Word recognition — the process of identifying a word from its visual or auditory form and accessing its stored representation in the mental lexicon — is the foundation of language comprehension. Skilled readers recognize words in as little as 200 ms, accessing meaning so rapidly that the process feels instantaneous. Understanding how this feat is accomplished has been a central goal of psycholinguistic and reading research for decades.
Visual Word Recognition
In reading, visual word recognition begins with extracting letter identities from the visual input. The dual-route model proposes two pathways: a lexical route that accesses stored word representations directly (efficient for familiar words) and a sublexical route that assembles pronunciation from letter-to-sound rules (necessary for novel words and nonwords). The connectionist triangle model (Seidenberg and McClelland, 1989) replaces these discrete routes with a network that maps between orthography, phonology, and semantics through learned statistical regularities.
Word Frequency and Context Effects
The word frequency effect is one of the most robust findings: common words are recognized faster than rare words. Neighborhood effects (words with many similar-looking neighbors may be recognized faster or slower, depending on the task) and morphological effects (complex words like "unhappiness" show evidence of decomposition) further reveal the structure of lexical access. Sentence context facilitates word recognition — predictable words are read faster and produce smaller N400 responses, reflecting predictive processing during reading.
McClelland and Rumelhart's (1981) interactive activation model proposes that word recognition involves cascading activation across three levels: features, letters, and words. Activation flows both upward (features activate letters, letters activate words) and downward (word-level activation feeds back to letter level). This feedback explains the word superiority effect — letters are recognized more accurately within words than in isolation — because word-level knowledge constrains letter-level processing.