The mental lexicon is a component of the human language faculty that contains information regarding the composition of words, such as their meanings, pronunciations, and syntactic characteristics. The mental lexicon is used in linguistics and psycholinguistics to refer to individual speakers' lexical, or word, representations. However, there is some disagreement as to the utility of the mental lexicon as a scientific construct.

The mental lexicon differs from the lexicon more generally in that it is not just a collection of words; instead, it deals with how those words are activated, stored, processed, and retrieved by each speaker/hearer. Furthermore, entries in the mental lexicon are interconnected with each other on various levels. An individual's mental lexicon changes and grows as new words are learned and is always developing, but there are several competing theories seeking to explain exactly how this occurs. Some theories about the mental lexicon include the spectrum theory, the dual-coding theory, Chomsky's nativist theory, as well as the semantic network theory. Neurologists and neurolinguists also study the areas of the brain involved in lexical representations. The following article addresses some of the physiological, social, and linguistic aspects of the mental lexicon.

Recent studies have also shown the possibility that the mental lexicon can shrink as an individual ages, limiting the number of words they can remember and learn. The development of a second mental lexicon (L2) in bilingual speakers has also emerged as a topic of interest, suggesting that a speaker's multiple languages are not stored together, but as separate entities that are actively chosen from in each linguistic situation.

Although the mental lexicon is often called a mental "dictionary", in actuality, research suggests that it differs greatly from a dictionary. For example, the mental lexicon is not organized alphabetically like a dictionary; rather, it seems to be organized by links between phonologically and semantically related lexical items. This is suggested by evidence of phenomena such as slips of the tongue, for instance replacing anecdote with antidote.

While dictionaries contain a fixed number of words to be counted and become outdated as language is continually changing, the mental lexicon consistently updates itself with new words and word meanings, while getting rid of old, unused words. The active nature of the mental lexicon makes any dictionary comparison unhelpful. Research is continuing to identify the exact way that words are linked and accessed. A common method to analyze these connections is through a lexical decision task, in which participants are required to respond as quickly and accurately as possible to a string of letters presented on a screen to say if the string is a non-word or a real word.

In the sample model of the mental lexicon pictured to the right, the mental lexicon is split into three parts under a hierarchical structure: the concept network (semantics), which is ranked above the lemma network (morphosyntax), which in turn is ranked above the phonological network. Working in tandem with the mental lexicon, in particular with the phonological network, is the mental syllabary, which is responsible for activating articulatory gestures in response to the phonological network. According to the theory which this diagram illustrates, different components both within and outside of the mental lexicon are linked together by neural activations called S-pointers, which form pathways together with large clusters of neurons called buffers (e.g. "concept production" and "word audio" in the diagram).

One theory about the mental lexicon states that it organizes our knowledge about words "in some sort of dictionary." Another states that the mental lexicon is "a collection of highly complex neural circuits". The latter, semantic network theory, proposes the idea of spreading activation, which is a hypothetical mental process that takes place when one of the nodes in the semantic network is activated, and proposes three ways this is done: priming effects, neighborhood effects, and frequency effects, which have all been studied in depth over the years.

In the spectrum theory, at one end "each phonological form is connected to one complex semantic representation", at the opposite end, homonyms and polysemes have their "own semantic representation[s]". The middle of the spectrum contains the theories that "suggest that related senses share a general or core semantic representation". The "dual coding theory (DCT)" contrasts multiple and common coding theories. DCT is "an internalized nonverbal system that directly represents the perceptual properties and affordances of nonverbal objects and events, and an internalized verbal system that deals directly with linguistic stimuli and responses". Others work around Chomsky's theory that "all syntactic and semantic features are included directly in the abstract mental representation of a lexical word".