The COVIS model, developed by F. Gregory Ashby and colleagues (1998), proposes that humans have two qualitatively different category learning systems that compete to control behavior. The explicit (verbal) system learns rule-based categories through hypothesis testing and is mediated by prefrontal cortex. The implicit (procedural) system learns information-integration categories through slow, incremental association learning and is mediated by the tail of the caudate nucleus.
Two Systems
→ PFC, head of caudate, anterior cingulate
→ Optimal for: rule-based categories
Implicit system: procedural learning, stimulus-response associations
→ Tail of caudate, putamen, premotor cortex
→ Optimal for: information-integration categories
Evidence for Two Systems
Multiple experimental dissociations support COVIS: (1) feedback timing — the implicit system requires immediate feedback, while the verbal system tolerates delays; (2) cognitive load — working memory load impairs rule-based but not information-integration learning; (3) switching costs — switching between verbal and implicit strategies within a session produces interference; (4) neuropsychological evidence — Parkinson's disease impairs implicit category learning while sparing rule-based learning.