Descriptive Linguistics is the work of analyzing and describing the actual language spoken now, or in the past, by any group of people. Accurate description of real speech is a very difficult problem and linguists have often been reduced to very inaccurate approximations.
Almost all linguistic theory had its origin in practical problems of descriptive linguistics. Phonetics (and its theoretical developments such as phonemes) has dealt with how to pronounce languages. Syntax has developed to describe what is going on once phonetics has reduced spoken language to a control level. Lexicography collects "words" and has not given rise to much theory.
An extreme mentalist viewpoint appears to deny that the linguistic description of a language can be done by anyone but a competent speaker. Such a speaker has internalized something called "linguistic competence" which gives them the ability to correctly extrapolate from their experience to new but correct expressions and to reject unacceptable expressions. Be that as it may be there are practical immediate needs for linguistic descriptions and we cannot wait for a full exploration of linguistic competence.
There are tens of thousands of linguistic descriptions of thousands of languages that were prepared by people without adequate linguistic training. With a few honorable exceptions all linguistic descriptions done before, say, 1900, are amateur productions.
A linguistic description would currently be considered good if it:
- described the phonology of the language and established a practical orthography.
- described the morphology of words.
- described the syntax of sentences.
- described the lexical derivations.
- included a vocabulary with at least a thousand entries.
- included a few genuine texts.
The current controversial topics are usually morphology and syntax. For many years too much attention given was to English, which has a very meager morphology, over-emphasized syntax, but now morphology has revived as an active field of study.
The purpose of linguistic theory, so far as a practical linguist is concerned, is to make descriptions of morphology and syntax comprehensible. It is easy to see that the same data can often be described in different ways. For a while there was an active desire to find some measure which would allow some one description to be called the best. Today that goal seems to have been given up as chimerical.