Lexical access retrieves word meanings and phonological forms from memory during comprehension and production. Eye-tracking and priming studies reveal that access begins immediately upon encountering partial word information and is initially automatic, with competing candidates activated in parallel.
Review eye-tracking experiments tracking lexical access during reading; conduct or study priming experiments showing automatic activation of semantic and phonological neighbors.
Lexical access is not conscious deliberate lookup but rapid automatic activation of competing candidates; later stages filter implausible meanings based on context.
From your study of psycholinguistics and the neuroscience of language, you know that the brain processes language rapidly and in real time — comprehension doesn't wait for a sentence to end before beginning. Lexical access is the subprocess of retrieval: given some acoustic or visual input, the cognitive system must locate the corresponding entry in the mental lexicon — the internal repository of words with their phonological forms, meanings, grammatical categories, and typical syntactic contexts. The central finding from decades of research is that this retrieval is not a deliberate lookup (like searching a dictionary) but a massively parallel, automatic process that begins with partial information and resolves competition over milliseconds.
The cohort model captures how spoken word recognition unfolds. When you hear "cap-", you automatically activate all words in your mental lexicon that begin with that sound: *cap*, *captain*, *capture*, *capsule*, and so on. This is the initial cohort. As more of the acoustic signal arrives, cohort members that no longer match are eliminated, until at some point — the uniqueness point — only one candidate remains and the word is identified. The cohort model predicts that words are recognized before their acoustic offset, which eye-tracking and priming experiments confirm: listeners can begin planning responses to a word before the speaker has finished producing it. For reading, the analogous process uses orthographic input rather than phonological, but the parallel activation logic is the same.
Priming experiments are the main empirical tool for studying what gets activated and when. In a semantic priming paradigm, seeing the word BREAD makes you faster to recognize BUTTER — because activating *bread* also spreads activation to semantically related nodes in the lexical network. In phonological priming, hearing "cat" facilitates recognition of "cap" — evidence that phonological neighbors are co-activated even when they are semantically unrelated. These effects occur at very short stimulus-onset asynchronies (sometimes under 100ms), confirming that the activation is automatic — it happens before conscious deliberation can intervene.
The key theoretical question is when and how context constrains this initially unconstrained activation. The modular view holds that initial lexical access is autonomous and context-free: the word BANK activates both its financial and riverbank meanings regardless of prior context, and only a later stage selects the contextually appropriate meaning. The interactive view holds that context can penetrate even early access, biasing which candidates are most strongly activated. Most current evidence favors a weak modularity: initial access is largely automatic and parallel, but context speeds selection so rapidly that only the appropriate meaning appears to reach awareness. This is why the misconception matters: readers who encounter an ambiguous word like *bank* in a disambiguating context may never consciously notice both meanings were activated, yet the transient activation of the inappropriate meaning has measurable effects on reaction time and eye movements.