not possible. As it is, the object of the present study - grammatical subordination strategies in differing text types in the English spoken and written in Kenya - does not lend itself to a purely formal surface structure analysis based on the quantitative analysis of items such as lexemes.

1.2. Aims and objectives

The present study looks at an aspect of English in Kenya that has not before been investigated: the syntactic structures of specific text types. It is the aim of this study to investigate the form-function dichotomy as represented in the syntax of four text types from the Kenyan part of the Corpus of English in East Africa. It is assumed that variation in linguistic performance is determined by the choices made by the speakers/writers of the discourse and that these choices may be the result of a variety of influencing factors. Choices of syntactic strategy employed by ESL speakers/writers may differ from those preferred by ENL users but, not being obvious on the surface, these can only be investigated if grammatical descriptions of a large amount of comparable language data are available. Specific syntactic structures may be preferred in texts in specific categories⁵ (referred to here as text type), and different text types might reveal different syntactic patterns. Comparisons with previous studies on other varieties will be made and the quantitative analysis of syntactic strategies will show if and to what extent stylistic variation is restricted, as has been claimed for non-native varieties. The detailed description of the distribution of a particular structure such as, for example, the relative clause, will allow a comparison with its distribution in other varieties. This may lead "purists" to adapt the English taught to the norm by encouraging or discouraging the use of a particular type of relative clause, but others may decide that the feature is typical of non-native varieties and therefore acceptable.
The subordination strategies used in differing text types were investigated to determine which forms of grammatical subordination are employed in which text types. It is hypothesised that similarities and differences in syntactic strategies will be quantitatively and qualitatively established between the text types chosen for analysis.
It has been suggested that specific forms of grammatical subordination are used to serve certain discourse functions, such as descriptive, "packing" or persuasive functions (Biber 1988: 221ff). This analysis will examine the texts in the light of these suggestions to determine whether specific forms of grammatical subordination are used to serve certain discourse functions in differing text types of English as it is spoken or written in Kenya.
Allocation of texts to particular categories is determined by external criteria and only an analysis can reveal whether there is linguistic evidence to support the categorisation. It is hypothesised that certain texts within a category are similar to one another and different from other text types to such an extent that the categorisation can be termed linguistic, whereas in others the criteria are purely external. If it can be shown that significant differences in the use of syntactic strategies can be established quantitatively or qualitatively between the text types, then the external text type categorisation employed here is also linguistic.
The categories chosen for this investigation are freely spoken language, short stories (creative writing), newspaper personal columns, and institutional editorial articles, also from the press. Both the newspaper categories, the columns and the editorials, are part of the ICE Persuasive category, which will be discussed below. The former is, however, more personal and "creative", the latter institutional and formal. The hypothesis here is that a comparison of the former with the short stories could substantiate the claim to creativity on a linguistic basis. It is to be expected that the language of newspaper (personal) columns shows in some respects structural similarity to that of the creative short stories, but also on other counts to that of the other persuasive texts, the editorials.
The nature of language is such that no clear borderline is to be established in issues related to usage but rather that a continuum or gradient may be determined. It is hypothesised that an informal-formal continuum exists between the four text types - freely spoken language, short stories, newspaper columns, and institutional editorials - which can be seen in the significantly differing variation in the frequency of specific subordination strategies from one end of the scale (freely spoken language) to the other (institutional editorial articles).

1.3. Definitions of key terms

1.3.1. Subordination

Such a variety of clauses are subsumed under the term "subordination" that Thompson (1984: 86) cannot accept it as a grammatical category at all but rather as "at best a negative term which lumps together all deviations from some "main clause" norm".
According to Quirk/Greenbaum/Leech/Svartvik's A Comprehensive Grammar of the English Language (from now on designated CGEL) subordination is a feature of a complex sentence (§10.1 Note [a]). A simple sentence is an independent clause that does not have another clause as one of its elements, as subject, verb, object, complement or adverbial, but a clause may be part of one of its phrases. This would mean that postmodification of a noun in the form of a relative clause adds to the complexity of the noun phrase, not to the sentence. It is "only indirectly embedded within a larger clause" (CGEL §2.9: 44). Other grammars would consider such sentences complex as they contain more than one finite clause (cf. Beaman 1984: 55). On the other hand a simple sentence for Beaman (ibid: 54) - as her examples given below show - may in addition to a single finite clause (her underlining) also include a number of non-finite phrases (non-finite verb form now in italics):

In CGEL, however, non-finite constructions are treated as clauses because they can be analysed into the functional elements of a finite clause (§14.5). They are "intrinsically subordinate", so cannot be part of a simple sentence. Thus in the second example (ES12: 132) the two -ing constructions would be considered in CGEL to be supplementative adverbial participle clauses (§15.60), and in the third (EW12: 38) in order to rub his shin would be an adverbial infinitive clause of purpose. The first example would be analysed as containing a nominal -ing clause functioning as prepositional complement; it also includes an adverbial comment clause (I guess).
Subordination implies that more than one clause is involved. There is, however, some dispute as to whether certain constructions consist of more than one verb phrase. CGEL recognises a scale of verbs with auxiliaries at one end and main verbs at the other, whereby the former consists of one verb phrase, as for example she may come, and the latter two, she wants to come (§3.40). Along this gradient are central modals (e.g. can, could), marginal modals (e.g. dare, need, ought to), modal idioms (had better, would rather, be to, have got to) and semi-auxiliaries, used in CGEL (§3.47) to refer to expressions introduced by the primary verbs be or have, as for example be about to, be bound to, be going to, be likely to, be supposed to, be willing to, have to. These all relate to mode or aspect. Do well to [1.3.1.1] is an example of another modal idiom occurring more often in the corpus.

