Abstract
Newell’s challenge remains: experiments often study cognitive mechanisms separately, although real learning requires their coordination. We present CAVL, a functional, multilevel, precomputational architecture of socially embedded vigilant learning. It links working memory, long-term knowledge, expertise, epistemic vigilance, source representation, metacognition, and social context. Structural embedding shapes knowledge, opportunities, and verification costs. Interactional embedding shapes how learners evaluate claims and their sources. CAVL makes three testable claims. First, valid prior knowledge and processing conditions alter the balance between content quality and attributed authority. Second, private acceptance can diverge from public expression when social costs are high. Third, unequal cognitive resources and allocated epistemic credit jointly shape participation and later learning. Six propositions specify critical tests and reasons to revise the architecture. CAVL treats ecological validity as a problem for research, not as a label attached to realistic settings. It proposes controlled and systematically varied experiments that preserve relevant person–task–environment relations. Its purpose is to connect experimentally established mechanisms with socially situated learning, not to replace controlled experiments or existing theories.