In Chomsky's transformational generative grammar theory, there are four most important types of linguistic knowledge. Though the theory and nomenclature has changed for some concepts as Chomsky has refined his theory over time, the underlying concepts remain constant. The first two of the most important types of linguistic knowledge are that of performance and competence.
Linguistic competence is the deeper concept of the two and has to do with a native speakers' comprehension of the "ideal" language system shared among a community of native speakers. In other words, while this ideal language includes syntax, phonetics and phonemics, morphology and semantics, it is not related to nor affected by factors such as errors, memory, distractions, etc. These nonaffecting factors are grammatically irrelevant and bear no relationship to competence in knowing the language ideal and are further irrelevant to forming a theory of a community's shared knowledge of their ideal language structure.
Linguistic performance is the practical versus the theoretical representedby linguistic competence. While competence can only--at this point--be a theoretical construct, performance is the factual enactment of linguistic utterances. Performance also includes syntax, phonetics and phonemics, morphology and semantics, as does competence, but, in this instance, performance relates to these in terms of practical actualizations of spoken and/or written utterances, not in terms of theoretical ideal language structure. Whereas competence does not include situational factors, performance does include situational factors such as errors, both habitual and situational; memory loss or limitation; general distractions; level of education; clumsy speech construction; etc.
Actual speech behavior, speech performance, for him is only the top of a large iceberg of linguistic competence distorted in its shape by many factors irrelevant to linguistics. (John R. Searle, 1972).
Two other concepts, the terminology and theoretic understanding of which have changed somewhat as Chomsky's research has broadened the definitions of transformational generative grammar theory, are deep structure and surface structure (D-structure and S-structure). These correspond in general concept with competence and performance in that competence and D-structure have to do with innate aspects of linguistic operation and performance and S-structure have to do with variable individual or group production of linguistic operation. While competence describes the theoretical ideal of a native speaker's language knowledge, D-structure describes the universal grammar underlying every language--a grammar that has universally applicable characteristics and is the foundation for every language of the world. Similarly while performance relates to how an individual speaker may utter the language, with errors, dialectical variation or slang etc., S-structure describes the language systems that are built upon the universal grammar as theorized by the D-structure. Examples of S-structure are Japanese, Indian English, and Portuguese.
According to Yergin, the surface structure "'faces out' on the world and, by certain phonological rules, is converted into the sounds we hear; it corresponds to the parsing of sentences which we all learned from our indefatigable junior high English teachers. (Major Twentieth Century Writers, 1991)