The final group along the gradient towards the main verbs are catenative constructions. Here definitions vary. CGEL uses the term catenative for such verbs as appear to, come to, fail to, get to, happen to, manage to, seem to, tend to, and turn out to, which are followed by an infinitive (§3.49), and start out, keep on and go on used in progressive constructions with the -ing form, and get with the past participle in a passive construction. These verbs express aspect or modality. It states explicitly that they are not syntactically related to transitive verb constructions such as expect (to), want (to) or attempt (to) because in contrast to catenatives, these may be followed by a direct or prepositional object. Begin, continue, cease and start are specifically not referred to as catenative verbs (§16.40).
Crystal (1991³: 50), on the other hand, defines the term catenative as referring to a "LEXICAL verb which governs the non-FINITE form of another lexical verb" giving as examples he likes to go, he wants to see, he hates waiting. For Gramley/Pätzold (1992: 168) catenatives are "those predicators which can have gerunds, infinitives, and nominal or interrogative clauses as complements." This would apply to "some 500 to 600 verbs".
This analysis takes only those constructions into account where the finite verb is followed by an object clause. Thus the complementation of central and marginal modals, modal idioms, semi-auxiliaries and catenatives, as defined in CGEL⁶, are taken as part of one clause and not regarded as subordinate.
The terms simple and complex sentences will be avoided here and Huddleston's brief and succinct definition of a subordinate clause as "one functioning as dependent within a larger construction that is itself a clause or a constituent of one" (Huddleston 1988: 152) will form a starting point for the analysis as it covers a whole range of dependencies, differing in form and syntactic function and serving various discourse needs, which will be described and discussed below.

1.3.2. Text type

The term "text type" is used here to refer to texts grouped together as of one kind according to external criteria. The major division in the ICE (International Corpus of English) is into spoken and written: Appendix 1 gives the text categories and their codes in detail. Written texts may be printed or non-printed; printed texts may, for example, be from newspapers or monographs. They may be editorials or instances of creative literature in the form of short stories or novels; they may constitute academic scientific writing, or academic papers on the arts. Scripted texts are written to be read and include, for example, news broadcasts. The printed written and scripted texts are intended to address a large audience. Spoken texts may be private or public, dialogue or monologue, as, for instance, day-to-day conversations, where only one person is addressed, broadcast discussions or talks, aimed at more recipients. These categories, among others, derive from discussions between ICE participants.
Four text types constitute the object of the present analysis - written creative (short stories), written persuasive press editorials, written persuasive personal columns and a spoken text type which is difficult to incorporate in the ICE categorisation. The spoken texts are monologues which originated in guided interviews, dialogue situations, where the interviewer(s) just "started the ball rolling" but, of course, took part in the communicative situation by means of paralinguistic responses, nodding, smiling, and occasional prodding.
There has been a lot of discussion (cf. Section 2.1. for a review of the relevant literature) as to the adequacy of such text classification and this work is an attempt at investigating the linguistic foundations of the specific text types chosen for analysis.

1.3.3. Function

As the stated aim of this study is to investigate the form-function dichotomy in the syntax of four text types taken from the Kenyan part of the Corpus of English in East Africa, it is essential to consider what is meant here by "function". This term is one of the most widely used in linguistics⁷ and in this study grammatical or syntactic function plays an important role when the various forms of subordinate clauses are discussed. Apart from the syntactic function, social and communicative functions are central to the present analysis. The social function of language is defined by Crystal (1991³: 146) as "the role language plays in the context of society or the individual", used to "communicate ideas, to express attitudes" and to "identify specific SOCIOLINGUISTIC situations, such as informality or intimacy, or VARIETIES of language such as science and law". Biber (1988: 33) speaks of "the desire to link the functions of particular linguistic features to variation in the communicative situation". Brown and Yule (1983) look at language from a functionalist perspective. For them discourse is "language in use" and

As such, it cannot be restricted to the description of linguistic forms independent of the purposes or functions which those forms are designed to serve in human affairs. (ibid: 1)
A speaker or writer of a language chooses the forms with which he or she wishes to fulfil his/her communicative intention. Sociolinguistic studies have shown that choice is determined by a number of factors of the speech situation. The lexical items selected, the choice of syntactic structures employed and variation in pronunciation are means by which specific purposes are served. Enkvist (1985: 11) states explicitly that the "syntactic form is governed by textual and discoursal forces" and that "linguists must learn to relate linguistic structures to functions of language in society" (ibid: 12). Major parameters for classifying domains of language use are according to Leech (1985: 40) those of formality, medium and communicative function. He writes:

It is a basic assumption of general stylistics that correlations can be established between situational domains ... and formal characteristics of language use. For example, in scientific English the passive voice occurs frequently, and the imperative mood rarely; whereas in advertising English, the passive is rare and the imperative is common.

Functions of discourse are descriptive, narrative, expository, argumentative, and instructive (cf. Section 2.1). Particular text types may be realisations of such discourse functions, as for example instructions for use of a public telephone, but a number of text types, such as novels, short stories, newspaper articles will make use of different functions at different stages.

The functions particular syntactic structures fulfil in stretches of speech or written language, in discourse, is an issue that has interested a number of linguists in recent years, among others Biber, who has developed a typology of English texts based on dimensions of linguistic features. This approach will be described, together with the research interests and findings of the other linguists who have provided the background to the present study, in the review of previous literature in the following section.