The preclinical development of drugs to treat the cognitive symptoms of neuropsychiatric and neurological disorders is a formidable challenge. Evidence from a wide range of preclinical behavioral and neuropharmacological tests has formed the basis for predicting drug-induced cognition enhancement in normal volunteers and in patients with cognitive impairments. However, the limited validity of preclinical predictions of this enhancement in humans indicates that conventional screening for "broadly active" compounds represents a below-optimal research strategy. This article conceptualizes the evidence needed to improve the predictive validity of preclinical research designed to discover and characterize cognition enhancers. We suggest that the investigation of reciprocal relationships among molecular, cellular, behavioral and cognitive processes modulated by candidate drugs represents the core of such research. By contrast, the usefulness of simple and high-throughput screening tests for the detection of cognition enhancers might be restricted to advanced drug-finding programs that are guided by evidence of the modulation of neurocognitive relationships by cognition enhancers and that are informed by iterative preclinical-clinical cross-validation of research approaches. We stress the need for basic biopsychological research approaches in preclinical programs to find and characterize drugs to treat cognitive disorders.