]> This ontology is implied and not formalized This document is a draft of a grammar of Saramaccan. There are two parallel streams of research which are going into its creation. The first is traditional descriptive research on the properties of Saramaccan grammar. The second is research into the digital representation of grammatical information via a structured markup language, specifically XML. The goal of this research is both to create a useful descriptive resource covering the Saramaccan language and to develop standards for a new generation of ``twenty-first century'' grammars. Traditionally, grammars have been published as books. However, in many respects, this has been an awkward format for them since the information contained in a grammar tends to be highly interrelated---something which is hard to convey via the linear organization inherent in any paper document. Accordingly, while the presentation format of this document that you are looking at now has many similarities to a book, it has actually been designed very differently from one. For example, its stored form contains descriptive prose, but no example sentences. These sentences are stored separately, in a resource containing annotated sentences from a corpus of Saramaccan sentences. Only references to the sentences appear in the grammar itself. These references are ``converted'' to full presentation sentences everytime the grammar is accessed. In addition, at certain points below, there are links allowing the user to see other sentences marked for a particular grammatical construction in the corpus. While these seemingly direct the user to a static document containing a list of sentences, this document is actually dynamically constructed, again, based on the annotated corpus. This means that, as the corpus expands, any new instances of some construction will automatically be presented when the link is selected. A further dimension of the design of this system is that gloss words in the texts and Saramaccan words referred to in the prose are not stored as simple text strings. Rather, they are stored as references to entries in a separate word list being constructed. Thus, the user can click on any one of these ``words'' and be directed to a word list entry. And, using the same basic architecture for corpus access found in the grammar, when a word list entry is accessed, it is possible to easily get a list of all the sentences in the corpus using some word in the dictionary. The greatest deficiency in this resource, at present, is the lack of information about Saramaccan itself. This does not represent a technical problem. Rather it derives from the fact that the labor-intensive task of text annotation is far from complete. Furthermore, while the collaborators on the grammar have gathered data on a range of grammatical phenomena in the language, they have yet to write up the relevant descriptive prose for them. As of now, the grammar contains two short chapters, one on causative constructions and one on relative clauses. In many cases, sentences in the corpus have not been fully annotated, therefore, they will appear with ``missing'' glosses if they are accessed via a corpus search. As work progresses more ``chapters'' will be added to grammar. In addition, it is anticipated that some of the technical aspects of the grammar's construction will be refined as more data is entered. The archival format of the grammar document, in XML format and containing an internal DTD can be accessed at http://2809d.linguistics.pitt.edu/Saramaccan/SMGrammar.xml and a paper describing the rationale behind the particular XML encoding employed can be found at http://emeld.org/workshop/2004/jcgood-paper.html. Some notes on transcription: in some instances, a word may appear with an acute mark, indicating tone, while in other uses it may appear without such a mark. This is due to the fact that, often, tone was not transcribed during elicitation. At this stage of preparation of the grammar, the transcription has not been normalized in any way, and the words in the example sentences are faithful to their original transcription. The sequence ``ε´'' is intended to represent an epsilon with an acute accent. The combination renders properly on some browsers but on most it does not. Hopefully, as new versions of browsers become available, this problem will fix itself. Please direct any questions or comments to Jeff Good (jcgood@pitt.edu). The standard formation of the causative construction in Saramaccan makes use of the verb in a way which is very similar to the English ``make...do'' construction as can be seen in the translation of the two example sentences below. However, there are also other auxiliary verbs which form causatives. For example, the verb can also give causative semantics to a sentence, as in the sentence below. The verb can also be used in causative contexts, as seen in the example below. While these latter two causative constructions are found in the language, they are far less common than causatives with , which represent by far the most productive causative construction in the language. Relative clauses are not as commonly used in Saramaccan as in English. Often, for example, Saramaccan will juxtapose two independent clauses in contexts where English uses a relative clause. Nevertheless, a wide range of relative clause constructions exist. A basic example of relative clauses can be seen in the sentence below. As is generally the case in relative clauses, the relative pronoun introduces the clause. Relativization can occur on a wide range of argument types. For example, indirect objects can be relativized upon, as seen below. Relativization can also occur on objects of prepositions---but in this case the preposition remains in the relative clause and a resumptive pronoun appears after it. In addition to relative clause which make use of the relative pronoun , similar to English, it is possible to form relative clauses on adverbial arguments using lexical items which are formally identical to question words as in the ``where''-relative below which makes use of the relative .