Lexical Bias Effect

The lexical bias effect is the finding, drawn from studies of experimentally induced speech errors, that when a sound-exchange error such as a spoonerism is produced, the result is more likely to form two real words than would be expected if sound exchanges occurred randomly with no regard for whether the outcome happened to be an actual word. The psycholinguists Bernard Baars, Michael Motley, and Donald MacKay demonstrated the effect in the mid-1970s using a laboratory technique for inducing spoonerism-like errors under controlled conditions, finding that error-prone word pairs more often slipped into a real-word outcome than a nonsense one. The effect is generally interpreted as evidence for an internal editing or self-monitoring stage within speech production, a mechanism that reviews an utterance for well-formedness before or as it is spoken and is more likely to catch and suppress an error that would produce a meaningless nonword than one that happens to produce another real word, since a nonword error is more obviously wrong to the mechanism doing the monitoring. The lexical bias effect has been an influential piece of evidence in building interactive and cascading models of language production, in which information can flow both forward from meaning to sound and backward from sound back toward word-level representations during the assembly of an utterance.