A descriptive grammar is, by definition, one based on naturally occurring linguistic data¹ and describing the usage producing them. The object of the present website, just as of most linguistic grammars, is the descriptive grammar. Its definition must, however, be complemented by some provisos.

Actual usage is full of variation. In principle, it is the linguist's task to identify the underlying principle of such variation instead of suppressing it. There are, however, exceptions to this principle:

Orthographic variation

The language description comprises a chapter on writing. Texts written in the object language usually display orthographical variation. It makes a difference whether the variant writing is found in earlier linguistic descriptions of the language or instead in written products of (non-specialist) members of the speech community.

• Variant writing in earlier descriptions of the language is (hopefully) not random, but the product of a difference in the underlying orthographical rules. It is one of the topics of the section dedicated to writing, since the user is to be supported in his task of using different descriptions of the same language.
• The latter kind of variation in writing is typically a consequence of insufficient standardization or insufficient familiarity with or adherence to the standard orthography. This is generally of no interest to the user of the language description, and the latter therefore suppresses such variation and is not descriptive in this respect [despite the terminological clash]. An exception may be provided by principled variation. This may occur if writers who do not follow the standard follow their own principle. To this extent, they behave like a specialized linguist who follows deviant principles, and such writing may reflect a feature of the language system. Consequently, their usage may be mentioned just as the former kind of variation.

If the language already has a standardized orthography which the present description adheres to, then the latter may be considered descriptive with respect to the orthography insofar as it describes the existent standard. Otherwise, the chapter on orthography is prescriptive.

Phonetic variation

Representation of language data may not reflect all kinds of phonetic variation occurring in them:

• Systematic phonetic variation is accounted for in the description of the phonology of the language. Those sections of the language description not concerned with phonetics and phonology are based on more abstract representations of the data. These may be phonological or even morphophonological or orthographic; and the orthography may, again, be based on a morphophonemic representation. As long as the description makes such reduction of variation explicit, it is not only admissible, but useful for most users. Reduction of this kind of variation is not actually prescriptive, but is rather a step in the abstraction which is necessary in descriptive work.
• Individual speakers may have peculiarities in their pronunciation. For instance, a speaker may have a laminal [s̻] where the phonetic system of the language provides for an apico-laminal [s]. In such cases, the majority pronunciation is the object of the description.

Performance defects

Natural texts are replete with speech errors, lapses, false starts, hesitations and so forth; oral texts more than written ones. While there are several branches of linguistics whose methodology depends precisely on such aspects of linguistic data, a language description is generally not interested in them. Editing such data is therefore commonly regarded as useful for the purpose of the description. Some principles are then to be observed:

• The raw data must be kept for counterchecking and for any user of the description who may need them for his purposes.
• The editing must be limited to performance defects; it must not pertain to features of the language system.
• The kinds of modification applied by the author of the description or text edition must be clearly identified.

¹ Naturally occurring data are data whose origin is independent from the researcher (Lehmann 2004).