In linguistics, the term lexis (from Ancient Greek: λέξις 'word') designates the complete set of all possible words in a language, or a particular subset of words that are grouped by some specific linguistic criteria. For example, the general term English lexis refers to all words of the English language, while more specific term English religious lexis refers to a particular subset within English lexis, encompassing only words that are semantically related to the religious sphere of life.

In systemic-functional linguistics, a lexis or lexical item is the way one calls a particular thing or a type of phenomenon. Since a lexis from a systemic-functional perspective is a way of calling, it can be realised by multiple grammatical words such as "The White House", "New York City" or "heart attack". Moreover, since a lexis is a way of calling, different words such as child, children, child's and children's may realise the same lexical item.

Lexical grouping may be:

A major area of study, psycholinguistics and neurolinguistics, involves the question of how words are retrieved from the mental lexical corpus in online language processing and production. For example, the cohort model seeks to describe lexical retrieval in terms of segment-by-segment activation of competing lexical entries.

In recent years, the compilation of language databases using real samples from speech and writing has enabled researchers to take a fresh look at the composition of languages. Among other things, statistical research methods offer reliable insight into the ways in which words interact. The most interesting findings have taken place in the dichotomy between language use (how language is used) and language usage (how language could be used).

Language use shows which occurrences of words and their partners are most probable. The major finding of this research is that language users rely to a very high extent on ready-made language "lexical chunks", which can be easily combined to form sentences. This eliminates the need for the speaker to analyse each sentence grammatically, yet deals with a situation effectively. Typical examples include "I see what you mean" or "Could you please hand me the..." or "Recent research shows that..."

Language usage, on the other hand, is what takes place when the ready-made chunks do not fulfill the speaker's immediate needs; in other words, a new sentence is about to be formed and must be analyzed for correctness. Grammar rules have been internalised by native speakers, allowing them to determine the viability of new sentences. Language usage might be defined as a fall-back position when all other options have been exhausted.

When analyzing the structure of language statistically, a useful place to start is with high frequency context words, or so-called Key Word in Context (KWICs). After millions of samples of spoken and written language have been stored in a database, these KWICs can be sorted and analyzed for their co-text, or words which commonly co-occur with them. Valuable principles with which KWICs can be analyzed include: